Trump’s Trade Tariffs and Game Theory

Trump’s Trade Tariffs and Game Theory On April 2, 2025, President Donald Trump announced, on his so-called “liberation day”, a 10% universal tariff on all imports (except for Mexico and Canada). He also announced so‐called reciprocal tariffs, which would retaliate against other countries for alleged unfair trade practices (which were mostly not real). From April 9, the EU would face 20% tariffs, China an additional 34% tariffs, and some other countries would face a 50% tariff on all imports. Global markets reacted instantly: the S&P 500 slipped 12%, while the Euro Stoxx 50 lost 13%. Early estimates from JP Morgan suggested to a 0,7 percentage point hit to world GDP growth in 2025, and Bruegel (a European think tank) a 0,3 percentage point hit to European GDP. Goldman Sachs believed it was more likely than not that the US would enter a recession. After a lot of pressure President Trump decided to delay the “reciprocal tariffs” for 90 days; however, he left the 10% universal tariffs in place.

Responding to US Tariffs
For world leaders, the main question is how to respond to these tariffs, when they might be reinstated in a few months. Several articles have urged their countries to not retaliate and sit
the trade war out, arguing that counter‐tariffs would damage countries’ economies more thanwhen they don’t retaliate (Swank & Van der Windt 2025; McKibbin & Noland 2025). These articles consider a trade war a single non-repeated game as could be considered in Game Theory. However, I believe that this perspective is flawed, and when considering a repeated‐game perspective, and when considering the flexibility governments enjoy in
designing counter‐measures, governments should adhere to different advice.

Both Swank & Van der Windt (2025) and McKibbin & Noland (2025) model using
Econometric models the outcome of a trade war, with different responses for each country. Swank & Van der Windt considers two main options: no retaliation to 25% tariffs from the US, or respond with a general 25% tariff to all US imports. In the Table below you can see
the main consequences of their either retaliating or not to Dutch GDP.

The authors conclude based on this modelling, that not retaliating should be preferred by the Dutch government, as this would limit the drop (or more accurately, lower expected growth)
of GDP experienced over the next ten years. Interestingly, this model also suggests that by implementing these tariffs, the US hurts its own economy more than the Dutch economy.
The other article as written by McKibbin & Noland (2025), similarly suggests that Mexico and Canada should not retaliate to US tariffs; however it suggests that the Canadian and Mexican economies will suffer more than the US.

Flawed conclusion
I believe that both articles are flawed in their conclusion, since they consider this trade war a single non-repeated game. However, tariffs can be raised, rescinded or tweaked on any day.

President Trump has already modified some of the tariffs he introduced, and he will likely grant exemptions for categories of imports that cannot easily be replaced (or are needed by some of his supporters). That transforms the interaction into an indefinitely repeated game (with discounting and a continuous action space), where reputational incentives, and not just one‐period pay‐offs, drive behaviour and where specific actions can be chosen.

Repeated Games and Robert Axelrod
In 1980, political scientist Robert Axelrod organised two computer tournaments in which different strategies for an iterated Prisoner’s Dilemma¹ competed with each other, where every strategy was aware of the decisions made by the same opponent in previous games.
Surprisingly, this tournament found that the best strategy was the simple Tit‐for‐Tat (TFT) algorithm, which cooperates on the first move, and then copies the opponent’s previous move, winning both events.

Axelrod distilled four properties of the most successful strategies, described in his book about these tournaments the four main components of good strategies. Winning strategies should be: 1) nice, they should never be the first one to refuse cooperation; 2) retaliatory, they should punish defection immediately; 3) forgiving, thus returning to cooperation once the opponent does; and finally 4) clear, they should behave in a way that is easy to understand. (If you want to learn more about this very interesting experiment, you could watch this very good YouTube video by Veritasium.)

We could consider these results of this experiment also to trade wars, as implementing trade barriers could easily be considered a repeated game. This would suggest that countries should only retaliate with trade barriers only if other countries raise them. Clearly, Donald
Trump is not following this advice of Axelrod, which would result in the best outcome for his country. However, that does not matter, as the TFT strategy also performed reasonably well against so-called “bad” players that did not prioritize cooperation, meaning that government
leaders should retaliate using some sort of measured, credible retaliation.

Measured and Smart Retaliation
The two articles mentioned in the introduction only consider retaliation by governments using exactly the same general tariffs that the US government implemented. This is a form of retaliation that could be chosen by some governments, for example China retaliated using this exact strategy.

However, governments have more flexibility when imposing retaliatory tariffs. For example the European Commission chose a more surgical response, by introducing tariffs on just a part of US imports, the negative impacts of implementing tariffs could be limited. Additionally, by targeting specific parts of the United States, these tariffs were designed to maximize political pain, for example by hitting products produced in US swing states, it would incentivize important Republican allies to pressure President Trump to start negotiating with the Europeans. Finally, by stating in advance when the retaliatory tariffs would be implemented (15 Apr, 16 May, 1 Dec), it would incentivize the Americans to start to negotiate early and quickly. This fulfils the important criteria: retaliation is retaliatory (politically costly for the US), forgiving (tariffs can be lifted instantly), and clear (publicly communicated lists
and timelines).

Conclusion
The Smoot‐Hawley tariffs, introduced in 1930, have shown that prolonged tariff battles rarely deliver economic prosperity, often precipitate recessions and incentivize a less efficient global economy. Repeated‐game logic offers a roadmap out: retaliate just enough to deter ongoing defection, but keep the door open for a mutually beneficial climb‐down.

With smart EU countermeasures on the table and both sides signalling willingness to talk, there is enough room to restore tariff-free trade before permanent damage is done.
However, there should be at least some willingness for governments to respond with smart countermeasures, and the general public deserves better articles which incorporate the full breath of economic and game theory, instead of just a simplistic view.

¹ For those that are unaware of the Prisoner's Dilemma: the Prisoner’s Dilemma is the idea that if two suspects were caught by the police, they can either cooperate with each other or defect by betraying the other. If both cooperate, they each get 1 year in prison; if one defects and the other cooperates, the defector goes free and the cooperator gets 5 years; but if both defect, they each get 3 years. The surprising conclusion is that even though mutual cooperation gives a better outcome, both suspects are likely to defect because it seems safer for each individually, leading to a worse result for both.