“Because economists can’t carry out experiments, economics is not a science. Discuss.” A classic exam question, but the premise is false: economists have been experimenting for decades. In the 1970s, John Kagel and Raymond Battalio tested the economic rationality of animals. For example, they plied rats with root beer (yum) and tonic water (yuck), varying the effort required to access either drink to see how they coped with changing prices. The rats turned out to be perfect rational consumers and could even manage a budget, which is more than can be said for most economics students.

Undeterred, Vernon Smith experimented on students, giving them fictitious commodities, and looking for evidence that they were able to trade and converge to the market prices predicted by theory. He shared the 2002 Nobel economics prize with psychologist Daniel Kahneman, who showed through experiments the ways in which we depart from the axioms of microeconomic theory.

This sort of research was for low stakes (coffee mugs, chocolate bars) and in odd, artificial conditions. Perhaps people behave differently outside the laboratory. So economists such as John List have recently tried to conduct experiments in more natural settings. More than once, List has found behaviour “in the wild” that looks much closer to the prediction of economic theory.

Peter Riach and Judith Rich are economists who use randomised trials to sniff out discrimination. They send out large numbers of fake job applications and randomise whether a man’s name or a woman’s name goes on each one. Testing for racial discrimination means using a “white” name or a Vietnamese name. Since the applications are of comparable quality and the names are assigned at random, if employers respond differently to the names, discrimination is almost certainly the explanation.

You might be wondering why anyone would suggest that economic experiments are impossible. Yet sceptics have a point. It is hard to set up experiments that have really high stakes, and it is even harder to set up experiments that test complex questions. It is useful to ask whether some people are sexist or respond irrationally to risk; it is quite another thing to ask what the effect of a change in the money supply will be.

But a small group of researchers now believes that a new experimental tool is at hand. One of them, the economist Edward Castronova, calls it “a social-science supercollider”. He is talking about the new breed of online computer games, in which players log into a fantasy universe and control characters who trade, argue, fight or, depending on the world, have sex with each other.

That sounds a little crazy, but in many ways this is a more realistic setting than Vernon Smith’s lab. Players invest a lot of energy into getting what they want. Real people work hard to produce things that have real value to themselves or to others. Make no mistake, some games support real economies inside make-believe universes.

Experiments are easy because virtual worlds exist in parallel. For technological reasons, a popular game might run 10 or 20 subtly different versions on separate servers. The researcher can adjust the money supply or the system of property rights in some of the worlds and observe the differences.

Castronova has secured funding to construct a game world where he can put his ideas to the test. Not everyone is convinced that real insights will emerge, but at least this 21st-century macroeconomic experiment will be less damaging than the 20th-century version. That was to build a wall across Germany or Korea, make markets illegal on one side, and see what happened.