Testing Business Decisions in the Lab
by Anjali Vaidya
Ronald Reagan famously called economics Òthe dismal scienceÓ, alluding to a title bestowed on the field after economist Thomas Malthus grimly linked population growth to inevitable global starvation. But to many scientists and engineers, the field of economics can hardly be considered a science at all. Unlike physics or chemistry, economics relies heavily on unverifiable assumptions. Most of the few ÒlawsÓ that exist in the field are not confirmed through the rigors of experimentation, but are mere interpretations of real world phenomena. And unlike, say, classical mechanics, a field built entirely upon NewtonÕs universally accepted and authenticated Laws of Motion, economics is a field with many schools of thought in which conflicting and even contradictory ideas are allowed to coexist. Of course contradiction and uncertainty are anathema to scientific purists, and for this reason economics has been relegated to the fuzzier world of social sciences. But economics as a field of study is evolving, as evidenced by a former physicist in Palo Alto who is using Stanford students as economic test subjects and in doing so, changing the way business decisions are made one computerized simulation at a time. The Founding of Experimental Economics
Dr. Kay-Yut Chen, principal scientist at Hewlett Packard (HP) Labs, is a Physics major from Cal Tech. As an undergraduate, he fully intended to pursue graduate research in experimental physics until the innovative work of a particularly famous professor put him on a different path to scientific fame. The professor was Dr. Charles Plott, pioneering researcher in the world of experimental economics whose work has contributed to the recent explosion of experimental methods in the economics profession. Plott himself had worked with the researcher largely considered to be the ÒfatherÓ of experimental economics: Vernon Smith, awarded the Nobel Prize in 2002. It was Smith who conducted seminal experiments on the convergence of price and quantity to their theoretically-predicted market equilibrium. In these experiments, Smith assigned his students to the category of ÒbuyerÓ or ÒsellerÓ of a particular good, each buyer with his own particular valuation of how much the good is worth, each seller with his own cost of production. Each student thus came to represent a point on the downward sloping demand and upward sloping supply curves. When students were allowed to competitively ÒbidÓ or ÒaskÓ on their commodities, the price and quantity transacted quickly and inevitably converged to an equilibrium theoretically predicted by perfect competition. And so was born the field of experimental economics, in SmithÕs small classroom at Purdue University, autumn semester 1955. His experiments consistently showed that abstract economic principles could accurately predict the real-world behavior of autonomous human beings. During the past half century, economic experimentation has grown to validate several theories on economic issues such as auctions, bargaining, finance, learning, games and decision making.
In 1994, Hewlett Packard made quite possibly the largest investment in experimental economics to date by establishing the first economics lab inside a corporation and hiring Chen to lab-test their business ideas. It is here where Chen has worked for the past twelve years, hiring Stanford students to take part in computerized simulations that help HP structure its decision making in the most strategic way possible. Gaming the System
So how exactly do Dr. ChenÕs experiments work, and what has he discovered? Over the years, Chen has worked on numerous projects for HP and his standard approach is the same: observe the economic behavior of human players in games with different payouts. ÒJust like IÕd model the behavior of atoms in a physics experiment,Ó he notes, Òexcept now the atoms can think.Ó Stanford students are incentivized with monetary payoffs depending on how strategically they play the game Ð payments range from $25 for participation to as high as $100. Chen deliberately avoids introducing a concrete context for the game in order to see how gut intuition guides his players. And his goals? Chen is chiefly concerned with issues directly related to HPÕs business strategy. He is, after all, a scientist for an $80 billion company, where scientific research is chartered with the explicit goal of improving the bottom line. As such, according to Chen, ÒI take more of an engineering approach. I am more interested in knowing what works and what doesnÕt: If I have to structure a supply chain or write a retail contract, whatÕs the most profitable way to do it?Ó
In one example of ChenÕs experiments, he researched how retailers get around HPÕs minimum advertised price (MAP) policy. Common among large companies, such policies dictate the minimum price that each product can be advertised to ensure all retailers have decent margins. Because HP has hundreds of products advertised every week, the company has neither the time nor resources to take people to litigate if retailers violate the policy. So a penalty is instituted instead (often fines, or restrictions on what products can be sold in the future). The question is: how do you write the most effective penalty protocol? In simulation with an early draft of the protocol, Stanford students quickly discovered a way to Ògame the systemÓ by playing upon the end of a productÕs lifecycle. Each product that HP sells has a lifecycle (around 18 months for a printer, for example) after which the MAP penalty has no teeth. Stanford students exploited this fact to skirt the minimum price policy but avoid the penalty fines. After Chen discovered this, he recommended the penalty protocol be revised to reflect a different time structure. He redid the simulation and found that even the cleverest Stanford Student could no longer game the system to get around the fines. In doing so, Chen prevented millions of dollars in lost revenue for HP. ÒThe moral of the story,Ó as he says, Òis that if you are going to lose that much money, itÕs better to do so in the lab than in the real world.Ó
Economics as a science?
One of the primary criticisms of ChenÕs field is that humans are too complex for experimental treatment. We act according to free will and our economic decisions could never be as accurately predicted as, say, NewtonÕs laws predict motion. In response, Chen asserts that it was precisely this complexity of human decision-making that first intrigued him, luring him from experimental physics to experimental economics. Further, Chen argues that ÒHuman economic behavior is no more complex than other scientifically-researched areas like protein folding. In fact IÕd argue itÕs less so. Besides, it is precisely because humans are so complex that they are scientifically interesting!Ó When asked about economic laws versus the laws of physics, most notably those famous Laws of Motion, Chen is quick to concede that all scientific theories have their limitations. Because economics is about human behavior, we may never get to a point where we can predict with 100% certainty what a human will do. ÒBut we can get pretty close,Ó says Chen. ÒSay I give you a choice between $5 and $10. IÕd say I can predict with 99% certainty which one you will choose.ÓPerhaps most controversially, Chen also contends that human beings have less free will than we may think. Time and time again in his simulations, students have proven to react in the same way to given incentive schemes. Most of us will choose the $10 over $5. The interesting question becomes: how does our greed change/manifest itself when things get complicated? When you start to deal with uncertainty and strategic considerations, the predictability of human behavior declines. But it still exists. ChenÕs experiments show that while you may not always understand the fundamental principles driving economic behavior, you can often identify statistical patterns that will predict what people do. ÒAnd thatÕs a lot better than even physics can do in certain areas,Ó Chen asserts. ÒPhysics may be great at predicting how the planets move around the sun. But when it comes to predicting the weather, for exampleÉ well, physics is as bad as economics!Ó The Future
While Chen is one of very few experimental economists and even fewer in the corporate world, he is quick to give credit to others when asked how his field is changing the way business is done. He cites a growing research trend where not only economists are using behavior in their experimental methodology, but also researchers in management science, supply chain management, and business operations. Entirely new areas like experimental finance, for example, have grown from the same incentivized payoff experiments and computational tools that allowed experimental economics to take off. Chen claims that he is merely one of many researchers putting the science and engineering into business process designs Ð reducing the guesswork in corporate decision-making.
As for any regrets about switching from physics to economics, Chen is sure he has none. He believes that research in experimental economics has at least as much potential to change the world as did the proverbial apple that fell on NewtonÕs head. After all, ÒWhen you look at the history of physics, itÕs a long one Ð Sir Isaac Newton lived more than 3 hundred years ago. The only real difference between physics and economics really is that the physicists have had a much larger head start.Ó