Imagine you’re a wine producer and you’re changing your prices. Your customers have an infinite range of preferences — some of them require very high quality wines while others are happy with the two buck chuck. You obviously can’t make a separate bottle for every customer, each with a different cost suited for their individual budget. Instead you need to figure out some way to provide a finite range of qualities — a gold, a silver and a bronze label, say — and stamp each with an optimal price tag.
But how do you do that? How do you identify optimal pricing when your million or so customers all have different needs? The standard economics method is to do some kind of qualitative market research study. “There was no systematic way of arriving at the optimal wine qualities that one should make,” said Electrical and Computer Engineering professor Edmund Yeh.
In June, Yeh presented a paper at the Electronic Commerce meeting of the Association for Computing Machinery that adds an entirely new method to the field of economics. “We can use information techniques to analyze economics problems,” said Yeh. The paper uses a method called quantization to identify your optimal wine (or any other product) prices.
Typically, quantization is a data compression technique that turns things like analog audio sound into a digital file for your iPod. “It’s a powerful tool that economists had not used before,” said Yeh, who was apparently approached by many interested economists after the paper’s presentation.
The method has the potential to reach a range of economics problems that require solutions based on limited information, Yeh said. He is already in discussions with several local economists who wish to implement the method in their own research programs.
In the case of a digital audio file, the information (an infinite collection of sound waves) needs to be turned into a finite number of bits (or regions). You cannot represent every possible sound, but rather need to deliver a representative collection of sounds.
In the case of product pricing, you have an infinite number of preferences but only three or four price regions. “The trick is how to choose those regions,” said Yeh. You need to choose three or four types of wine whose qualities “maximize either the aggregate social welfare or the seller’s revenue,” he explained.
“It is quite serendipitous how a standard problem in economics becomes a problem in information processing,” said Yeh. He and his colleagues mapped the variables in an economic pricing problem to the variables in a data quantization problem. The product’s quality becomes the representation point while the preference parameter becomes the signal to be quantized.
The work is a collaboration between Yeh, his graduate student Yun Xu, Professor Dirk Bergemann from the Yale Economics Department and Ji Shen, a graduate student at the London School of Economics.