The Big Mac index, created in 1986 by the Economist to estimate the over- or under-evaluation of currencies, is based on the price comparison of a uniform good across countries. In the best case, purchasing power parity would hold, and it typically does not. But the choice of McDonald's Big Mac always struck me as poor for such an exercise. Its major ingredient, meat, is subject to regulation and subsidies that vary considerably across countries, including trade barriers. And a major part of its price is the "service" which is offered by the restaurant, which is non-tradable. In short, it would even be a surprise were purchasing power parity to hold.
Kenneth Clements, Yihui Lan and Shi Pei Seah have looked at this more formally than I just did. They find that the Big Mac index indeed suffers from biases, and thus its predictions are biased. But they can be corrected. The index even beats the best predictor of exchange rates, at least at medium to long range, the random walk. This should not surprise us, however. After all, if there are strong deviations from purchasing power parity, they should correct themselves in the long run.
I prefer the iPod index that the Commonwealth Bank of Australia computes periodically. iPods are traded, identical (once you choose which one to index), and widely available. Only drawback: Apple has the ability to price to market thanks to its market power. I am waiting for someone to use it to see whether it beat the Big Mac index.