Foreign exchange: Why Does the Carry Trade Work?

As the article notes, it shouldn't.
From The Economist's Buttonwood blog:

FEW articles on the foreign exchange markets are complete without a mention of the carry trade, under which investors borrow in low-yielding currencies and deposit the proceeds in higher-yielding currencies, in the hope of making a turn. Over the last 40 years, this has been a highly successful strategy; research by Elroy Dimson, Paul Marsh and Mike Staunton of the London Business School (published in Credit Suisse's global investment returns yearbook) found that the annual return since 1971 (when Bretton Woods collapsed) has been 2.3%.

Of course, theory suggests there should be no return from such an approach. High interest rates should be the return investors demand for the risk of currency depreciation*. A reliable return from the carry trade suggests that investors are consistently overestimating the risk of depreciation; creating a "free lunch" for those who take the opposite view....MORE

The DMS trio impacted my thinking eight years ago with their paper "Irrational Optimism".
Even the abstract is profound in its simplicity:

We address the tendency of many investors to overestimate the rewards and underestimate the risks of investing in stocks over the long term - that is, investors' irrational optimism. In particular, we examine the widely held belief that stocks are a "safe" investment for the long run. The probability of experiencing a real loss on equities depends on the expected real return and standard deviation of stocks. Judgments about the future magnitude of these two parameters typically involve extrapolating from history. We use a global database of real equity returns from 16 countries during the 103-year period from 1900 through 2002 to confront the optimism of investors with the reality of history.

Since 1900, the worldwide real return on equities averaged close to 5 percent a year (before costs, fees, and taxes). This is appreciably lower than is frequently quoted from historical averages, a difference that arises because we use a longer time frame than other studies and adopt a global focus. Prior views on the long-run safety of equities have been overly influenced by the experience of the United States. Furthermore, the US evidence that, over the long haul, stocks have beaten inflation over all 20-year periods is based on relatively few nonoverlapping observations and is hence subject to large sampling error.... 

We overweight the experience of a country that has never been invaded during the period of its ascension to the position of largest economy in the history of the world while using a relatively small number of samples.
Hmmm...

Good one guys, got me thinking.