"This paper develops a simple but general methodology to estimate the expected intertemporal marginal rate of substitution or "EMRS", using only data on asset prices and returns. Our empirical strategy is general, and allows the EMRS to vary arbitrarily over time. A novel feature of our technique is that it relies upon exploiting idiosyncratic risk, since theory dictates that idiosyncratic shocks earn the EMRS. We apply our methodology to two different data sets: monthly data from 1994 through 2003, and daily data for 2003. Both data sets include assets from three different markets: the New York Stock Exchange, the NASDAQ, and the Toronto Stock Exchange. For both monthly and daily frequencies, we find plausible estimates of EMRS with considerable precision and time-series volatility. We then use these estimates to test for asset integration, both within and between stock markets. We find that all three markets seem to be internally integrated in the sense that different assets traded on a given market share the same EMRS. The technique is also powerful enough to reject integration between the three stock markets, and between stock and money markets"--National Bureau of Economic Research web site.