The Handbook of News Analytics in Finance

9.4 MONTE CARLO–STYLE SIMULATIONS

It behooves us now to consider the worst possible investment experience from sentiment reversals. Imagine a scenario in which we see the marvelous forward 1-year excess return number of 16% over the entire universe U₁₇ eagerly buy a portfolio of stocks matching the appropriate parameters, then underperform the market by 10% each year over the next four years. This is possible. Unfortunate, but in such a case we would not have considered carefully enough the real-time quality of excess returns. It could be that the documented excess returns all came from a bull market in the late 1990s or, for that matter, a bear market in the early 2000s. In a sense these outcomes are unavoidable; the stock market is not like the physical world, and the principle of scientific induction (that the future will be like the past) should be applied with the utmost reticence. There is more comfort, however, in the belief that something never will happen because it never has happened, as opposed to turning a blind eye towards history.

Generally speaking, Monte Carlo simulations attempt to estimate the range of a function by taking randomly generated arguments, performing the function on those arguments, then aggregating the function values into a final result. There is an obvious application of Monte Carlo methods to estimating portfolio performance in an event study. Starting the portfolio formation process at different events, if there is a sufficient number of them ...