We revisit Kendall's (1953) conclusion that "the interval of observation may be very important". Contrary to his other conclusions on return predictability, this conclusion has received surprisingly little attention. Most tests in finance and economics routinely regress daily, weekly and monthly observations on daily, weekly and monthly observations, respectively. This is especially surprising because, while convenient, this convention lacks economic reasoning in many applications. Using similar data to Kendall (commodity prices and US, UK and World Stock Market indices) we show how conclusions regarding stock market return predictability vary drastically once we deviate from this convention. Even more surprising: conclusions whether or not stock returns are predictable fluctuate strongly for almost similar intervals of observation. In other words, had the "Demon of Chance" in 1953 offered Kendall slightly different intervals of observation, Kendall might have concluded that stock market returns were predictable.
Read the whole thing here.
Essentially, they find that while many series don't exhibit predictable behavior when using the prior month's return (one of the "standard" intervals used by researchers) as a predictor, they do when using other less standard but equally defensible interval lengths.
The key take-away of the paper is not that markets are necessarily predictable (or even that they're not), but that the choice of interval can make HUGE differences in the results.
HT: Financeprofessor.com
Essentially, they find that while many series don't exhibit predictable behavior when using the prior month's return (one of the "standard" intervals used by researchers) as a predictor, they do when using other less standard but equally defensible interval lengths.
The key take-away of the paper is not that markets are necessarily predictable (or even that they're not), but that the choice of interval can make HUGE differences in the results.
HT: Financeprofessor.com