sábado, outubro 05, 2013

Big data, pois!

Como eu aprecio estas histórias, sobretudo agora que leio tanto sobre os amanhãs que cantam baseados no big-data
"Imagine that you are a subject in the following experiment, conducted by pioneering behavioral economist Richard Thaler and his colleagues. You are told that you are in charge of managing the endowment portfolio of a small college and investing it in a simulated fi nancial market.
The market consists entirely of just two mutual funds, A and B, and you start with a hundred shares that you must allocate between the two. You can put all of your shares into A, all of them into B, or some into A and the rest into B. You will be running the portfolio for twenty-five simulated years. Every so often, you will be informed of how each fund has performed, and thus whether your shares have gone up or down in value, and you will then have the opportunity to change how your shares are allocated. At the end of the simulation, you will be paid an amount that is proportional to how well your shares have performed, so you have an incentive to do as well as you can. Before the game begins, however, you have to choose how often you would like to receive the feedback and have the chance to change your allocations: every month, every year, or every fi ve years (of simulated time).
The correct answer seems obvious: Give us information, and let us use that information, as often as possible! Thaler’s group tested whether this intuitive answer is right— not by giving people the choice, but by randomly assigning them to receive feedback monthly, yearly, or every five years. Most people initially tried a 50/50 allocation between the funds since they knew nothing about which might be better. As they got information about the per for mance of the funds, they shifted their allocations. Since the simulated length of the experiment was twenty five years, the subjects in the five- year condition got feedback and could change their allocations only a few times, compared with hundreds of times for the subjects in the monthly condition. By the end of the experiment, subjects who only got perfor mance information once every five years earned more than twice as much as those who got monthly feedback.
.
How could having sixty times as many pieces of information and opportunities to adjust their portfolios have caused the monthly- feedback investors to do worse than the five-year ones?"
Trecho retirado de "the invisible gorilla - And Other Ways Our Intuitions Deceive Us" de Christopher Chabris e Daniel Simons.

Sem comentários: