Question

1. The effectiveness of data mining has been criticized by the Wall Street Journal. In one arti- cle, the author notes that academic studies have shown that by using data mining, analysts could accurately predict changes in the stock market based on either the population of sheep in Bangladesh, the number of nine-year-olds at a given time, or whether it is smoggy on a given day.4 While the statistical correlation may be valid, there must be a logical reason that a particular factor will predict stock returns. LO 10-2

   Required:

a. Do you think these findings represent valid relationships or spurious correlations?

b. What measures do you think might be valid predictors of stock market returns?

Answer 1

(a): I believe that these finding are more of spurious correlations and do not hold any valid relationship. This is because there can be no underlying and logical connection and relationship between changes in stock markets (or stock market returns) and population of sheep in Bangladesh. Using data mining and advanced quantitative analysis techniques one can develop a correlation between any two variables. What matters is whether any underlying cause and effect relationship exists between the two variables or not. Here clearly no underlying cause and effect relationship exists between stock market returns and population of sheep in Bangladesh.

(b): Valid predictors of stock market returns will be factors like be various macro and micro factors affecting directly or indirectly a stock. It can be GDP growth rate, inflation rate, interest rate, number of housing starts, number of new jobs created etc. All these factors will cause an impact on stocks. A very good example can be given of the stock of Starbucks. One can correlate Starbuck stock’s figures with that of rainfall figures in Brazil. This is because Starbucks buys a large amount of coffee from Brazil and the rainfall in Brazil directly impacts the coffee prices there. This, in turn, will impact stock prices of Starbucks that is listed in USA.