Investors or any other market participants dealing with the stock markets would inevitably project some form of forecasting. Forecasting tends to use ex post data to determine what the future holds. This thus creates some form of data measurement errors and biases. Similar to behavioral biases and cognitive errors mentioned in the previous articles An Introduction to Behavioral Biases – Part 1 and An Introduction to Behavioral Biases – Part 2, individuals should be made aware of the problems with forecasting. Below are some of the limitations:
1. Individuals using economic data should be well aware that a time lag (which can extend as far as 2 years) usually exists between gathering and distributing such data. (e.g. International Monetary Fund) might report data with a lag of as far as 2 years. In addition, although these data might be revised often, the revisions might not be made at the same time with the publications. Data definitions and methodology tend to also change over time. For example, the basket of goods which are included in the CPI changes over time and the data sets are often rebased over time. Although rebasing might not be a substantial change in the data itself, individuals might incorrectly calculate changes in the indices value if no appropriate adjustments are made.
2. Individuals might incidentally uncover patterns in security returns that are unlikely to occur in the future, which can produce bias in the data set. Some variables might appear to have a relationship with security returns when in fact these relationships are unlikely to persist. An example would be testing the relationship between stock returns and 40 randomly selected variables with a 5% significant level. Just by pure luck, 1-2 of the variables might show statistically significant relationship with stock returns. In order to avoid these biases, individuals should thus critique if there are any economic basis for the variables found.
3. Lastly, the misinterpretation of correlations is common when it comes to using historical data for forecasting. For example variable A might have a correlation with variable B and hence, individuals might conclude that A have an influence on B. However, that is not to say that B influences A as well. It is possible that could be a third variable influencing both variables
Value in Action
There are limitations to using historical data for forecasting, and individuals should be made aware of such measurement errors and biases.
Join us on Facebook for more exciting updates and discussion about value investing. Submit your email address for important market updates and FREE case studies!
We will only provide you with information relevant to value investing. You can unsubscribe at any time. Your contact details will be safeguarded.The information provided is for general information purposes only and is not intended to be any investment or financial advice.
All views and opinions articulated in the article were expressed in Willie’s personal capacity and do not in any way represent those of his employer and other related entities. Willie does not own any shares in the companies mentioned above.