To use a single data point to determine where a future data point will lie is really nothing more than a guess. It isn’t scientific, and yet, we see this practiced in areas of finance and economics reports to which people tend to pay a lot of attention.
Financial institutions see this problem constantly when they file their call reports. Anything reported on your first quarter call report will automatically be multiplied four times to assume this is the trajectory of the entire year ahead of you. Did you charge of a loan in the first quarter? Now your regulator might assume you will be charging off four times this amount over the coming year. Did you have an unforeseen expense? It will be multiplied by four, and your financial performance report (FPR) assumes it is recurring over the next three quarters. Of course this could work to your advantage if you also realized a significant gain or recovery in the first quarter.
This problem with annualizing data was recently pointed to the Bureau of Economic Analysis (BEA) by CNBC. The reporters at CNBC caught on to a recurring issue where first-quarter GDP numbers continually seemed subpar when annualized, which caused panic and policy concerns on a regular basis.
In April, CNBC released a report showing that first-quarter GDP was consistently lower or “weaker” than the following three quarters for the past 30 years. Economists noted that they attempt to adjust out seasonal variations, which might cause this anomaly in the data, but still a “residual seasonality” exists.
The BEA, a government office, responded directly to CNBC noting it was “aware of the issues” and “is developing methods to address what it has found.” Read more about it at http://www.cnbc.com/id/102695676 .
If our government’s economists cannot effectively forecast the future, then who can? Nobody, that’s who! The problem we must accept is that a single data point cannot enlighten us to where the next data point will lie. At best, we can look at a trend using several data points, but that itself will still only give us a guess; albeit, a more informed guess.
If we want to dabble in the art of financial fortune telling, it behooves us to at least take a logical approach towards forecasting. We want a good span to our data, meaning we want it to cover several relevant periods. One way in which this is approached is to use a trailing twelve-month (TTM) analysis. This means no matter what month you examine data, you are getting a snapshot of the last 12 months, giving you a true annual year-to-date picture.
Another method to assessing a trend is to look at data points only from the same period. This is why we might always look at just year-end results. If we want to look at the most relevant month, say May in 2015, then we should not compare it to other year-end numbers, but rather to May 2014, May 2013, and so on.
To summarize, we understand that annualizing data is desirable and well intentioned, but it is a poor tool for forecasting and is a random guess given the unscientific nature of the method. We should pay more attention to groups of previous data points that correspond with the same period. This means seasonality will naturally be adjusted for and will not require complicated adjustments that further mystify a murky guess.