Forecasting with Econometrics

Posted By: Advisor Analyzer Team

Econometrics: a word many chose to leave behind with their wrinkled AC/DC posters and futon couches at their old alma mater.  Some cringe at the mere thought of statistics and normally distributed probability functions, while the more mathematically inclined enjoy venturing deeper into a world of regression and forecasting.  While a tad more challenging than English 101, econometrics can be an effectively simple tool for analyzing one’s portfolio performance.

 

The main benefit of econometrics stems from its forecasting abilities.  Without going into too much math, econometrics takes a set of data and tries to predict an outcome via a regression equation.  In simple linear regression, we take two data sets, the dependent variable and independent variable and use the data to forecast an outcome.  In the case of multiple regression several independent variables are used to predict the dependent variable’s value.  If these last statements left you more confused than a kung fu movie with no subtitles, please read on.

 

Rather than using stock market data for this example, I’ve decided to use a more common data medium: cars.  We all drive them or at least have ridden in one.  Many would agree cars are all different, especially when it comes to fuel efficiency, but why?  There can be many reasons for the differences; the car’s horse power, the country it was engineered, or the number of cylinders in the engine to name a few.  So how could one “guesstimate” a car’s fuel efficiency without looking at the sticker or recent issue of Automobile magazine?  For those that haven’t been paying attention, I’ll repeat: econometrics.

 

First off, we are trying to forecast fuel efficiency, so this is our dependent variable.  But what does fuel efficiency depend on?  To keep it simple, we’ll say a car’s horse power, the number of cylinders, and the 0-60 time are the three independent variables that determine the MPG.  A word of caution for the mathematically squeamish: please brace yourselves for minor number crunching. 

 

To forecast, we must take the data of each variable, and obtain a regression equation.  Since our argument was that fuel efficiency depends on the number of cylinders, the horse power, and the 0-60 times we will gather the data for these variables:

 

Upon entering the necessary data and running a regression (using Microsoft Excel), we obtain the equation:

 

Y             =      40.989     -   .309 X1   -     .0487 X2 -   .9128 X3

(MPG) =      (Intercept)     –     (Cylinders)     –     (HP)     –     (0-60 time)

 

Using this equation, we can essentially “plug and chug” the independent variables to predict a car’s fuel efficiency.  For example, the Cadillac STS’ fuel efficiency is ~ 18.5 MPG.  Using our acquired equation we predict the STS’ fuel ergonomics to be:

 

Cadillac STS = 40.989 – (.309 X 6) – (.0487 X 302) – (.9128 X 6.5) = 18.39

 

Although not exact, we were able to approximate the Cadillac’s MPG by plugging the needed variables into our regression.  The forecasted result (18.39) is very close to the actual MPG (18.5) of a Cadillac STS.  Had we used a different set of variables, such as torque, curb weight, cylinder size and ¼ mile times, we would obtain a completely different equation yielding different results.  Ultimately, it is up to the individual to determine which variables provide the most accurate forecasts.  Also note, as the data set increases, the forecasted MPG will be closer to the actual value: the more car data entered, the more accurate our equation predicts MPG.

 

So what does a car’s fuel efficiency have to do with stock market?  Not much unless your portfolio includes GM, Ford, Toyota, etc…    However, imagine discovering a regression equation that can predict a stock’s future price given the volume, previous close, and daily highs and lows.  Here the stock’s future price would be the dependent variable, and the others the independent.  Needless to say, anyone who can predict a stock’s future price will earn enough money to make Warren Buffet blush.  Of course finding this equation is easier said than done, but it doesn’t keep many market statisticians and quantitative analysts from trying.

 

While far from an all encompassing guide, the goal of the article was to reintroduce a concept that was long tucked away with our memories of Friday night fraternity parties and ramen noodles.  Econometrics can be a very useful tool to better gauge a stock’s price movements, and analyze the causation of those movements.  It can be used to estimate future inflation rates, and GDP expectations.  But before I bid adieu, a word of caution: stock prices are not easy to predict, after all if they were, would I be sitting here writing this article?