Trend forecasters tease out patterns via a mixed toolbox of techniques, but their predictions are only as good as their data, assumptions and skills. You can run a trend analysis in Microsoft Excel but, without a guiding rationale, what you end up with is likely worse than useless -- it's probably misleading to boot.
If you want your forecasts to say something meaningful, you can't just "plug and chug." You must gather good data, make sure it says what you think it says and check that it doesn't contain hidden relationships that will torque your results. You might need to adjust it, too, for inflation or seasonal variations. And, most importantly, you must know and respect your core assumptions -- including whether you can reliably project present patterns into the future.
Three of the most commonly used forecasting methods are called linear trend (aka simple linear regression), multiple regression and autoregressive integrated moving average (aka ARIMA). Here's the gist of each:
- Linear trends fit a line to scattered data. They make for fairly vague trend gauges, and analysts typically turn to them when stuck with scarce or unreliable data [source: Nau].
- Multiple regression provides a handy way to deal with variables when more than one influence is in play, as is the case with interest rates and other economic indicators. Such models only work properly when you know and have data for all of the relevant forces [source: Nau].
- ARIMA allows forecasters to deal with events that are not independent from one another, and excels at smoothing out noise, outliers and random fluctuations. Seasonally adjusted unemployment figures are a good example of trends typically analyzed using ARIMA [sources: Meko; Vogt].
However many fancy charts and graphs you have, and however solid your theoretical framework, predicting the future remains an uncertain prospect. Trends are blunt tools. They trace out average inclinations, often through a field of data that scatter wildly or that contain other patterns when viewed at larger or smaller scales. It's one thing to stare at data, Magic Eye-poster style, and perceive a pattern; it's quite another to then assume that pattern will repeat in some predictable way, without random events tossing a wrench in the works.
Good trend forecasters wear many hats. They are part philosopher, part historian, part enthusiast, part scientist and part artist. They are big-picture thinkers who don't skimp on details, team players who rely on their singular hunches and knacks. If that sounds overwhelming, we don't blame you. But if it sounds like the challenge you've been looking for, then we foresee an exciting career for you in trend forecasting.