Although Dr. Stephen Millet does not exclusively discuss economic trend analysis in his article entitled Trend Analysis as Pattern Recognition, he does provide some good insight into trend analysis as an intelligence tool. Millet believes that analysis of trends (repeated phenomena and data that display an inclination toward a certain direction over time) is able to combine both quantitative and qualitative data in order to produce actionable intelligence regarding the future.
Millet begins his article by defining trend analysis as “the disciplined and systematic way of extracting meaning and recognizing patterns in trends.” He quickly makes the point that such trends—whether they be economic, social, or historical—do not follow immutable laws, lending any prediction a degree of uncertainty.
Millet postulates that there are three types of pattern recognition, and therefore three types of trend analysis: 1) Background, 2) Signals, and 3) Scatters. Background looks at some point in known history and attempts identify both continuities and deviations that may arise from this past norm. Signals identify known trends while filtering out the background “noise”. Finally, scatters deal with apparently random signals that are inductively interpreted by emerging trend analysis.
Based upon the situation, economic trend analysis could fall under any of these three types. Case studies in economics are typically Type 1, as they use static or dynamic backgrounds to understand norms, detect deviations, and predict future conditions (as long as there are no critical deviations). Type 2, on the other hand, focuses on specific patterns and isolated trends, placing emphasis on dramatic changes and not background continuity. This type of economic analysis would be useful for forecasting, and perhaps avoiding, economic recessions and depressions. Finally, Type 3 economic analysis deals with unfamiliar trends or patterns, and the attempt to arrange them in a sensible fashion. Millet warns against quickly placing high estimative probability in these types of new models.
In closing, Millet again reiterates that his proposed trend analysis is not designed to produce, nor is it capable of producing, 100 percent certainty regarding the future. Instead, he hopes that “this line of thinking may lead us to new and more prescient methods for forward looking and actionable intelligence.” In order to achieve this goal more sophisticated trend analysis methods must be developed for analyzing background and signal patterns, and scenarios and Bayesian probabilities should be incorporated into scatter pattern models.
MiIlett, S. M. (2009). Trend Analysis as Pattern Recognition. World Future Review, 1(4), 5-16.