Early in my career as an economist, I worked for a large bank that employed a group of more than twenty economists, most of which held either PhD’s or ABD’s (all but dissertations) – so really smart people.

Once a month, we would all convene to develop the group’s economic forecast. The first half of the day we devoted to briefings on global economies, major events, government policy, etc. The second half we spent formulating the short-term GDP forecast. Each forecasting session started with announcing what the current consensus estimate was.

Being the young economist, my task was simple: be quiet, listen, gather data and feed it into my long-term forecast model. The official short-term forecast however, was still derived from the figures that were agreed to at the end of the day.

The group was incredibly accurate in its short-term forecasts, rarely missing the next reported GDP estimate, and frequently “hitting the number,” and even winning the prized Blue Chip Consensus award a time or two. Oddly however, the model that I maintained seemed to not work at all. After inputting the senior economists estimates for consumer spending, business investment, government spending and net exports, my long-term model produced a short-term forecast that was nowhere near what was actually reported.

So I tinkered and even got on the phone with the company that built the model - an even larger group of PhD's - who helped me navigate my way around inside the model. After a few months poking and prodding at the model, which was an endless maze of interrelationships, I discovered something. The model was good for short-term forecasting, but the data that was fed into it from the meetings was not. In fact, the short-term forecasts prior to “massaging” the data were really good, one-quarter, two quarters, and even a full year in advance of what was ultimately reported.

It took a while, but I eventually realized the problem. The group was merely "forecasting" what the consensus was, and the consensus was merely “forecasting” what was already known. While the group was good at was addition and subtraction, they were horrible at explaining how we moved from today to tomorrow, or in essence how the economy actually works. As a consequence of their goal of "hitting the number" they were completely ignoring the impact their numbers were having in other areas of the modeled economy - and for that matter the real one too.

As my career in the industry continued, I came to realize that it is simply human nature to focus intently on today, and in particular on what the consensus expects; even if it is widely at odds with rational thinking and longer-term probabilities. In other words, they were afraid to be wrong in the short-term, at the expense of being accurate and useful long-term. Ultimately, my economic peers were practicing what is known as “groupthink” – smart people, when together, making really dumb decisions as a result of their attempt to be appear smart.