Don't think of them as forecasts--the market can do anything.

Think of them instead as what a reasonable person should expect. If you remember that, they're spectacular and directly useful work.

They explicitly assume that valuation levels will be at some kind of "normal" at the end date, similar to the average in the past. Maybe they use too much or too little history, but that's a quibble. They key point is that one should never expect more than that. If valuations expand to unusually high levels during the term addressed, then returns will of course be much higher than the figures they suggest. Sure, it might happen, so it would turn out to have been a poor "prediction". But what sane person would plan on that assumption?

When GMO publishes something called a “7-Year Asset Class Forecast” we are not supposed to mistake it for an actual forecast. It is what a reasonable person should expect. When reality disagrees, it is reality that is unreasonable.

Their entire defense rests on assuming valuations will revert to some historical normal. What if the composition of the market, corporate profitability, and economics of the business have structurally changed?

Nobody is criticizing GMO for failing to predict the future perfectly. The criticism is that their framework repeatedly underestimated US equities. But they continued producing remarkably confident projections from assumptions that should have deserved serious rethink.

When the model works - brilliant forecasting. When it fails - it was never a forecast in the first place.

GMO may have discovered the holy grail of investment research - a forecasting model that can be spectacularly wrong for decades without ever having to admit defeat.