This does not represent one of Jerry Coyne’s best introductory paragraphs:
This article by Gregory Paul in the new journal Essays in the Philosophy of Humanism (click on screenshot for pdf) argues, based on polling data, that the proportion of Americans who are atheists is rising by 5-10% a decade, leading to the conclusion that within a century America will comprise mainly nonbelievers.
Seriously? By that “logic”, the fact that American sugar consumption used to rise steadily with each decade would lead to the conclusion that by the year 2000 sugar should have represented over half of the typical American diet. No. Statistics and trends don’t work that way. (Despite the long rise, sugar consumption actually dropped in recent decades.)
This misuse of statistics reminds me of YEC origins-ministries’ arguments claiming that because the moon is moving away from the earth at a steady annual rate, the moon must be very young. (After all, extrapolating that annual rate backwards would mean that the moon couldn’t be millions of years old—or so say Jason Lisle, Ken Ham, et al.)
As I’ve observed many times previously, the worst arguments across the belief and non-belief spectra tend to share many of the very same logical fallacies.
What do you mean??? In just 2 centuries, the US will be 200% atheist!!! That sounds great! 
Removing tongue from cheek, I agree with you. You can’t extrapolate from this data. Religiosity in the US has fluctuated in the past, and I would expect it to fluctuate in the future. I find in very improbable that religiosity will be as common as it was in the past, but there is no reason to expect consistent trends through long time periods.