> You're plotting changes rather than levels,
> right?
Neither nor. Proportions, shares, percentages (of total answers), or -- if you wish -- probabilities. An index is not bounded (doesn't fall between zero and one). And being an average, an index is not as volatile as the probabilities (weights) used in computing it.
I was not referring to the evolution of *average* "confidence" levels, but rather to that of the "Betters" (optimistic responses) and "Worses" (pessimistic responses) -- i.e. the two extreme values of (say) a variable like the one captured by a confidence measure. (Hence the subject line.) The "monthly" changes I reported are the slopes of the linear trends of these proportions.
But, you get basically the same dramatic picture if you use the data you just posted. One can imagine that for the indices to have evolved the way they did, a lot of "Betters" switched to "Worses" over the last decade.
Consider the indices alone: Take their evolution over the decade. If you normalize 1997 to 1 for both indices, you'll see immediately the indices declining 1.5 to 1.6 index points per year. Or from 100 to 86 -- that is, a 14 index-point drop over the decade. Isn't that dramatic?