Relevant: http://mises.org/daily/4223
Income can still be held constant in an analysis, even with a price change, if you simply assume the price change’s concomitant income effects are negated by other equal and opposite income effects. That way you’re isolating the direct effect of a single price change on action given a single set of value scales, instead of looking at a shift from one set of value scales to an entirely NEW set that emerge from a complete revolution of the market data.
It hardly makes sense to invoke an “all
else equal” condition in cases where all else is of necessity never equal !