A new study,Activity energy expenditure is an independent predictor of energy intake in humans, published this year in the International Journal of Obesity, set out to look at this phenomenon.
Now to be clear, the study certainly wasn't designed to explain exercise's impact on weight. It was just 7 days long and it involved the retrospective analysis of data from 5 prior studies and did not directly measure energy expenditure or energy intake. Instead researchers utilized estimated energy expenditure by way of heart rate and indirect calorimetry data, and energy intake by way of known to be problematic food diaries.
My stats skills are nowhere near good enough to comment on the various treatments of the data, but here's the scatter plot of the impact of energy expenditure on energy intake.
The increase in energy intake the authors attributed to energy expenditure wasn't high, roughly 3% of total daily calories (around 70 in this sample), an amount too small to explain away exercise's often uninspiring impact on weight loss.
Truth be told, I'd have predicted the difference to be larger as eating more consequent to exercise is something I know many people do for one or more of the various reasons mentioned above.
Of course none of this changes the fact that exercise has tremendous health benefits at any weight and that weight shouldn't be your driver for upping yours if you're able.