Let’s consider a power supplier with market power and zero marginal cost. Capacity suffices for ten units but five units are sold at p = 10; selling more would lower profits. Now, using carbon offsets, bribe the fifth buyer to stay out of the market, say by walking to work rather than flying his jetpack. Even better, just shoot him.
The company has two options. It can stick with selling four units and raise price. Or it could drop price a bit and pick up a fifth buyer again. Hard to say what will happen. Alternatively, if buyers stand along a continuum, is there a general proof one way or the other?
Rather than bribing the fifth buyer to walk, invest the "carbon offsets" money in building a nice comfy sidewalk. In principle all buyers could walk on this new path.
It is then easy to see how the power company might lower price and expand to six units or more. Otherwise they might lose all their customers.
A key question is the cost structure of the alternative clean technology. Non-scalable technologies, with little potential for expansion, are the least likely to backfire and least likely to lead to more dirty power. Scalable technologies, such as the sidewalk, are most likely to backfire and make the world dirtier. They require a bigger competitive response on the part of the dirty power supplier. (At least in the short run this is true, in the longer run the scalable technology might eliminate dirty power altogether.)
This counterintuitive conclusion is one reason why we have economic models.