This is actually a question of the feedback of systems with a long delay. In system control theory, systems which have a long delay in them (between corrective action and output) are difficult to control through feedback, but are controllable through "feedforward".
For information: feedback is a general technique where the output of a system is sampled, compared to a desired output, and a corrective input is given which should give a change in output in the opposite direction as the measured distance. The whole theory of feedback is about how to determine the amount and timing of this corrective action.
A simple example is the linear proportional feedback loop.
Consider a system, which has an input, and an output. Imagine that the system is a strong amplifier. It means that even a small input that deviates from 0 will give a huge output.
Now, consider that we sample the output, divide the result by 10, and subtract this from the input signal before giving it to the system. This means that from the moment that the system input starts to deviate from 0, it starts to give a big output, but this output will (divided by 10) be fed negatively into the input, which would then make the output lower again strongly. One can show that the solution of the system equations is such, that the input of the system is near-zero, and hence the output is equal to about 10 times the input signal. Indeed, one will then subtract from the input signal one tenth of the output, or one tenth of ten times the input signal, which results in about 0.
If the input signal rises a bit, then this will make the system input go slightly positive, and the output signal will rise strongly, until the subtraction will again make the input to the system equal to about 0.
So in this case, if the system amplifies "enough" (without any specific nice properties), the feedback will entirely determine the response of the overall system.
But the problems start when the system has a delay: when the output doesn't change *immediately* as a result of the change of the input, but a bit later. Indeed, with an above type of feedback, one can get oscillations. When the output is too high, the feedback system gives back a negative input signal, but this doesn't change the output immediately. So the feedback system continues to give a negative input signal, until finally the output reacts to this, but becomes now too low. As a response, the feedback system will give a positive input signal (in order to increase the output), but again, it will not have any effect until some later time. As a result, the output stays too low for a while, until this input has its effect, where the output becomes too high again etc...
Feedforward is different: one tries to anticipate what the system will do for a given input, and tries to give a corrective signal right away, without sampling the output. But this means one needs a model of the system, and if the model is wrong, the output will not be what one desires.
Now, I don't know in how much one can identify "government interaction" as "feed forward" and "capitalism" as a feedback system, but in as much as it is, the delay between "inputs" (CO2 production etc...) and the "outputs" (changing climates, effects on investments, costs induced by it...) is probably way too long for a feedback mechanism to be efficient.
The question of whether government stuff is really feed-forward can be put in question: after all, there's feedback from the electorate, and this is the "time constant" by which politicians work. Politicians which work for the good of people 30 years from now, but which place a burden right now, don't often get elected, or re-elected.