The trade off is usually ~same functionality but better graphics. Compare Office 1997 with Office 2017 and the average user sees a very modest improvement relative to how much better of a computer you need.
Another thing is that users also have a lot more applications running in the background, calling servers, caching data, disk I/O.
I don't think that's entirely clear. At least for cpus, less transistors in practice means you get rid of super-scalar architectures, and over-engineered microcode decoding and branch-prediction units. This leads to a modest reduction in processing speed with far larger gains in battery life for mobile devices (think X86 versus Arm). Client side applications are mostly unaffected in terms of end-user functionality.
I think this is far more difficult to evaluate than your comment admits.
More transistors -> more software that can do more things, written by developers with a wider range of skills, in a society with increased tech permeation, and so on...