To solve this riddle meself, I ripped a song from a favorite CD of mine to WAV, and then to 192kb CBR(constant) and 192kb VBR(variable) mp3s.
I then performed Null tests: WAV vs the CBR file inverted, then WAV vs the VBR file inverted. The null for the WAV vs VBR was only half as loud(to my ears) as the CBR null, so according to my tests, yes, VBR is the way to encode MP3(or AAC for that matter).
I don't think Variable Bit Rate codecs for lossy formats actually *lower* the bitrate, from that selected, in less complex parts of a song, but I do think that the amount of variablity - UPward - can be controlled in the settings menu of your preferred software(iTunes, Exact Audio Copy, etc.).
IE: In my case I selected iTunes 192kbps, VBR "medium" quality, normal stereo. iTunes reported the bitrate of those files as 202-208kbps Variable. Very likely, that is what the bitrate was 'throttled up'(from 192kb) to during the busier sections of the songs I ripped.
In any case, I have actually re-ripped several dozen tracks, just as as an experiment, from constant bit rates of 256kb or higher, down to 192VBR Medium, and have not heard any difference at the lower setting. And the space savings was about 50mb!
I also did some test rips of(all other things equal) normal(full) stereo vs joint stereo, and the space savings(insignificant compared to VBR vs CBR) are not worth the slight loss in separation(again, confirmed by null tests - the joint nulls were very wide and panned out!).
20th century folks:
Your mileage may vary.
21st century/Millennials:
"YMMV"!
😉