No, it isn't simply noise shaping. I looked at the patents some years ago and summarised what they say here:
http://www.audiomisc.co.uk/MQA/origami/ThereAndBack.html
and
http://www.audiomisc.co.uk/MQA/bits/Stacking.html
The result tends to be additional 'noise' and anharmonic components in the HF region.
However the patents omit key details so make it difficult to assess the full impact. AIUI MQA also block input-output comparisons. Although 'GoldenOne' on ASR seems to have managed this.
I formed the view that, indeed, noise shaping would do as well in terms of file size reductions and stream rate reducing as per:
http://www.audiomisc.co.uk/MQA/intoshape/NoiseShapingHighRez.html
without the added deterministic anharmonics indicated by the patents. (And which GoldenOne's measurements seem to confirm.) Also, being free and open, cost nothing and can be fully scruntinised. Another alternative based on the cause of 'high rez' being inefficient is also point out here
http://www.audiomisc.co.uk/MQA/cool/bitfreezing.html
However I wrote that mainly to show why genuine lossless compression a la FLAC doesn't do so well with 24 bit high rez due to the 'sea of noise bits' that carry nothing.
If you look up-thread you'll see I'm currently doing some more examination/analysis. I'll report on that when I get more info. Based on GO's files and ones released by 2L.