Is it worth the cost to overcome the inertia of billions of keyboards and typists? People who really really need every bit of speed they can get (e.g. court reporters) either already use an alternate or specialized layout.
With court reporting, it's not just a matter of sheer keying speed. In both pen shorthand and mechanical shorthand, there are "short forms" where a single penstroke or key chord can reproduce a lengthy string of text. The difficulty with this is that there needs to be a "re-translation" from the shorthand into the expanded text. This has a couple of ramifications. Firstly, to learn shorthand, either pen or machine, is a very lengthy and expensive process, usually a minimum of a couple of years. This is because the shorthand may have to be translated at a later time, possibly even after the original scribe is no longer around, and hence must be practised thoroughly, since it must be exactly the same for everyone. This still applies to pen shorthand (where used - very rarely) (If I were given my Pitmans practice notes from 1979, I wouldn't be able to read a single thing.) and until recently it also applied to machine shorthand (Stenotype), which is called "scoping". This has been reduced somewhat now by the use of computer assisted typing (CAT) which does some of the scoping in real time. This scoping/ proofreading can involve the employment of another person, increasing cost. (The time taken to do this does not seem to be factored into calculating a "typing speed".)
http://en.wikipedia.org/wiki/StenotypeThe Wiki article gives examples of the output and "code" of a Steno machine.
Next, the specialised machines are very expensive, since they are only produced in small numbers, far more expensive than even the dearest "alternative" keyboard.
However, modern software in the field of word processing has the facility to produce its own form of "shorthand" to a greater or lesser extent, depending on which software is used. The gift here is that the abbreviation expands immediately, (only) whenever the operator wishes it.. In one instance, it can be demonstrated that there is a Ki:Ko (Keystrokes in to keystrokes out ratio) which can be as high as 40:100 ie for every 40 keys stroked, 100 characters can be placed on screen. And this is entirely dependent upon the software, not the keyboard layout, since the layout can be factored into the creation of the abbreviations.
An example of how layout can affect abbreviations: When I was using QWERTY I would consistently (ie 80%+) key the word "the" as "t-e-h [backspace] [backspace] h-e". So I created a shortform of Ctrl T printing "the". However when I changed over to Maltron, I no longer made that error, but to this day I still use Ctrl T for "the". Which actually proved advantageous, because I was able to use this in words like "them", "they", "there", "their", "other", "mother", "father" etc etc.
When I work on specialised subjects such as medical reports, I've calculated that sometimes I've "typed" at 2-4000 wpm, ie I've taken 10-12 seconds of speech and reproduced it onscreen with about 4-10 keystrokes, a Ki:Ko of hundreds.
For those of a curious mind, this link describes (in detail) how to do keyboard shorthand.
http://proword-transcription.blogspot.com/Joe