Network effects / path-dependence / lock-in

There's quite a bit of evidence that Dvorak improves typing speed and comfort.

In the first place, the world record-holder for typing speed to this day is a Dvorak user. Dvorak users, though a tiny minority of typists, have been disproportionately represented among typing speed and accuracy contest winners and high placers.

Second, QWERTY was designed specifically to reduce typing speed. At the time of its design, type bars used gravity, not springs, to fall back down after striking the ribbon. But this worked too slowly for the type bar to get out of the way of the type bars with the following letters of the word coming in because the typist had hit the next keys quickly. Rather than tackling the mechanical problem of speeding up post-impact typebar movement, rather than fixing the flawed design to serve the convenience and needs of the user, the early typewriter entrepreneur Christopher Latham Sholes focused on slowing down the typist, thus on imposing an anti-feature on users to serve the needs of an inherently flawed system. The same paradigm as unfree software, of putting someone other than the user first.

Accordingly, the men who worked for him (Carlos Glidden and James Densmore) had the goal of physically separating keys that are frequently struck in sequence. Amos Densmore (James' brother), an educator, provided data on digraphs, commonly paired letters, which Sholes then arranged as far from each other on the keyboard as he could. While making typing unnecessarily slow and cumbersome had some positive effect on jam prevention, it didn't work from a sales perspective. Typewriters sold very slowly.

Typewriters only became a breakout, mass market product when their second generation came out in 1875, which had two innovations: first, at last having springs that pulled back the typebars, thus solving the jamming problem the right way, and second, showing the typist the page he was typing on as he was typing on it (previously you were typing blind - you couldn't see the paper until you were finished and pulled it out). Unfortunately, the manufacturer (Remington) retained the now-obsolete and totally unnecessary QWERTY layout with this generation, and thus QWERTY came along for the ride, setting the standard we are bedeviled with to this day. Sholes even proposed a new layout to Remington that made more sense in the new context, but Remington ignored him.

Third, by drastic contrast, the Dvorak layout was designed and intended from the outset, from the proverbial clean sheet of paper, to maximize speed and efficiency. Dr. August Dvorak did extensive research on typist behavior and on the makeup of English words. Because the typist's natural resting position is on the home row, Dvorak put the most frequently used letters there. That's why with Dvorak you can type over 3,000 words (including many of the most frequently used words) without your fingers ever leaving the home row, while in QWERTY you can only type about 300 words on the home row, many of them obscure. In Dvorak, 70% of your keystrokes are on the home row, while in QWERTY only 31% are, forcing you to the slower and more tiring effort of reaching above or below the home row much more often. Even within the home row, Dvorak puts the more common keys under the strongest fingers (index and middle) and the less common ones under the ring and pinky, while QWERTY forces you to stretch to the corners and use your weaker fingers like the pinky much more often. Dvorak reduces finger movement over QWERTY by a factor of THREE. Dvorak also has intelligent features such as the vowels being on the left side of the home row and the consonants on the right, which speeds up typing by spreading a word out between both hands rather than requiring one hand to do disproportionate work -- it's easier to rapidly tap two fingers in succession back and forth on a table than to tap one finger from one hand as quickly.

Dr. Dvorak's studies showing that new typists get up to 40 wpm in only 18 hours via Dvorak while taking 56 hours in QWERTY have been dismissed with claims of conflict of interest, but the study Dvorak deniers cling to, done by the GSA in 1956, was conducted by a man who had established his anti-Dvorak bias before the study in a publicly quoted statement. Even a relatively Dvorak-skeptical study showed that Dvorak is more efficient:

http://www.webmeets.com/files/papers/lacea-lames/2013/539/keyboard-efficiency.pdf