EH.Net Mailing List Archive: EH.Res

EH.R: Path dependency

Douglas Puffert (Douglas.Puffert at econhist.vwl.uni-muenchen.de)

Thu Dec 2 15:34:14 EST 1999

----------------- EH.RES POSTING ----------------- 
Deirdre, Tony, Philip, Jack, John, and Peter: 
 
1. Retraining costs: The empirical retraining studies reported by Liebowitz 
and Margolis find that it takes 100 hours or more training time (4 hours a 
day) on the Dvorak system to catch up to one's former QWERTY speed, and that 
further improvement comes slowly. I do not find this inconsistent with the 
claim that the Dvorak keyboard is as much as 10 percent more efficient for 
those who learn it from the start. What I understand about the neuropsychology 
of trained reflexes is that these become 'hard-wired' in the nervous system in 
a way that takes time and endures long. Thus one's QWERTY skills interfere 
with learning a different system--and continue to interfere for a long time. 
Indeed, it seems to me that the fact that one can catch up so quickly is 
evidence for the superiority of Dvorak; I doubt that those typists attained 
their original QWERTY speeds in anything close to only 100 hours, and their 
new Dvorak skills have not yet had time to become as 'hard-wired' as their old 
QWERTY skills.  (But, yes, it would be better to ask an expert about that than 
to ask me.) It is not clear that retrained former QWERTY typists can attain 
the full advantages of Dvorak--however high they are--, at least soon. 
    A shortcoming of these studies is that they do not give the complete 
skill-improvement trajectory that would be necessary for a cost-benefit 
analysis of retraining. A curve extrapolated from experimental data in an 
Oregon State University study suggests that Dvorak speed reaches 110 percent 
of former QWERTY speed after 165 hours, but I'll quite agree with L&M that one 
can hardly be confident of such an out-of-range estimate. (In the GSA study, 
the marginal retraining benefit for Dvorak typists who have caught up to their 
old (qwerty) speed turns out to be less than the marginal return to a short 
course of additional training for QWERTY typists. As the latter training has 
not yet reached the point of onset of (rapidly?) decreasing returns, this 
comparison says nothing about the relative asymptotic speeds attainable by 
QWERTY and by retrained Dvorak typists.) 
    So, Philip, given that few typists really spend a substantial fraction of 
their work time typing at full speed, and that retraining periods are 
relatively long, I think even the figure of 10 percent superiority can be 
reconciled with your way of setting up the cost-benefit problem. (One 
additional relevant fact here is that the best typists take the longest 
training time to surpass their old speeds.) 
    As to what self-employed typists would do, Jack, some have indeed 
switched. It is not difficult to find their wwweb sites with testimonials. 
 
2. A counterargument: In view of retraining costs, it seems to me that the 
best counterargument to 'my' position (which is of course quite provisional) 
would relate to new trainees. Corporations could establish a 'Dvorak track', 
separate from their QWERTY track, and arrange with high schools and 
secretarial schools to provide original training on that keyboard. This might 
be the more fruitful application of John Howard Brown's entrepreneurial 
experiment. What could prevent such a thing working is transactions costs, 
particularly in giving trainees a sufficient variety of employment 
opportunities. There would be particular problems for trainees who hope to 
move to a different city, who just aren't sure they want to commit to the pool 
of firms that have Dvorak tracks, or who otherwise don't want to cut 
themselves off from the larger world of QWERTY. (This happened to a friend of 
mine who learned Dvorak in a small-city Oregon high school--I think in an 
extension program of Oregon State U.) 
 
3. A qualification: Deirdre, indeed I don't know how firmly established the 
figure 10 percent is, where the lower 95% confidence bound may lie, or what 
problematic assumptions (as opposed to tested human behaviors) underlie the 
ergonomic simulations--which are, as you point out, simulations. I'm not 
committed to the specific number or its possible implications. 
 
4. 'What path dependence was supposed to be all about': 
The answer, Tony, is the historical contingency of current features of the 
economy. Some of these things, as Joshua Rosenbloom and others noted on this 
list back in June and July, do not necessarily involve Pareto inefficiency but 
do affect economic structures and people's lives in substantial ways. There's 
no point in my repeating what I wrote to this point on July 20, but those 
seeking a more authoritative answer might delve into Paul David's 1997 essay, 
'Path Dependence and the Quest for Historical Economics', available on our 
community web site at http://www.eh.net/Publications/ . (I am tempted to call 
this Paul's Apologia Pro Vita Sua--or perhaps pro CV suo.) Paul discusses both 
the interpretation of path dependence and various misinterpretations, and he 
also addresses the pre-history of path dependence to which Deirdre has 
recently alluded.  W. Brian Arthur's views on What It's All About can best be 
found in the introductory sections to his collected papers, _Increasing 
Returns and Path Dependence in the Economy_, Univ. of Michigan Press, 1994. 
    Peter Klein notes that the fact that 'history matters' is painfully 
obvious to the non-economist, but of course Paul's and Brian's point was to 
make this obvious also to economists, who are trained to ignore the painfully 
obvious and see all things converging to a unique equilibrium, independently 
of the dynamics of allocation and thus of history. 
 
Douglas Puffert 
 
 
------------ FOOTER TO EH.RES POSTING ------------ 
For information, send the message "info EH.RES" to lists at eh.net.