« The Fiscal Stimulus: Lessons from Katrina, Iraq, and the Big Dig | Main | This is either very good news or very bad news »
Do people get stuck on QWERTY?
No, and Hossain and Morgan explain their tests:
In this paper, we offer new evidence regarding the economic importance of QWERTY type outcomes. We use laboratory experiments to study platform competition. Experiments have several advantages in studying platform competition: the identity of the inferior platform is clearly defined; the degree to which a platform has a “head start” is controlled; and the “life cycle” of platform competition is reproducible. So far as we are aware, we are the first to study QWERTY in the lab.
We can easily summarize our results: Somehow, the market always manages to solve the QWERTY problem. In sixty iterations of dynamic platform competition, our subjects never got stuck on the inferior platform—even when it enjoyed a substantial first-mover advantage. The remainder of the paper describes in detail the experiments and the results.
This is another theory which probably should be laid to rest. I do think it can explain being stuck in an inefficient language (switching is then truly difficult), but traditional economic examples are hard to come by.
Posted by Tyler Cowen on January 19, 2009 at 07:46 AM in Education | Permalink
Comments
lsw ,jak odappt makkdo; (alh ,jak g; lsk kd;kdh yso jdod) g; kjd ;gkfakgsl ,jdod a mslsrspt g; kd;kdh auagl;k a yoaumdlkdh lsl'mslsrspt maovdkrpaide a ;glupd mslsrspt auagl;k a ;glupd ismrdkdkso g; lsk a odapg;kgi mshdpe
Posted by: nanao at Jan 19, 2009 8:21:08 AM
Laid to rest?! First, "get it right eventually" is not comforting if it takes a really long time. Our limited use of nuclear power fits into this category.
Think of inefficiencies you observe/experience in daily life -- rules in your house, department, institution. Most of the time the explanation is these made sense to someone at some point and switching was not and still is not worth it. I don't think their model is particularly good at showing anything -- the fact that they make people reevaluate and make the costs apparent to people is not great (I skimmed). People know with some certainty the benefits of the dominant technology and have a limited sense of the benefits of the superior technology.
Could a facially reasonable experiment be built in which the dominant technology did not win? Yes.
Posted by: NiceGuyAddie at Jan 19, 2009 8:34:43 AM
Perhaps Alex's new language should coincide with the creation of a libertarian society for The Exodus.
Whereas an ultra-efficient language entails switching costs to those embedded in a legacy culture, it may be a draw towards joining an overall more efficient culture.
My little addition is that there will be no time zones (or other nonsense like daylight savings): "One-Time"
Posted by: Andrew at Jan 19, 2009 8:34:48 AM
It's funny because there's no conclusive evidence that the DVORAK system really is any better than the QWERTY system.
Posted by: Jason at Jan 19, 2009 8:46:24 AM
I have no love for the QWERTY theory, but I'm a bit puzzled by this paper. The argument for the QWERTY capture is that you build up a skill set learning QWERTY and this keeps you using it in the future. In this paper, I can't see any learning advantage to knowing the earlier platform, the only difference is in terms of money.
Posted by: Tracy W at Jan 19, 2009 8:53:51 AM
Is the ocean really deep? We recently laid this theory to rest by study in the laboratory. In our simulated oceans, the temperature is clearly defined, and the number of fish per cubic meter was unambiguous and reproducible.
We can easily summarize our results: Somehow, the market ocean is always no more than 10 meters deep. In sixty iterations of filling a tank with seawater and then measuring the depth, our fish swam around but never went below a depth of 10 meters. The remainder of the paper describes in detail the experiments and the results.
Posted by: Curt Fischer at Jan 19, 2009 9:09:17 AM
What Curt said.
Posted by: Russell at Jan 19, 2009 9:31:31 AM
People everywhere, and often on this blog, give scientific studies a lot of flak. Curt and Russell are a great example here. Yes, many studies are limited in scope. Yes, many studies are not optimally designed for any number of reasons (budget, time, feasibility, talent). Nonetheless, a study reaches a conclusion, and the author must describe that conclusion.
That doesn't mean we should ever answer one of life's many questions with one study. Nor, I think, do most scientists ask you to accept their study as the answer to life's many questions.
However, if we spent the time and resources crafting the ideal study every time, very little would actually get studied. The fact is you often NEED many inferior studies to develop a knowledge base, learn from past mistakes, and justify the costs related to a superior study.
Posted by: Jason at Jan 19, 2009 9:47:24 AM
Jason (at 9:47:24 am), I think Curt was reacting against, was Alex and the authors seeming to say (I'd have to read the paper to know for sure) that they just demonstrated the actual QWERTY keyboard is efficient. But obviously, that's not at all what they tested (again, I haven't read the paper).
For my own part, I always thought the claims of professional typists being able to wail on a DVORAK (or whatever) keyboard were odd. If they were true, then why didn't temp firms and other appropriate employers pay for a limited batch of the "efficient" keyboards, that their secretaries would take around with them? Or in a big firm, why wouldn't they have a training program for people on a secretary track? It just seemed that if the advantages were as great as its proponents claimed, somebody would figure out a way to exploit it. As Alex alludes to, you personally can use a different keyboard even if everyone else uses QWERTY.
Posted by: Bob Murphy at Jan 19, 2009 10:01:53 AM
Whoops, Tyler wrote this post. Somebody in the comments referred to "Alex's new language" and so I got confused. Lock-in and all...
Posted by: Bob Murphy at Jan 19, 2009 10:02:49 AM
Nonetheless, a study reaches a conclusion, and the author must describe that conclusion.
Good studies have conclusions, true. But any study worth its weight in salt has a conclusion that is follows from its contents. This study (at least as excerpted by Tyler) overreaches because it ignores the elephant in the room: the QWERTY problem is in fact not solved in the real world.
I might put it as:
"Our simulated market always manages to solve our simulation of the QWERTY problem....we suggest the difference between our simulations and the real world stems from these X particular parameters in our study design. We hope someone will study the importance and effect of these parameters in more detail later on."
Posted by: Curt Fischer at Jan 19, 2009 10:11:29 AM
What Curt said x2.
Posted by: Andrew Edwards at Jan 19, 2009 10:14:17 AM
Curt,
Which elephant? Is there evidence that QWERTY is in fact substantially inferior? (Evidence that wasn't produced by Dvorak, I mean ...)
Posted by: ryan at Jan 19, 2009 10:26:08 AM
The study was interesting, but it would be interesting to incorporate sunk costs(Learning how to type QWERTY, buying a blue ray player, etc). This, to me atleast, would be the main driver of QWERTY effects.
Posted by: David Shor at Jan 19, 2009 10:26:15 AM
I recently saw this paper presented at the AEA's. The author's stressed the paper was a test of a value-added type hypothesis. "I will use QWERTY because it was introduced first and I expect everyone else to use it as well." Issues like switching costs were not examined, and I'm sure the network effects literature has much more to say on alternative hypotheses.
Posted by: Ryan Webb at Jan 19, 2009 10:26:27 AM
People, keep in mind that is not a stand-alone result. The QWERTY claims, in real world settings, have been refuted for almost twenty years. This is icing on the cake. There may be other applicable examples, but at this point the burden is on the other side.
Posted by: Tyler Cowen at Jan 19, 2009 10:27:32 AM
I didn't notice any network effect in the study. Similarly, QWERTY has no network effect -- you can type an email to someone with a DVORAK keybd and they can read it and type back.
Beta/VHS, hardware platforms (e.g., Intel and compatible, OSs (e.g., Windows), applications (e.g., Word, Excel) all have network effects that dominate once a market position is established. So do human languages....
Switching costs are very, very high when your customers expect Microsoft Word documents in English. The penalty for a not-quite-perfect compatibility is a potentially devastating loss of credibility -- there's no point risking it even if Mac is better or Google-whatever is free.
OTOH, they don't really care if you typed it on a DVORAK keybd. Yet that switch has not been made.
Posted by: Anon E. Mouse at Jan 19, 2009 10:30:04 AM
I don't know why so many people are beating up on laboratory experiments here - the growing number of experimental results + the theoretical arguments (Leibowitz) provide a fairly convincing case that lock-in is quite rare and generally does not prevent a more efficient technology (including language) from taking over. Most of the "real world" examples of the QWERTY problem are controversial - what one person sees as a clearly inefficient standard that persists, others see as more efficient (e.g., Microsoft, the keyboard, etc.). The fact is that these are multidimensional products and it is never clear which is the more efficient technology. Where else can we look for guidance but under laboratory conditions, where we can unambiguously set up conditions where we know the efficiency ranking?
Posted by: dale at Jan 19, 2009 10:32:30 AM
This thread seems to have engendered confusion because two different topics are being discussed.
1. Is the QWERTY keyboard worse or less efficient than many easily available alternatives?
2. Grant that the QWERTY keyboard is less efficient, and call all situations where a less efficient alternative becomes entrenched in the marketplace QWERTY problems.
I was trying to talk about topic #2. If the real-world QWERTY keyboard is not in fact less efficient than its alternatives, a relevant parameter "X" (referencing my 10:11:29 comment) would be the magnitude of the efficiency difference used in the simulations as compared to the real world.
Posted by: Curt Fischer at Jan 19, 2009 10:43:03 AM
What nanao said (perfectly understandable to Dvorakians):
"no, what really matters (and what is not tested for here) is the situation where a monopoly is tested against a fragmented non-monopoly marketplace. a single monopoly against a single competetor is not a realistic model."
Posted by: lalas at Jan 19, 2009 10:56:51 AM
Can we acknowledge the dead horse on the table here? QWERTY is a bugbear for ideologues - they feel it might be a damper on the efficient market theory and so they fight it tooth and nail, often offering some really poorly designed studies that supposedly refute the advantages of Dvorak.
The whole discussion is a proxy, which makes it pointless, and there's no point going down this rabbit hole - the ideologues are simply signaling here. It would be better to be truth-seeking, even if it offends one's long-term political identity, folks.
The larger issue is that the keyboard sucks as an input device, and it would be better to work on improving voice and BCI. Especially BCI - there are companies now in a race to bring these to market by this summer, namely Emotiv & Neurosky.
Mindball is already in the market, but not for home use. NIA for gaming is already in mass-production and will certainly explode once connected to MS Office.
Posted by: StreetWalker at Jan 19, 2009 11:00:41 AM
So, this study did not in fact study Qwerty keyboards in an experiment, but a class of problem called QWERTY problems, even though QWERTY itself probably does not fall in the category?
I mean, why make it less confusing if it can be more confusing?
Posted by: Zamfir at Jan 19, 2009 11:02:16 AM
This reason article from 1996 does a pretty good job of examining the evidence on the "DVORAK is superior" thing:
http://www.reason.com/news/show/29944.html
The result is that DVORAK had not been credibly proven to be superior for typing speed or ergonomics. My google-fu doesn't reveal anything to have changed since then in regards to studying the topic. I wonder if the real measure of how fast you can type isn't much more related to how your brain handles the key strokes than to how much your fingers need to move (meaning sub-optimal layouts might matter a few percentage points, but nothing significant).
Posted by: Wes Winham at Jan 19, 2009 11:38:12 AM
The Paul-David story works well when adapted to culture, particularly the culture of government organizations.
I adapt the theory in the following paper:
http://www.gmu.edu/departments/economics/klein/PdfPapers/IfGovtVill.pdf
In other words, this failure theory applies better to government than it does to free private enterprise.
Posted by: Daniel Klein at Jan 19, 2009 12:00:36 PM
Actually, lock-ins are pretty common, and if you need a more recent example just go look at the alpha keyboard on your cellphone and bear in mind that it was designed in order to provide a mnemonic aide to remember phone numbers but is now largely used to write text messages, a task for which it is poorly designed. The main reason why this topic is not laid to rest is because far too many economists are unable to put their ideological horse blinders aside and try to understand the problem at hand rather than use it as ideological gunpowder in the "Are markets really always efficient?" culture wars. I'm inclined to say that the last useful contribution to this topic came from Schelling in 1978, when he pointed out that the QWERTY keyboard is inefficient because it requires touch typists to use the left pinky finger to type the letter A. Everything after that is simple ideology-pushing, including (but by no means restricted to) Liebowitz & Margolis' much-heralded but really piss-poor analysis.
Lock-ins occur when the social returns of a technology shift are positive but the private returns are not large enough to warrant individual investment under a real possibility that new technology adopters will remain in the minority and thus will be excluded from the network benefits. Simply put, if a new technology offers a productivity gain Λ social and private returns are aligned and universal adoption occurs, but if it offers λ < Λ where social and private expected returns diverge even if λ > 0 and social returns from coordinated adoption would be positive. Examples for this scenario abound. The λ tends to be small in percentage terms (Liebowitz-Margolis estimated the productivity gain of DSK to be around 3-5% rather than the 15% touted by Apple and others), but a simple 3% productivity gain can quickly turn into a very large number economy-wide. If this is the only problem to be solved public incentives can be successfully used to drive universal adoption (as demonstrated by the tax-subsidy scheme created to drive adoption of unleaded fuel in the EU in the 1990s), but in a dynamic technology regime a static comparison between generation 0 and generation 1 is rarely sufficient to make efficiency statements: it is very possible that the adoption of generation 1 will in turn preclude the adoption of generation 2, etc. Iow, it is sometimes better to skip a technology generation even if it is superior to the status quo.
Oh, and regarding the experiment: I didn't try to figure out the incentives from the payoff matrices, but the big jump in initial adoptions after the switch to the duopoly stage is a strong signal that they picked a gain close to or above Λ, while the interesting question would be which equilibria emerge if the initial rate of adoption is at 20% or lower (i.e close to λ). Another opportunity wasted.
Posted by: ogmb at Jan 19, 2009 12:10:59 PM
