The End Of Innovation or How To Lie With Statistics?
References:
Entering a Dark Age of Innovation?, Adler, Robert, NewScienctist.com
(http://www.newscientist.com/article.ns?id=dn7616)
How to Lie With Statistics, Huff, Darrell, , WW Norton&Co, 1954
.1 This is a Joke, Right?
The article in the NewScientist.com about an upcoming paper, had all the trappings of a "tongue in cheek" joke: Someone from someplace respectable proposes something highly controversial by cooking the numbers.
When I read the article, I found it didn't pass the "laugh test".
The fundamental "errors of statistics" so totally bias the statistical results that could have been used to advertise Darrell Huff's book!
But, looking on the web, I found that "controversy" seems to win out over intelligence and common sense. (This was just like the time in Intel when some undergrad at a 3rd rate college in New Hampshire managed to send a mail to Andy Grove requesting a computer for his "important research" and got attention from over 4 levels of management.)
For a bit of fun, you read the article: can you identify the 10 or 16 "problems", each of which means the results are totally useless?
Then, do a quick search on the web and look at what "respected reporters" say.
Below is my summary and analysis.
Have fun.
.2 Summary of NewScientist.com Article "Entering a dark age of innovation"
The article is a "review" of an upcoming article by Jonathan Huebner, a physicist working at the Pentagon's Naval Air Warfare Center in China Lake, California to be published in Technological Forecasting and Social Change.
Based on his insight that "promised advances in technology were not appearing as quickly as predicted", he wondered if "there was a reason for this". "Perhaps there is a limit to what technology can achieve."
Huebner does some statistical analysis on inventions and, voila, concludes that innovation is declining. His conclusions (as summarized in the review) include:
- Using the 7200 "key inventions" in "The History of Technology" and dividing by world population, you get a peak in innovation in 1873
- Currently, in the world, there are approximately 7 "important inventions/billion people" each year
- "Linearly extending the downward curve" moves below the 1400's Dark Ages and hits "zero important inventions" in 2055
- Using US patents in a decade (1990's being the most recent), dividing by US population yields a peak in 1914 (World War I) (although 1990's decade was up significantly over 1980's)
Huebner concludes that:
- "It is more difficult now for people to develop new technology
- "We are approaching the 'dark ages point', when the rate of innovation is the same as during the Dark Ages"
- Perhaps there is a limit to what technology can achive.
- Currently, in the world, there are approximately 7 "important inventions/billion people" each year
- "Linearly extending the downward curve" moves below the 1400's Dark Ages and hits "zero important inventions" in 2055
- Using US patents in a decade (1990's being the most recent), dividing by US population yields a peak in 1914 (World War I)
.3 What's Wrong With This Picture?
I've made the analysis a lot easier by picking out the "results" from the chaff it was embedded within.
My assessment is;
1) "7200 key inventions" - just what does it take to be "key"?
-
First, any book that covers all of history and remains finite size is going to have to do some pruning.Inventions are NOT "key inventions" from day one: it takes a bit of time for the innovation to since in.
- Thus anything in the past 25 year will have the advantage of "recent" and the liability of "impact not known"
- Author's knowledge constraints- what you don't understand, will be hard to judge (See my commentary on "Communication, Mis-communication and Speaking Gibberish"). I doubt the author of the 7200 is terribly savvy on semiconductors, nano machines, bio-tech, robotics, etc. Ex. is the Prius (hybrid car) "one innovation" or "1000 innovations"?
- Small changes in small numbers - classic trick, constrain the data set to make the numbers small, then carefully round up and down to whole numbers to make your point
Bottom line: biased data - the 7200 is a totally meaningless number for "entertainment purposes only"
2) "7 Important technological developments/billion people
- So, only 42 "important technological developments" last year? Really?
- Why were you able to get "important technological developments for last year but
- I bet the criteria for "important" is a bit different today from 1600.
- Why is it meaningful divide by population?
- Why the "whole world" when the "whole" world isn't equally invested in innovation either now or in 1600?
- The rate of population growth has, as I understand it, exploded since WW II. What is the impact of "young people" on the mean age to be productively innovative?
- What is the relevance of a "productivity" metric when your conclusions are "volumetric"?
=> does this nicely turn the graph downward to meet your needs?
Bottom Line: draw the curve, then divide by something that gets you there to generate controversy
3) Patents granted per decade are declining
-
Why drop the recent data: the 1990's are large up tick, why ignore 1/2 of 2000 (which is an even bigger up tick)?
- So, not much was "invented" during World War II but the Great Depression was highly innovative? This doesn't pass the "smell" test.
- Why aggregating by numerical decade? Excel can easily handle that many years.
- Why are all patents of the same "quality" and has the ratio of "quality to quantity" changed?
- Why does he think ROI on filing patents hasn't changed?
Bottom Line: opaque statistical analysis - drop/ignore data you don't like to get the curve of your choice
Now, it could be that the author has all this covered in his paper...
but the summaries don't read that way.
=> They read as if he is about to go job hunting and wanted to get his name known!