How much more durable, really, are higher end cameras?

Question

I notice this webpage seems to come up pretty often when people discuss the durability of their gear.

http://www.olegkikin.com/shutterlife/

The site has some neat information. It allows people to enter the number of shutter actuations they've put their camera through and its working state into a database. You'd expect the statistics to show that higher end models tend to outlast lower end ones. I checked the gathered data for Nikon's lineup and to my surprise, they indicate exactly the opposite.

According to the data this site has gathered, the average D5000 will outlive a D300 by over 100,000 actuations and a D3 by over 50,000 actuations. I guess a number of things could be going on here:

1. Only outliers are entering statistics. I think the guy who's taken 3 million remote-fired macros of moving train wheels is more likely to document his camera than the guy who takes maybe like a few pictures of his sock puppet collection or whatever on Thursdays. The guy with a bad D3 will go and enter that his camera broke after 9 actuations but everyone else expects theirs to be reliable and don't bother.
2. More people own lower end cameras so more statistics get added for them.
3. Bitter people are entering false statistics because they are spiteful for not being able to afford higher-end cameras

Are these statistics truth or dumb? Even from how little I've handled different models there's just no way you can't notice the difference in build quality between a D3 and a D5000. Maybe there really isn't very big a gap in the durability of these models if they are properly used and cared for.

Answer 1

Another thing that could affect this is that a pro with a D3s may get the D4 as soon as it comes out. Someone not as wealthy may have a D5000 for a much longer amount of time. If the people with the high end cameras replace them more often, they rarely have as many actuactions as the D5000 will by the end of its life.

Answer 2

That site suffers from self-selection bias. It is therefore statistically meaningless. Fun, but meaningless.

For another example of this problem, take a look at The Linux Counter. This site has been going for 18 years in various forms, it has gotten national magazine press several times, and unlike this shutter activation site, its users are almost certainly all happy Linux users. Yet, right on its front page, the top-level stats are clearly very odd:

• more users registered than machines

• that in a world where a Linux sysadmin is likely to be personally responsible for more than one machine, sometimes dozens or hundreds

• a guess four orders of magnitude higher than the measured value

• doesn't jive with measured evidence, such as the ever-controversial Netcraft stats, which has the great weakness that they only count public-facing web servers and can be confused by proxies, so they greatly undercount the actual number

For those who think I'm getting away from your actual question, I'll spell out the application:

• measurement beats out self-reported stats every time

• accurate measurement is often difficult

• those in a position to do accurate measurement often have poor incentive to release the results publicly

• in the end, experts' guesses might well end up being closest to the actual truth