Friday, April 03, 2009

Net Productivity

I just came across this study from a group of researchers in University of Melbourne. In brief, it claims that those workers who use internet at work for personal leisure activities (in moderation) are on an average 9% more productive than the rest.

Although I haven't read the actual study yet, this post is based on assumption that the University of Melbourne is reporting the study accurately. If so, I have serious concerns over the way researchers used the data to fit their theories. Using the same data, which I believe is a questionnaire filled out by many professionals when asked by the researchers, I can draw many more independent conclusions:
  • People who are more productive at work are more comfortable using the internet at work for personal use.
  • People who are more productive at work are more comfortable accepting the fact that they use internet for personal use in a (supposedly) anonymous survey.
  • The more computer savvy people in workplace are more productive. (This is different from what the study concluded. The study did not take into account whether the same people also use computer more effectively for official purposes.)
As you can see, it is easy to fit many theories on to same facts. Unless they do controlled experiments, such studies should be taken with a pinches of salt. A controlled experiment here would have been monitoring people who didn't use internet at work start using it without letting them know they are being watched. That is, monitoring employee productivity in organizations that previously didn't have internet access (or had draconian laws punishing people caught surfing for private work) start having liberal attitude towards the issue. Keeping workers in dark is essential because if they knew they are being monitored, they naturally start performing their very best.

Without these control measures in place, it looks like an attempt to pass off correlation as causation.

1 comment:

Anonymous said...

Several covariates were included in the analysis to control for potential confounds. Internet "saviness" is not an issue in this country -people don't get hired if they are not "savvy"
Randomness controlled for the effects of computer task difficulty, and in fact most of the concerns you raised are invalid because systematic error was evenly distributed. The study does not claim causation -you are right, only an experimental design can provide evidence of causality. The study finds clear evidence (ANOVA)that those who WILB are more productive than those who don't.