Friday, June 22, 2012

Margin of Error

Is trust and collaboration beneficial? Seems doubtful from the results of one group's survey as reported in Forbes. In a comparison of high performing organizations with low performing organizations (as it relates to increase in profits):

  • 41% of survey respondents said they had effective leadership in their high-performing organization, while only 26% in low-performing organizations said the same.
  • 38% at high-performing organizations trusted management, while 24% at low-performers did.
  • 37% at the better performing companies reported highly collaborative environments, while 30% did at the other group of companies.
The survey group shows that these numbers have slid downward dramatically in the past three years. However, it's not compelling evidence that trust, collaboration and effective leadership work because...
  • 59% of high-performing organizations are thought to have less than stellar leadership
  • 62% of those organizations have management that is distrusted.
  • 63% of those organizations practice some less than stellar levels of collaboration.
Now, I believe high levels of trust and collaboration are necessary to succeed in the long term. However, this survey seems to indicate otherwise.

They surveyed 300 companies and only 440 people. The margin of error in this kind of survey is huge. The swing is 7% in one direction. In other words, the real answer to high performing collaborative companies could be anywhere from 30-44%, while low performing organizations have a range of 23-37%. With that much overlap they could be 'equal' in likelihood to have highly collaborative environments.

The number of US companies is 22 million. Assume that half are active, so we can reduce the margin for error. That means we have a total population of 10 million. If we want to survey them for these factors and have a margin of error of +/-2%, we would need to sample 4,159 companies. That narrower margin is necessary to prevent any overlap on the collaborative dimension.

At the proportion they used for their sample, it would be like guessing the composition of the longest ever tapestry/embroidery (Prestonpans) by looking at 300 random stitches over hundreds of feet, and being correct about the proportion of the 21 colors and the hometown of the stitchers based on 200 possibilities.

I'm not disputing their results. I can't. But I am cautioning that their conclusions are not necessarily telling us what's happening in the real world. As much as they'd like to say collaboration, effective leadership and trust lead to high performance, they can't. Not based on this survey.

No comments:

Post a Comment