The numbers don’t lie

Take a look at these three statements. They’re conclusions that myself and others have made about the results of a survey that asked students, among other things, if they support the incorporation of a Starbucks franchise into the new EV3 building.

“Majority of respondents don’t support Starbucks in EV3”

“Majority of respondents are in favour of or ambivalent to Starbucks in EV3”

“Student opinion is split on Starbucks in EV3”

You really can use statistics to prove anything. For clarity’s sake, here’s the actual breakdown of responses:

How do you feel about having a Starbucks franchise in EV3?	Number of respondents	Percentage of total*
I support it	250	42.5%
I don’t support it	236	40.1%
I don’t care	99	16.8%
No response	3	0.5%

*adds to 99.9% due to rounding

Now that you know the numbers, which statement from above is the most fair? I would say the third one best represents the numbers. It’s clear that no one opinion has a majority. So the first two statements above are misleading because they co-opt the “don’t care” category by implying that it supports one of the first two responses.

It’s inappropriate to add the “don’t care” category with one of the other responses and say that the majority of students “support or are ambivalent to Starbucks”. By the same token, saying that the majority “don’t support Starbucks” because the three last responses are not explicitly in support of Starbucks is equally inappropriate.

So when drawing conclusions from this, or from any poll where there is a large block of undecideds, only make statements about the people that have actually made up their minds. So it’s most appropriate to calculate the results as a percentage of decided respondents. This way, it becomes 51% for “I support it”, and 49% for “I don’t support it”. That’s a very different picture than declaring victory for one side or the other.

Side note: this particular set of questions is flawed from the start because the responses are not mutually exclusive. I can not care and at the same time not support a Starbucks in EV3. A better set of responses would be “Support”, “Oppose”, or “Don’t Care”. That way, “oppose” is equally forceful as “support”.

Sam Nabi