Survey companies like Nanos Research often survey about 1000 people in order to
get a narrow standard deviation on their results and hence significant results. To see
the effect of using a much smaller sample, let’s suppose that only 30 people had been
surveyed.
QUESTION: 

If the survey of whether people would consider voting NDP had been
done on only 30 people this week and on only 30 people a year ago, and resulted in 61%
and 36%, respectively, would this indicate a significant difference over the past year on
this issue at the 5% level?