Statistics Tip of the Week : Sample Sizes and Margins of Error for Proportions

​This may be a handy table to keep around somewhere. How big a Sample Size do we need if we want to differentiate between 2 choices in a survey or election? It's more than people usually think.
 
Some might have in mind the guidance on when to use the t Distribution instead of the z (Standard Normal) Distribution. We're told we can use z when n, the Sample Size, is "large". And then we learn that some consider 30 to be large enough, while others say 100.
 
But as you can see from this table, n = 100 barely gets you into the game when you're doing a survey or poll. When n = 100, you have a 10% Margin of Error (MOE). That is, you can say that you have a Statistically Significant difference if your Proportions are wider spread than 44% and 55% for the 2 candidates.
 
But to get to a 2% MOE, you'd need a Sample Size of 2,400. Notice also, that diminishing returns set in. To get to a 1% MOE, you'd need a sample 4 times larger than you would for 2%.