Gallup Poll Of U.S. Gun Laws – Small Sample Size?

A record-low 26% of Americans favor a legal ban on the possession of handguns in the United States other than by police and other authorized people.

The results are based on Gallup’s annual Crime poll, conducted Oct. 6-9. This year’s poll finds support for a variety of gun-control measures at historical lows, including the ban on handguns, which is Gallup’s longest continuing gun-control trend.

I was excited until I saw the poll was conducted only over a 3 day period, with a random sample of 1005 adults. :/  I took 3 statistics courses in university, but I squeaked by.  I’m pretty sure surveying 1005 people when the total population is 307 Million is an awfully small sample size.

You can check out the rest of the questions, and the results – HERE

Anyone care to weigh in?



JeremyS October 29, 2011 at 12:08 am

It’s been a little while since I have taken stats, but from what I remember, the key to getting a good prediction, is not getting a huge sample size.

The key to accuracy is to sample a truly random segment of the population. That is not easy.

If you are random, then going over 30 for your sample size is usually enough to get a picture. Since there are three possible responses, they would require more than 30. With 1005, the sample size is definitely enough. What I wonder, is how they sample.


ENDO-Mike October 29, 2011 at 12:13 am

They said they made sure it covered all 50 states though, so right off the bat there’s ~20 people from each state. That’s enough people to accurately represent $307 Million? Proportionally if someone wanted to represent the population of Canada for instance in a survey, would they only need to ask 107 random people some questions?


retro_joe October 29, 2011 at 12:56 am

1000 respondents would result in a margin of error of about plus or minus 3% with 95% certainty (give or take). It would need to be properly distributed but n=1000 is typically considered the standard for national polling and is entirely acceptable. When I wrote and conducted surveys 1000 respondents was my threshold.

Anything much less than 500 respondents can put the margin of error in national polling above 5% for 95% certainty and that’s not acceptable. Although if you pay attention to the presidential race and the various polls some news outlets play fast and loose with this.


Dayid Alan October 29, 2011 at 12:59 am

It’s a loaded question – “Authorized people” is really vague. Technically, we’re all authorized to have handguns, as per the 2nd Amendment; thus, the lack of clarity in the question probably caught some folks with different interpretations of the vocabulary in their scope also. I’d agree with the statement completely: I think police and authorized people (that is: any citizen) should be allowed to have handguns, naturally anyone unauthorized to (on a perma-ban-style list: felons, etc) shouldn’t.


Benson October 29, 2011 at 01:23 am

The sample size could be troublesome, though it’s about the same used to poll political opinions and presidential approval. So there’s precedent for it.

Another way to look at it is, even if the poll had a 10% margin of error, that would mean just 36% would support a handgun ban. 63% would oppose such a ban.

Not too long ago we pro-gun folks would have given our right arms for those numbers.

And on top of that, look at those assault weapons ban numbers. A majority now opposes the ban. If you told that to me in the 90s I’d have suggested you seek mental help.


HMPlatinum October 29, 2011 at 11:06 am

Their “statistics” are wholly dependent upon where they took their pitifully small sample from. You will get WILDLY varying statistics taking polls in rural America compared to urban America.

Don’t believe me? Just look at a map showing national voting trends.

This being the case, it boils down to WHO is taking the poll and WHAT political agenda they’re trying to foist upon the masses.


Ted N(not the Nuge) October 29, 2011 at 11:57 am

Good news if it was a truly random poll. I don’t understand the griping though, usually the polls and statistics are bent and twisted the other way and we end up busy trying to prove things like, “no, 68% of America does not think jackboots are an acceptable fashion statement.”

Still hate the “authorized persons” phrase. .Gov bans em, .Gov better get rid of theirs first.


Frank October 29, 2011 at 12:00 pm

Actually a sample of about 1000 is enough to be representative. Going with a 10,000 sample would only lower the uncertainty from 4% to about 2%(rough figures), 19 times out of 20.

The methodology of the poll ensures that if you were to do it all over again 20 times the results would be the same within +/- 4% 19 times. Pretty standard.


RomeoTangoBravo October 29, 2011 at 09:05 pm

It almost seems like an irresponsibly small sample size…


Borepatch October 30, 2011 at 08:29 pm

Pretty typical sample size. You almost never see a sample size of more than 1500. When you consider they probably had to call 5000 people to get that number, the sample size is most of the poll cost.

Any topic that gets regularly polled will probably have a similar sample size.


A Sociologist October 31, 2011 at 07:57 pm

I am a grad student in Sociology and I TA a stats course as well having taken several advanced stats courses. Like what many people have said here, if the sample was truly random then the sample size is of little concern and we can take the findings to be valid. However surveys have been having hard go of it lately. I think Gallup used a telephone survey here and if that is the case then the results may not be random (how many of you all answer the phone when you see an unknown number, or even if you do would you take time out of your day to answer questions for some random survey?) Telephone surveys are beginning to be seriously questioned in modern social science research


Poppy November 1, 2011 at 12:25 am

Telephone surveys have always been suspect, all the way since the FDR elections.


Jim November 1, 2011 at 07:09 am

I was just reading about this, if I remember where I’ll source it for you later: Gallup found out a long time ago 1,000 people at random is the sweet spot. Making the pool any bigger doesn’t drastically change the outcome but it does cost a heck of a lot more. No sense in paying to poll 2,000 people if the outcome would be within a few points of polling 1,000.


Jim November 1, 2011 at 07:16 am Here’s straight from the horse’s mouth, “Gallup’s Daily tracking process now allows Gallup analysts to aggregate larger groups of interviews for more detailed subgroup analysis. But the accuracy of the estimates derived only marginally improves with larger sample sizes.” Not where I’d read it, but the info is the same.



Older post:

Newer post: