Misc blog.
Published on September 30, 2004 By XX In Politics
I always wondered about the polls. News always show them, websites has em, etc.

The polls usually is very accurate and truthful. (I said usually!) However, what is truthful can also lie too. It's kind of like one of those statements that is truthful but intent or content is distorted.

You say, "Hey! How can it possibly be truthful and deceitful?" Basically, it goes like this: For example, a poll may show that 35% of people prefer to eat road kills and 40% likes to eat roadkill only if it has cheese. It may sound odd, but it's truthful. 7 hillbillies loves roadkills, and another 8 hillbillies only eats with cheese. *note: just a example I didn't do any poll or know any hillbillies.*

You can show basically any results by asking right people. Want higher than 80% anti-gay? Go poll only in conservative churches. Want post 80% anti-gun ban? Go to NRA. Any kind of results can be done that way.

Now, how did one of politic related polls show that Left is more popular while another shows that Right is more popular? It is simple but very insidious. Rights just go to richer communities for polls and Lefts goes to poorer areas.

It's impossible to know if a certain poll were conducted widely or pointed to certain group of people to shift the result. Basically, Polls can be a useful tool but people can also use it to lie. In politics it is much worse. Left and Right always wins on popularity all time. How is it possible? Poll shifts by targeting groups.

Politic polls is always worthless. It's pretty much like counting chickens before they hatch.

Object lesson is this: Polls can be useful, but like people, it can lie.

on Sep 30, 2004
Polls are not meant to be Truthful, all they attempt to do is measure and quantify something as ephemeral as public opinion. No poll can be 100% accurate and the margin of error is ofen wide. Yet politicians need an index against which to measure publis perception and hence polls.
on Sep 30, 2004
Yeah. The problem I have outlined is that you could get pretty much any results you want, and politic news, etc usually abuse that.
on Sep 30, 2004

It also has a lot to do with the way a question is asked. For example, on the partial birth abortion issue, if they got a good random sampling, but wanted a more favorable response, the question might be:

"Do you support late term termination of pregnancy if the life or health of the mother or fetus is in jeopardy?"

Note that by using the terms "life OR HEALTH", you make it sound as if it's about general medical concerns, when the truth is, in a large percentage of pregnancies, the HEALTH of the mother could be considered to be in jeopardy, so this question would elicit a fairly strong favorable response. It would change if you asked the question in the following manner:

"Do you support partial birth abortion on demand into the 9th month?"

By combining two "red flag" terms ("partial birth abortion; abortion on demand") for American conservatives, you're likely to get a very high negative response. And yet, you've asked the same question both times.

on Sep 30, 2004
It's worth noting that the most useful polls tend to be recurring polls that ask the same questions to the same population balance. These polls do not so much show national statistics, but change in preferences.

It's also worth noting that many of the political polls are run by independant organistaions which spend large quantities of money analysing populations to determine exactly what mix of people to interview for an accurate portrayal of the national statistics. Interestingly most political polls give a margin of error to them, which is fairly laughable for those who understand such thing.

on Sep 30, 2004
Gideon MacLeish, you hit another way to mislead using poll. It's one that I competely missed.

spend large quantities of money analysing populations to determine exactly what mix of people to interview for an accurate portrayal of the national statistics

The same data could be abused to give the results they want to have.
on Sep 30, 2004
If a poll is scientific, the actual data should be within the margin of error.

I have noticed that many news organizations don't understand the margin of error. When a new poll shows a change of 1% but the margin of error is 3%....you cannot make a conclusion of an actual change. The two polls agree with each other.
on Oct 01, 2004
I wonder if there was any polls that was made to mislead, but weas caught?
on Oct 01, 2004
Margins of error are normally based on the population statistics on which the poll is based. If the correct people are chosen to represent the correct population bands then that is the standard deviation of the error. They do not however cover any form of human error, such as the poll taker choosing the wrong person to represent a certain population band. This is normally the largest problem with such polls. They ask all the questions to determine the population band that the voter sits within, but they can still end up mis-representing a segment. Anyone who understands scientific statistics will tell you that the normal margin of error on statistical data is the square root of the number of measurements taken. So for a population base of 1000 that's about 3%. Strange then how poll margins of error are often this value or less when they have only interviewed this proportion of people.

It's also important to note that the margin of error is only the standard deviation. It does not mean that the real result could be outside this, just that the result is 67% likely to be within the error.