US Elections 2012 Part I: The Primaries/Caucus

[Mr. DJ]

yay! sanity prevails!

http://www.ajc.com/news/georgia-poli...o-1330300.html

[Ithekro]

Meanwhile, back in 1916...when California was a swing state....

http://en.wikipedia.org/wiki/U.S._pr...election,_1916


Republican/Progressive candidate Charles Evans Hughes' eligibility was questioned due to his father being a British citizen at the time of Charles birth in 1862.

(http://en.wikipedia.org/wiki/Natural-born_citizen)

Breckinridge Long, of the Chicago Legal News, argued that a native born citizen was not natural born without a unity of U.S. citizenship and allegiance and stated: "Now if, by any possible construction, a person at the instant of birth, and for any period of time thereafter, owes, or may owe, allegiance to any sovereign but the United States, he is not a 'natural-born' citizen of the United States."

As far as I can tell the notion didn't stick to anything then. Though had he won, we would have likely gone to War with Germany slightly sooner than we did historically.

[Ledgem]

Originally Posted by DonQuigleone

Also, the accuracy of a sample is based purely on it's size and composition. A 2000 person poll of a single university is just as accurate as a 2000 person poll of the entire USA, so long as those 2000 people are sampled representatively.

No wonder everyone thinks that statistics are a bunch of lies and BS - I completely disagree with that assertion. There's no way to actually make that accurately representative, no matter how you spin it.

[DonQuigleone]

Originally Posted by Ledgem

No wonder everyone thinks that statistics are a bunch of lies and BS - I completely disagree with that assertion. There's no way to actually make that accurately representative, no matter how you spin it.

Sure you can. That's why the sample is so big.

You randomly select a number of zip codes in each state, such that added together those zip codes are representative of the state (EG correct balance between urban/rural, even distribution of wealthy and poor neighbourhoods etc.)

Once you have randomly selected those zip codes, you randomly select a number of households in each zip code.

So let's say, you want 200 households from every zip code, that means you select 325 zip codes from the US to survey, making sure those zip codes are chosen according to a sound methodology.

Constructing a representative sample is one of the age old problems, but it's not as difficult as you might think, and the methodology towards doing it correctly is well documented. The question isn't "can I get a representative sample" but "how representative can my sample be for the cost of my survey?" The US does not have resource constraints like a private polling organisation does, hence they can have such an oncenely large sample size (most surveys are 2-10 thousand people, and you can get a decent sample with as few as 200).

Read a book on statistics sometime. Statistics is used everywhere. Most people don't read up on it, because it's simultaneously really boring.

[Zetsubo]

Originally Posted by SaintessHeart

Now that you have drawn up those examples, it seems that those who built their own American Dream are more inclined to share it than those who inherited theirs.

This.... This... This deserves the Pulitzer Prize !

[DonQuigleone]

If you guys want to know further about how they calculate unemployment, here it is.

It's a bit dry and dull, but what do you expect from the "bureau of labor statistics".

[Kyuu]

WASHINGTON -- At a private three-day retreat in California last weekend, conservative billionaires Charles and David Koch and about 250 to 300 other individuals pledged approximately $100 million to defeat President Obama in the 2012 elections.

A source who was in the room when the pledges were made told The Huffington Post that, specifically, Charles Koch pledged $40 million and David pledged $20 million.

The semi-annual, invitation-only meeting attracts wealthy donors, Republican politicians and conservative activists. Last year, hundreds of activists gathered outside the walled-off resort to protest the meeting. This year, however, the conference went off quietly.

http://www.huffingtonpost.com/2012/0...n_1250828.html

Just another example of big money corruption.

[Ithekro]

I contend that both the Republican and Democratic Parties are corrupted beyond reason and that someone needs to take a stand against them both for the sake of the country. And I don't mean some radical extremist party. We need something that returns us to what we think is the nation's ideals before the corruption took over the government. We don't need a revolution. We need a reasonable Party that can restore reason and honor to our nation.


The problem is...who....how...and to get people to vote for something that isn't the old party lines. Instead of just not voting at all.

[Bri]

Doubt a third party could operate for long without getting corrupted fast. A fundamental issue is the cost of campaigning. As long as lawmakers have to sell their independence in exchange for a chance to win elections the cycle will continue. Banning or severely limiting all political advertising in mass media might be a start to remove unhealthy donor influence.

[Kokukirin]

Originally Posted by Ithekro

I contend that both the Republican and Democratic Parties are corrupted beyond reason and that someone needs to take a stand against them both for the sake of the country. And I don't mean some radical extremist party. We need something that returns us to what we think is the nation's ideals before the corruption took over the government. We don't need a revolution. We need a reasonable Party that can restore reason and honor to our nation.


The problem is...who....how...and to get people to vote for something that isn't the old party lines. Instead of just not voting at all.

Ironically, US is the only democratic country where creating a third national party is all but impossible. The big two have dominated politics for so long that they are both political behemoths ingrained into the system. No new party has hope to rival their fundraising ability. The mainstream media do not pay attention to newcomers. No one can compete on the national stage without the necessary funds and coverage.

In United States of America, money means votes.

[ChainLegacy]

Originally Posted by DonQuigleone

Sure you can. That's why the sample is so big.

You randomly select a number of zip codes in each state, such that added together those zip codes are representative of the state (EG correct balance between urban/rural, even distribution of wealthy and poor neighbourhoods etc.)

Once you have randomly selected those zip codes, you randomly select a number of households in each zip code.

So let's say, you want 200 households from every zip code, that means you select 325 zip codes from the US to survey, making sure those zip codes are chosen according to a sound methodology.

Constructing a representative sample is one of the age old problems, but it's not as difficult as you might think, and the methodology towards doing it correctly is well documented. The question isn't "can I get a representative sample" but "how representative can my sample be for the cost of my survey?" The US does not have resource constraints like a private polling organisation does, hence they can have such an oncenely large sample size (most surveys are 2-10 thousand people, and you can get a decent sample with as few as 200).

Read a book on statistics sometime. Statistics is used everywhere. Most people don't read up on it, because it's simultaneously really boring.

I think what Ledgem is questioning is that a 2000 person poll from a university can be as representative as a 2000 person poll from the whole US. I took stats several years ago so the term eludes me, but doesn't the fact that these people all attend university, had the funds/family/social support to attend, and other related factors of that kind, make it impossible to truly represent the student body the way you could the myriad groups of people in the US who would never step foot in university themselves? The sample size is perhaps most important, but the population size can't be overlooked.

[Anh_Minh]

Originally Posted by ChainLegacy

I think what Ledgem is questioning is that a 2000 person poll from a university can be as representative as a 2000 person poll from the whole US. I took stats several years ago so the term eludes me, but doesn't the fact that these people all attend university, had the funds/family/social support to attend, and other related factors of that kind, make it impossible to truly represent the student body the way you could the myriad groups of people in the US who would never step foot in university themselves? The sample size is perhaps most important, but the population size can't be overlooked.

I never took much interest in stats myself... but IIRC, it depends. (Horrible answer, I know.) There's a bit of circular reasoning that goes in the making of a representative sample. The results are accurate because the sample is representative. The sample is representative because the results are accurate. The criteria you chose to determine that representativity are based on assumptions. Which you can check... a posteriori. And even then, you're not free of luck or other assumptions.

To get back to your question - it depends on what the question is. Does it matter that you go to a certain university or not? Maybe. Maybe not. You'd have to test it.

[monsta666]

Originally Posted by Ithekro

I contend that both the Republican and Democratic Parties are corrupted beyond reason and that someone needs to take a stand against them both for the sake of the country. And I don't mean some radical extremist party. We need something that returns us to what we think is the nation's ideals before the corruption took over the government. We don't need a revolution. We need a reasonable Party that can restore reason and honor to our nation.


The problem is...who....how...and to get people to vote for something that isn't the old party lines. Instead of just not voting at all.

Only a crisis—actual or perceived—produces real change. — Milton Friedman

I doubt any drastic changes will occur unless something really drastic happens like a financial meltdown. Otherwise it will be more of the same business as usual policies. It took the Great Depression for the US to reject the laissez-faire approach to the economy, add banking regulation, implement Keynesian policies and cause a surge in the popularity of labour unions. None of those movements would have gained traction had it not been for the depression. I fear something similar must happen where the status quo and politic system becomes untenable before drastic changes can be made. Another good point can also be gleaned if I finish the Friedman quote I mentioned earlier:

When that crisis occurs, the actions that are taken depend on the ideas that are lying around.

So perhaps the best idea would be to provide solutions and make sure these ideas just happen to be lying around if and when the crisis comes. Friedman applied this philosophy with great success although one must say the means of achieving it were highly suspect if not criminal...

[Ledgem]

Originally Posted by DonQuigleone

Sure you can. That's why the sample is so big.

You randomly select a number of zip codes in each state, such that added together those zip codes are representative of the state (EG correct balance between urban/rural, even distribution of wealthy and poor neighbourhoods etc.)

Once you have randomly selected those zip codes, you randomly select a number of households in each zip code.

So let's say, you want 200 households from every zip code, that means you select 325 zip codes from the US to survey, making sure those zip codes are chosen according to a sound methodology.

Constructing a representative sample is one of the age old problems, but it's not as difficult as you might think, and the methodology towards doing it correctly is well documented. The question isn't "can I get a representative sample" but "how representative can my sample be for the cost of my survey?" The US does not have resource constraints like a private polling organisation does, hence they can have such an oncenely large sample size (most surveys are 2-10 thousand people, and you can get a decent sample with as few as 200).

Read a book on statistics sometime. Statistics is used everywhere. Most people don't read up on it, because it's simultaneously really boring.

Originally Posted by ChainLegacy

I think what Ledgem is questioning is that a 2000 person poll from a university can be as representative as a 2000 person poll from the whole US. I took stats several years ago so the term eludes me, but doesn't the fact that these people all attend university, had the funds/family/social support to attend, and other related factors of that kind, make it impossible to truly represent the student body the way you could the myriad groups of people in the US who would never step foot in university themselves? The sample size is perhaps most important, but the population size can't be overlooked.

ChainLegacy got where I'm coming from.

I'll admit that my only statistical training was biostatistics. Even there, the primary weapon of choice is a t-test, and it's used just for the convenience of saying "look, this difference in data is statistically significant, so we're not making it up." The thing is, the statistical test is just a test. While useful for being arbitrary, it doesn't mean that you can suspend all other forms of interpretation.

In biology, suppose I perform an analysis using 2,000 cells, and find that there is a statistically significant difference with a group of another 2,000 cells. That is a laughably small number of cells when you consider that this cell type exists in the millions per organism, and that there's variation even between organisms. Such data may be published, but it had better be backed by a ton of other data - and even then, I would seize on that particular data as a weak link in the research.

Going back to your example, by looking at people sampled from a university, you might be able to stretch it and say that your findings are representative of opinions held by Americans attending university, but all of America? All 300 billion Americans? We don't even have to limit it to college students being polled, even randomly calling around the United States means that you're getting roughly 40 people per state and are trying to say that their opinions represent a correct breakdown for 300 billion people, spread all over the country, across different lines of work, and all with different experiences?

I don't care what the statistics show. Think about that for a second and I hope you'll agree with me that it's absolutely ridiculous. But it makes for a good news article that sounds credible, I'll give it that.

[Kokukirin]

Originally Posted by Ledgem

300 billion Americans

Just to point out that US does not have ~230x China's population.

But yeah, to claim 2000 samples taken in university, a place at best representative of a small portion of population (young and relatively well-educated), is as good as samples taken from all over the country is rather ridiculous.

[Vexx]

Statistics are only as good as the model used to construct the sampling, i.e.
"university students" do not equal 'American view" no matter how carefully you sample.

edit: and Kokukirin typed faster...

[DonQuigleone]

You're misreading me.

What I mean is that a 2000 person sample is just as accurate for a university, as for the entire country. IE the size of the population being sampled is irrelevant, so long as the sample is representative.

But.

If you took 2000 people from just a university, that sample would be accurate for the university, but not for the entire country, as I've introduced a systemic bias from just sampling from the university. To have a sample that's accurate for the entire country, you would need to draw the sample from the entire country, but the same accuracy would be obtained from the same sample size.

So if, instead of sampling 2000 people at the local university, I samples 2000 people by picking random names out of the national US telephone book, I'd get a pretty representative sample of the entire US. Note though, that I would be only sampling people who own a telephone. So if I wanted an accurate survey of what percentage of the country owns a telephone, that wouldn't really work.

I would never imply that a sample drawn just from a university would be accurate for the whole country. That's absurd. You need to sample widely. For one thing, if you only sampled from a University, 90% of the population would be between 17 and 23!

To improve accuracy, you can also use known information to make your sample more representative. So if I'm doing my university survey, I'll just be pulling people who are walking around, and asking them questions. If I know that 60% of the University is female, I can preselect so that 60% of my respondents are female, making my sample more representative. Likewise, if I know 20% are in the science faculty, I can make my sample more representative by taking 20% of my sample from the science building.

This is what BLS does, they divide all the counties into certain categories based on factors like whether they're urban/rural and other demographic indicators they know from the previous census, and they'll randomly pick counties from those categories in proportion to the country as a whole (EG, if half the US population is rural, half the counties chosen will be rural). They'll also weight certain respondents to compensate in case a group is over represented in their sample. For instance, if Blacks make up 20% of their sample, but the census indicates that blacks only take up 10% of the population, they'll weight black respondents downwards by half, and weight the rest of the sample up by 12 and a half percent.

[monsta666]

Originally Posted by Vexx

Statistics are only as good as the model used to construct the sampling, i.e.
"university students" do not equal 'American view" no matter how carefully you sample.

edit: and Kokukirin typed faster...

As much as we can argue how bad unemployment figures are I would argue the inflation statistics are even worse. If you have the time I would advice you check this video about fuzzy numbers by Chris Martenson. Very enlightening and it's ramifications are pretty significant. I would call it an eye opener especially since the inflation rate is needed when calculating economic growth:

Crash Course: Chapter 16 - Fuzzy Numbers (1 of 2) by Chris Martenson

If the government can skew inflation that much you have to wonder if they make rose tinted errors when applying that sample to unemployment statistics.

[DonQuigleone]

Originally Posted by monsta666

If the government can skew inflation that much you have to wonder if they make rose tinted errors when applying that sample to unemployment statistics.

Inflation is a difficult to define concept. Unemployment is an absolute thing, and can be measured. Either a person is unemployed and looking for work, or they're not. You can't really fudge that.

The only point of criticism of unemployment statistics is that they don't include "discouraged" unemployed. But if you go to the BLS, they do list stats including discouraged unemployed. That news companies don't include those people is their fault, not the BLS.

[monsta666]

Originally Posted by DonQuigleone

Inflation is a difficult to define concept. Unemployment is an absolute thing, and can be measured. Either a person is unemployed and looking for work, or they're not. You can't really fudge that.

True but if you watch the video you will see how inflation is systematically lowered to make it appear that GDP growth is higher than is actually warranted. It is notable that old government statistics actually provided more accurate information on the actual state of the economy than the current statistics we see today.

Basically since the 1990s inflation does not really measure inflation by its skewed system of substitution, geometric weighting and hedonics. Even when calculating the average basket of goods the government - at least the one in the UK - does not use the arithmetic mean but the geometric mean which provides a lower number than normal. In addition the weighting of various aspect of spending is also skewed to make the figure seem lower. For example spending on hospital maybe lowered than what is spent in the actual economy thus lowering the inflation rate. In certain areas they are entirely ignored such as fuel for the car.

The substitution effect is another cause for lower inflation readings and this comes about when the price of a certain good rises too quickly it is then substituted for a cheaper good. The biggest impact on inflation however is how products are hedonically adjusted. What this means is if you buy a faster computer then the government is likely to subtract the price of the computer as it is hedonically adjusted for inflation. More and more items are becoming hedonically adjusted and this has resulted in an artifically low inflation rate.

If the inflation is actually higher than stated then that means GDP growth is lower than stated and that could well mean that if we think we are in a period of low growth the reality could be we are still in a recession.

Granted, measuring statistics is difficult and there will always be flaws but if these flaws always paint a rosier picture than warranted then that should be a cause of suspicion. And these rosy pictures not only paint a dishonest picture of our situation but they also mislead the very people who need to use these statistics to apply economic policies i.e. the government itself. So in the end the measures are somewhat self-defeating. This inaccurate reading means when the stats say we have a recession of -1% growth the reality could actually be that we are experiencing -3% growth. Each situation requires a different economic strategy but by following inaccurate figures we are more likely to pursue inappropriate policies.