I understand that one motivation behind not providing crime breakdowns by ethnicity is to avoid providing information that might be used without context to justify hate.

But it is also a little racist to try to hide the problem, as if criminality is inherent to the ethnicity and cannot be solved, so may as well be ignored.

Quite true - and the level of Orwellian thought-control that bureaucrats are engaging in, in order to hide facts because it might "justify hate" - is much more terrifying than the hate itself.

But pointing out the positive statistics of immigrants (e.g. https://hbr.org/2016/10/immigrants-play-a-disproportionate-r... ) - is okay?

Ignore the data supporting one side of the debate, and of course the other side will seem more convincing.

Collecting data about ethnicity is always problematic because the definitions are fundamentally fluid.

Finnish people were considered Asian in some American states in the early 20th century. Should they be included in crime statistics with Chinese?

Also, if police are required to collect this data, who else is? Should you be able to look up the registered ethnicity of a person similar to how anyone can look up tax returns in Sweden (and other Nordic countries)? If not, why not?

Trump got less votes than Romney.

Your backswing is just the same old bigots getting a momentary upper hand because of the way the chips happened to fall in the last election. Whether they can leverage that into more lasting power remains to be seen.

Not sure what stats back up this claim. I've heard people say this movement is anti-globalization, about the economy, anti-PC, anti-establishment - yadda yadda yadda. Everyone has a theory, and it seems like people just choose the one they like best.

Trump at the moment has a very low approval rating - maybe not everyone got what they read into him?

>Trump at the moment has a very low approval rating - maybe not everyone got what they read into him?

Polls also predicted him to have less than a 10% chance of winning the election. Not all data is good data.

If the polls had said "0.1% chance", I would agree with you. But a 10% chance is actually quite high odds. Things much less likely than that happen to everyone, every day.

The predictions were essentially right. Hillary Clinton had a better chance of winning the election, but Trump successfully rolled the dice with his 10% odds: he campaigned in some critical states that were expected to go blue and won them over by a narrow majority.

That's how people interpreted the polls. But there were a lot of people who said that assigning a probability to the election was a foolish thing to do.

Also worth noting that accurate national polls wouldn't have captured a Trump victory (since Clinton lead in the popular vote). You would need state polls, but they're not always accurate. Trump's win in Michigan was a big surprise, but so was Sanders' win in Michigan a few months earlier.

So now I'm confused. Does a low approval rate matter, is it "bad data"? What defines "bad data"? Say the statistics running around this thread specifically about crime in countries, broken down by race and type or crime, or broken down further by region within counties too. How does one determine "good data" vs "bad data" here?

Sorry you're confused. Point was pretty clear: US polling data was not found to correlate strongly with reality, so treat it with some suspicion. Can't really break it down further than that.

The polling data was that it is possible for it to occur, although unlikely. It occurred. I'm not sure where the data said otherwise, or where it isn't correlating with reality.

So saying an event has a 10% chance of happening, and then it happening after one roll of the dice, is enough reason to doubt the data?

Nope.

Polling was incapable of accurately determining what the odds really were. If that die was six-side, but I told you there was a consistent 10% chance of rolling a 1, then I'm providing you with a bad probability. The media and pollsters did the same (for whatever reason).

>Polling was incapable of accurately determining what the odds really were.

Nope.

If polling couldn't determine what the odds really were, the probabilities, which showed it was well within the realm of possibility that trump would win, aren't evidence the polls were out of alignment with reality.

Yes :)

Semantics at this point. I'll rephrase your comment with consistent terms:

If polling couldn't determine what the odds really were, the odds, which showed it was well within the realm of possibility that trump would win, aren't evidence the polls were out of alignment with reality.

The possibility of Trump winning does not exonerate the accuracy of the odds presented by polls. The odds were not correct, and the polling data was flawed.

>The possibility of Trump winning does not exonerate the accuracy of the odds presented by polls. The odds were not correct, and the polling data was flawed.

If there are 10% odds of something happening and it happens after one event, that's absolutely not a reason to say the data points that generated the odds are flawed.

That is the reason you presented.

It's not.

I'll clarify, to help.

The reason not to trust polling data presented today is that the probabilities presented in the past were not accurate.

Based on what data?

Clinton ended up with a 2.1% victory; the RCP polling average before the election was 3.2% in favor of Clinton. So a 1.1% difference, which isn't huge. Some of that might be explained by the Comey announcement as well.

FiveThirtyEight pegged Trump's chances more in the 25%-30% range. Not all data is analyzed well.

Most analysis showed he had some chance of winning, and he won by a relatively low margin.

If you rolled a ten sided dice once and it came up with a ten, would you feel repeated tests that showed there was a 10% chance of that outcome should be ignored in future?