What moves swing voters

[Berkut]

Of course, the Dems are never going to get 100% message discipline as they can't control what grassroot movements and there will always be some progressive officials. Similar situation for Republicans except that it seems to be okay to swing to crazier and crazier positions.

It does seem like a totally asynchronoous problem. It makes no sense.

It's like the GOP is just expected to lie and race bait and make shit up, so it is just baked into their brand and accepted without consequence beyond what they have already paid.

It makes no fucking sense.

[Admiral Yi]

You and I remember that chanting and the NYT article very differently. I'd be interested to know if there was any polling/interviews done with BLM organizers or prominent members to establish exactly what they want. Of course the GOP doesn't need that level of detail, they need the chanting as is for their dog whistle.

As for Omar and AOC, clearly they don't have as much weight as you think. How many other congress members took up their talking points? 3? How many senators? 1, maybe? The top 3 democrats are Biden, Pelosi and Schumer. You'd be hard-pressed to convince anybody these 3 old geezers are revolutionaries...

I don't see what's wrong about abolishing ICE. It was founded in 2003. Hardly indispensable to the survival of the nation...

I need some clarification.  When you talk about the big GOP lie, what exactly are your referring to?  I took it to mean crazy ass shit doesn't exist at all on the progressive left, whereas on sober reflection I'm wondering if you meant something about the real possibilities of that crazy ass shit becoming real policy.

[Admiral Yi]

Defunding and abolishing the police it must be remembered isn't exactly extreme fringe anarchist insanity as much as it is painted that way.
Northern Ireland shows that this is a workable solution when the established police force is too baked into one side of a divide. In their case religious, so why not racial?
But this is an example of the problem being talked about. The overwhelming majority of left wing politicians have nothing to do with this at all. One or two politicians gave a little thumbs up to the slogans but even for them it isn't a policy they hold.
Yet listen to the media and its adherents and Biden is gunning for killing all the police.

The original texts of abolishing and defunding are exactly as crazy as they're made out to be.

The college-educated white person suburban-friendly reboots are much less crazy.

[Zoupa]

You and I remember that chanting and the NYT article very differently. I'd be interested to know if there was any polling/interviews done with BLM organizers or prominent members to establish exactly what they want. Of course the GOP doesn't need that level of detail, they need the chanting as is for their dog whistle.

As for Omar and AOC, clearly they don't have as much weight as you think. How many other congress members took up their talking points? 3? How many senators? 1, maybe? The top 3 democrats are Biden, Pelosi and Schumer. You'd be hard-pressed to convince anybody these 3 old geezers are revolutionaries...

I don't see what's wrong about abolishing ICE. It was founded in 2003. Hardly indispensable to the survival of the nation...

I need some clarification.  When you talk about the big GOP lie, what exactly are your referring to?  I took it to mean crazy ass shit doesn't exist at all on the progressive left, whereas on sober reflection I'm wondering if you meant something about the real possibilities of that crazy ass shit becoming real policy.

I was referring to "If you vote Biden or any Democrat in office, they're gonna defund the police".

[Admiral Yi]

I was referring to "If you vote Biden or any Democrat in office, they're gonna defund the police".

At the federal level that's probably true.  Small chance of the FBI, or Marshals or Park Police getting defunded.  It's a little more complicated at the city and local level.  After all the Seattle city council did vote to cut funding by 40%, and the mayor did permit the (temporary) establishment of the Seattle Free Zone on Queen Anne Hill.

On a personal note I got a kick out of Queen Anne being the cop free zone because I spent a couple weeks crashing in a buddy's sister's basement during a college summer.

[The Minsky Moment]

If you read the same page you quoted from earlier, but go one paragraph up, you would find out that error bars are exactly the visualization of confidence intervals.  Frankly, given the context of other things discussed, it wouldn't make sense for them to be anything else, even if that paragraph weren't there.

So I did read the page and the footnotes

The footnote says is that the central points are generated by a conjoint analysis and used to generate a marginal mean and then cites to a paper that outlines the concept. https://s3.us-east-2.amazonaws.com/tjl-sharing/assets/MeasuringSubgroupPreferences.pdf.  The body text then uses the word "confident" in relation to the bars, but does not indicate what standard of confidence is being used or how it is being determined (the Leepert, Hobolt paper indicates that it *should* be a measurement of standard error).

There is also a second footnote on the same page that states the following: "In some cases, however, the lack of statistical significance is caused not by the lack of a real difference between two estimates, but rather because we only have a small number of respondents for a given demographic group, which makes our confidence in the estimates for that group lower. In these cases we can often draw useful conclusions from estimates that are not statistically significant, even as we recognize that we should not draw strong conclusions from such estimates."  Now you might say that the size of sample should be and is taken into account in the size of the error bars and that could be true here but if that was so this is very odd way of phrasing the matter.

There are two possibilities here.  Either they fully understand and properly implemented the methodology they describe and are correctly and accurately interpreting its meaning in terms of statistical significance (albeit with the standard left unstated).  Or they are clueless mooks who can't add 2+2 in which case how confident can we be that they did any of the steps of the analysis correctly?

Bottom line is no matter how you interpret the situation, the report provides no basis to conclude that there is a meaningful and statistically significance difference in the "good" progressive populist message and the "bad" woke progressive message among the subgroup of independent swing voters (on page 45), which is the key takeaway of alleged "finings" reported in the NYT article.  If you contend otherwise, please show where I am wrong.

[The Minsky Moment]

Then there was AOC tweeting to "abolish ICE."  Here on Languish Joan and others claimed that she *actually* meant reform the culture of ICE.  How do they know that' what she actually meant?

I don't recall what I said but if what AOC means about "abolishing ICE" is to break up the over-wieldy superagency and parcel out its functions in a more rational way, it's a proposal worthy of consideration.  In fact it makes sufficient sense that it was already done in small part - pieces of the original agency were broken out over the years.

[DGuller]

If you read the same page you quoted from earlier, but go one paragraph up, you would find out that error bars are exactly the visualization of confidence intervals.  Frankly, given the context of other things discussed, it wouldn't make sense for them to be anything else, even if that paragraph weren't there.

So I did read the page and the footnotes

The footnote says is that the central points are generated by a conjoint analysis and used to generate a marginal mean and then cites to a paper that outlines the concept. https://s3.us-east-2.amazonaws.com/tjl-sharing/assets/MeasuringSubgroupPreferences.pdf.  The body text then uses the word "confident" in relation to the bars, but does not indicate what standard of confidence is being used or how it is being determined (the Leepert, Hobolt paper indicates that it *should* be a measurement of standard error).

There is also a second footnote on the same page that states the following: "In some cases, however, the lack of statistical significance is caused not by the lack of a real difference between two estimates, but rather because we only have a small number of respondents for a given demographic group, which makes our confidence in the estimates for that group lower. In these cases we can often draw useful conclusions from estimates that are not statistically significant, even as we recognize that we should not draw strong conclusions from such estimates."  Now you might say that the size of sample should be and is taken into account in the size of the error bars and that could be true here but if that was so this is very odd way of phrasing the matter.

There are two possibilities here.  Either they fully understand and properly implemented the methodology they describe and are correctly and accurately interpreting its meaning in terms of statistical significance (albeit with the standard left unstated).  Or they are clueless mooks who can't add 2+2 in which case how confident can we be that they did any of the steps of the analysis correctly?

Bottom line is no matter how you interpret the situation, the report provides no basis to conclude that there is a meaningful and statistically significance difference in the "good" progressive populist message and the "bad" woke progressive message among the subgroup of independent swing voters (on page 45), which is the key takeaway of alleged "finings" reported in the NYT article.  If you contend otherwise, please show where I am wrong.

I decided to cut to the chase and look at their code on their Github page.  That way I can see exactly what they did, and I can also analyze for myself which differences are significant. 

They do say in three different places that their results can be fully replicated in the code on their GitHub page, which adds to my confidence in their work.  Putting your code that can be fully replicated out on GitHub is a good practice, and says that you're not afraid of someone calling bullshit on your analysis.  Unfortunately, they do not actually provide a link for their GitHub page, and none can be found by any Google search term I can think, which subtracts from my confidence in their work.   

Absent GitHub, I'll just what I said before, and which also seems to be alluded to in your latest link in the section about reference category:  differences can be significant even with error bars overlapping, regardless of how the error is calculated.  No matter the standard, no matter what you're estimating the error of, no matter how you do it, that is still the case.  I can't phrase it any differently.

[The Minsky Moment]

I will repeat my conclusion from the beginning:

Bottom line is no matter how you interpret the situation, the report provides no basis to conclude that there is a meaningful and statistically significant difference in the "good" progressive populist message and the "bad" woke progressive message among the subgroup of independent swing voters (on page 45), which is the key takeaway of alleged "finings" reported in the NYT article. 

[DGuller]

I will repeat my conclusion from the beginning:

Bottom line is no matter how you interpret the situation, the report provides no basis to conclude that there is a meaningful and statistically significant difference in the "good" progressive populist message and the "bad" woke progressive message among the subgroup of independent swing voters (on page 45), which is the key takeaway of alleged "finings" reported in the NYT article.

I disagree that these are the only two possible conclusions.  They stated that they judgmentally interpreted some difference as significant even if the math showed them not to be.  If it turns out that they messed up the math, which so far it sounds very definite to me that they did, then they might have reached the right conclusions with the wrong logic.  If they overruled math that actually didn't need overruling in the first place, then their conclusions can still be valid.  Data doesn't change just because it's not properly interpreted.

[DGuller]

I e-mailed the author for the link to GitHub.  Let's see what happens. 

[HVC]

Odds on a cease and desist letter?

[The Minsky Moment]

I will repeat my conclusion from the beginning:

Bottom line is no matter how you interpret the situation, the report provides no basis to conclude that there is a meaningful and statistically significant difference in the "good" progressive populist message and the "bad" woke progressive message among the subgroup of independent swing voters (on page 45), which is the key takeaway of alleged "finings" reported in the NYT article.

I disagree that these are the only two possible conclusions.  They stated that they judgmentally interpreted some difference as significant even if the math showed them not to be.  If it turns out that they messed up the math, which so far it sounds very definite to me that they did, then they might have reached the right conclusions with the wrong logic.  If they overruled math that actually didn't need overruling in the first place, then their conclusions can still be valid.  Data doesn't change just because it's not properly interpreted.

The null holds until there is evidence warranting its rejection. 

[The Brain]

The null cannot hold.

[DGuller]

I will repeat my conclusion from the beginning:

Bottom line is no matter how you interpret the situation, the report provides no basis to conclude that there is a meaningful and statistically significant difference in the "good" progressive populist message and the "bad" woke progressive message among the subgroup of independent swing voters (on page 45), which is the key takeaway of alleged "finings" reported in the NYT article.

I disagree that these are the only two possible conclusions.  They stated that they judgmentally interpreted some difference as significant even if the math showed them not to be.  If it turns out that they messed up the math, which so far it sounds very definite to me that they did, then they might have reached the right conclusions with the wrong logic.  If they overruled math that actually didn't need overruling in the first place, then their conclusions can still be valid.  Data doesn't change just because it's not properly interpreted.

The null holds until there is evidence warranting its rejection.

It's not that cut and dried in practice.  Despite popular belief, you are allowed to use judgment in reaching conclusions from statistical data.  If the authors on some level felt that the "bars not overlapping" standard was too strict for practical considerations, they would not be automatically wrong in going "we believe that the differences are significant despite wide confidence intervals due to low sample size".  It would be a case of their judgment making up for the overly strict implicit assumption that all categories are perfectly negatively correlated with each other.