Tell Me How This is Not Statistically Significant!

[deckerd]

Some background:

1.

My friends in Jersey, staunch liberals with whom I watched parts of the RNC, were quite impressed
(in a disturbed sense) by the speeches at the RNC -- especially Arnie's "you ARE a Republican" bit,
directed at undecideds and the last of the Reagan Democrats, and Zell's appeal to the last of the
Dixiecrats. They weren't happy about it either -- they felt that as discusting as it was,
it would appeal directly to the target audience and that soft voters WOULD be convinced to
realign with the Repubs. There's a risk that may be where the country
is headed.
I argued with them about it, too.

2.

Undecideds (not so much voting Independents who seem to be staunchly pro-Kerry) have been
giving * second looks over and over, resulting in continual dips and spikes in *'s overall
declining popularity numbers.

Then, in July, Bush bottomed out. No more declining popularity numbers. I noticed this at the time
and thought it ominous back when people here were still trumpeting this pollkatz graphic.

It's not like the polling methodologies of every organization somehow changed in mid-July,
either. That would have produced a statistically noticeable spike. Instead we have a
statistically significant bell curve that is pointing upwards.

3.

The polls where this movement in Party ID is showing up are the ones that fail to CORRECT for
the unexplained phenomenon (i.e. ignore it and "cover it up") by applying sauce. There's
something to be said for sauce ("normalizing" party ID on the assumption that overall strength of the two parties will remain constant and not spike
or trend) but if we were on the other side of this much sauce, we'd be the ones trashing Zogby
and Rasmussen. All the other polls are being criticised for failing to reject "on faith"
a trend in their raw data. The trend needs to be analyzed and determined what is
causing it, across the board, including in the Rasmussen internals, before it can be
legitimately corrected for. I am not saying Gallup doesn't eagerly embrace the results,
mind you. I'm sure they have no interest in looking critically at the source of the trend.

4.

Assuming we are not down to an extent that Kerry MUST ace the debates or have an October
surprise, and I think the polls are off by a scalar margin of about 5% extra Republicans
in every poll, so we're not quite there yet --

What we may be looking at is an election where Bush has bottomed out amongst the VERY
UNDECIDEDS who are supposed to break for Kerry unexpectedly, i.e. the EARLY undecideds.
We could be seeing a major spike in Republican Party ID that, while sure to dissipate
if * wins re-election (how many people say they voted for Nixon?) IS STATISTICALLY ACCURATE and sure to leave a lasting impression on these voters as they
drift back to the Democrats and move the country even further to the right in years to
come. This is exactly what happened in 68 and 94, followed by periods of relatively increased
conservatism (Compare early 90's to late 90's, even among Democrats, or late 60's to mid-70's).

Whether or not we reach November 2 before Bush's actual reelect numbers break 50% is purely a
matter of how large the scalar oversampling factor of Republicans is amongst e.g. young women,
students, and blacks. But it will break 50% again at some point if it has not actually done so,
and past that point the election ceases to be about base turnout and becomes entirely dependent
on an October Surprise or media praise of Kerry in the debates. The latter is MUCH less likely!


5.

When a candidate bottoms out, he gets some momentum back regardless of whether the trend lines
are off on EVERY poll by a scalar correction margin for party ID.

If the Party ID is NOT off, or even just exaggeration of a slight movement of Undecideds
to call themselves Republican (these are the voters who were NEVER energised for Kerry and
never known enough to be ABB, so don't call BS on me just yet.)

6.

The undecideds we are speaking of are EARLY undecideds who can and do very easily break
overwhelmingly for the incumbent. Only LATE undecideds break always for the challenger.
The reasons for this are common knowledge here on DU.

7.

The total number of undecideds, regardless of whether they are likely voters, who are
switching their self-declared alliegance from Independent to Republican (or from Dem to
Republican in the case of hawkish dixiecrats and a certain percentage of "security moms")
could very easily be overwhelmingly larger than the MARGINAL percentage increase of Dems
and "new voters" who are mobilized to vote against Bush, who were not already accounted
for in the last election (new voters in 2000 having since replaced deceased older cohorts.)

In other words, don't get too cocky about the scalar margin of oversampling
Republican-leaning voters which, if it does exist, has been going on all year and is only
INCREASING because Shrub bottomed out in July -- which means he has unfortunately nowhere
to go but up amongst certain early undecideds who have given * repeated second looks, may in
fact feel jilted by Kerry and are rallying to the Republican party -- potentially long-term --
as a post-9-11 security blanket.

This is common psychological pattern among liberal younger voters, especially women,
when they decide conservatism is the "grown up" thing to do in a time of instability.

I don't like it, but it happened in 1968 and if we keep our heads in the sand, it will
happen again this year.

Let's not imagine this can't be 1968 all over again, only this time the EC is cushioning
Kerry somewhat since independent likely voters (mostly socially libertarian whites
who are critical of the war) are concentrated in the battlegrounds.

[unfrigginreal]

1. And your solution is what?

Worry? We've all done enough of that since the scumbags in the GOP stole the last election. We know what we're up against.

Now back to my original question, I reread your post and still don't see anything other than "things are bad, really bad." How about a clue as to what you're proposing that we do, or was worry the intention of the post?

[deckerd]

5. My solution is KNOW what you're up against!

You don't seem to. EVERYONE on DU is counting on a groundswell of
unforseen ABB voters on Election Day. This groundswell could EASILY
EXIST and still be swamped or even strangled in the cradle by a
movement of YOUNG, SECULAR OR MILDLY RELIGIOUS, APOLITICAL, HAWKISH
undecideds to Republican Party self-identification. It happened in
68 and people here are doing nothing to prepare for that CONTINGENCY.

These are the voters ACT is NOT mobilizing. All those groups are
concentrating on base turnout -- just like Bush. * HAS bottomed out
and the Dems job should be to keep these voters from breaking * ONE
LAST TIME, and becoming Republicans on the "fool me twice" theory
about Kerry. This is significant ESPECIALLY if marginal base turnout
increases only just enough to offset the Bush GOTV efforts.

I'm doing something about it by asking people to think about it -- and
start focusing again on soft independents. I don't like these people,
I think they're mistaken but if they break for Bush in September, late
undecideds and unlikely voters turning out to vote ABB will not make up
the difference and we'll end up with a "Silent Majority" like in 68.

It IS too late, mind you, for DEMS to start canvassing to register
non-base voters in states where they have not been doing so
("concentrate on inner-city areas when you go out and register", etc.)

These are the people DUers have been saying "don't worry, if they're
undecided now, they'll surely break for Kerry -- if not, or if they
vote Nader or stay home because they don't like Kerry as a person,
then they're assholes who can't be helped."

[troublemaker]

7. Thank you for a very thoughful post

Unfortunately there are practical limitations on money, effort and time until the elction so I suspect that base-reaching efforts are, hour for hour, dollar for dollar, more effective.

[missouri dem]

2. This graph looks very bad for dummy

If w's approval ratings were for a stock I would short it. He's toast.

[HFishbine]

3. It is VERY STATISTICALLY SIGNIFICANT

There's a little something called wave theory which is very helpful in analyzing trends that can be plotted on a graph. (People use it in the stock market all the time).

Look at what you observe to be Bush "bottoming out." It is, in fact, only his most recent "bottom." There were two previously, each followed by a bump back up, each bump up lower than the previous bump and each followed by a new lower, low.

What a wave theorist will note right away is that a true bottom hasn't formed. What one sees is progressively lower lows and lower highs -- A DOWNWARD TREND. Eventually bottoms form, but they are formed by a "retest" of a previous bottom without setting a new low. We haven't seen that yet, it's coming.

Unlike the stock market though, this dataset has a closed end. There is not enough time for Bush to retest his bottom and begin an uptrend. What I'm seeing is another temporary peak that will soon start down again and probably finish at about 44% - 46% right around election day.

[deckerd]

9. You're both right and wrong. A wave theorist would point out

A wave theorist would immediately note that previous upticks in
* popularity numbers are SPIKES, produced by outside events.

A spike is statistically insignificant LONG-TERM (over the course
of years.) They are not wave forms at all. They merely interrupted
the trend.

But the trend you are looking at is a massive, statistically significant
parabolic curve which depicts an emotional roller coaster ride
of a frightened nation marching in lock-step with each other's opinions.

When Bush bottomed out in July, there WAS NO SPIKE OR UPTICK. He simply stopped
declining. His disapproval rate reached a CIELING. Having failed to keep the
playing field open as the stock market has been (hovering around that floor in
suspended animation awaiting a double-dip drop), what you're looking at is a
self-reinforcing bandwagon effect for Bush.

I see no statistically significant way this will drop back down on its own
if you look at the parabolic nature of the curve.

What you're looking at is a President who, if reelected, will only
increase in popularity over time like dictators everywhere.

It's like battered wife syndrome. The fact that their marriage to *
has been thru trials only makes them more reluctant to let go, now
that the self-reinforcing signs show the worst is over and
"it's safe to be a Reublican again."

The deification of Reagan amongst "moderate" Dems and "liberal" Repubs
in attempted contrast to Bush illustrates this more than anything
else --
there's no way Bush could lose such "negative" comparisons to another
Prez whose popularity curve followed the exact same pattern!

It's called a parabolic curve. It's obvious.

This means Kerry's still being kept afloat by a slowly dissipating ABB sentiment
amongst voters who are beginning to forget the bad things about Bush, and Kerry
was actually much higher in the polls last June than anyone realized.

Hence attempts by the media pollsters to correct it by unsaucing the numbers,
in an effort to show a close race -- back when Kerry undisputibly had the Mo.

There are several solutions:

1. Ride it out and hope that the polls are scalar oversampling by a significant
margin -- i.e. the notion that Bush's support is increasing from a much lower
"floor" than the graph indicates. In other words, shift the entire graph image
down 5% to see the REAL figures and Bush is STILL below 50% popularity, albeit RISING.
Ride out the clock like a strong offense team does in the NBA.

Narrow loss <--> Narrow victory. Exactly like 1968.


2. October surprise/outside event producing a SPIKE in Kerry's favor that the
trend lines do NOT predict. Spike would not dissipate in time, or --
if damaging enough, could alter perceptions and lower the "floor" for Bush,
thereby cratering him.

Narrow loss <--> Landslide Kerry.

Just like 1980 w/ failed hostage rescue,
or 2000 w/ DUI offsetting a stinging Gore loss.


3. Kerry trounces Bush in the debates, throwing opinions up in the air resulting in
stasis and a divided nation like 1960 that had previously been trending Republican.
This would reset things to 2000. Opinion polls open up again like they did before
9-11, resulting in a range of Landslide Kerry <--> Landslide Bush.

As in 1960 or 1980.

Option 3 depends on favorable media for Kerry in tbe debates.

Assuming that Bush's popularity IS on a parabolic uptick amongst the "soft" voters who supported him back in January,

Does anyone else have any additional --contingency-- solutions?

I have provided three.

[HFishbine]

11. I'l tell you what. Post that graph again in two weeks

Even if you have to re-register. Then we'll see who's right.

[unfrigginreal]

13. LOL

:thumbsup:

[deckerd]

17. Agreed.

I apologise for sounding hostile. I am not being Chicken Little.

I'm merely saying be prepared for this contingency and offered
three ways Kerry might go about to correct for it.

I agree base GOTV has to happen no matter what. It'd be cool if
those efforts were unnecessary given how soft voters still appear
more eager to vote for or against Bush based on the past 6 weeks
of their experiences, than disenfranchised Dems based on what he
did in 2000 and 2002/3.

OK, I'll even offer another strategy:

4. Assume the likely voter models are off due to soft Republicans
lying about their eagerness to vote. Soft Repubs and Bush-leaning undecideds
stay home on election night out of complacency or quiet doubt
coupled with a history of not voting. Kerry loses RVs polled on
Election night and wins on the basis of LVs and unpolled ABB voters.

This could well be.

As for re-registering, I'm not partisan enough to worry about that.
It's enough for me that I'd never vote for the Republicans. I don't
care whether or not you think I'm a loyal Democrat for pointing out
what is and is not statistically significant.

I agree there's oversampling going on but it can't be construed as
a deliberate trend increase, like some secret effort to call more
Republican RVs. Unless all the pollsters are in on it. The fact is
independents polled have dropped as self-identified Republicans have
risen. The raw numbers for Rasmussen show the same thing.

Does this mean Kerry is losing -- or is even behind? No it does not.

But it's a trend that needs to be counteracted, like the trend of
hostility to Kerry post-SBVT and the trend of hostility to those
preaching to the choir "quagmire" ads (speaking as a dove here.)

[troublemaker]

15. It is not obvious to me that it's a parabolic curve

It's merely drawn better than previous SPIKES because there are many more polls/week as the election nears.

I consider the Repub convention an event similar to capturing Saddam Hussein and I am not persuaded that the current move is not a spike

[deckerd]

18. Assuming this is a spike,

Then the trailing edge of the spike will go down gradually for * at the previously established rate, which is what has happened in all the previous spikes. Which means the election could be exceptionally close no matter what, and a narrow, mandate-less victory which I'd hate to see.

If, on the other hand, this is a gradual uptick in Bush support that would simply grow if left unchallenged, there's no reason to fret.

An outside event such as a stellar debate performance echoed in the media could produce a spike FOR KERRY that would result in a sharp drop off for Bush, followed by approval ratings all over the map in the last 3 weeks before the election, like they were prior to 9-11. This could result in an "anything could happen" scenario -- potentially the same as 1980 with the challenger winning decisively.

Or an October surprise could crater *, setting a new floor from which Bush could gradually regain support and still lose the election.

[NEDem]

10. if your assumptions are correct

then we should begin to see another dip in *'s approval rating any time now. Look at the graph, it is quite obvious that * is due to drop any day now.

[HFishbine]

12. Yes!

And we will. Weird how graphs can predict events, but this one points to a weak debate performance by Bush.

[HFishbine]

4. Question:

You claim:

The total number of undecideds, regardless of whether they are likely voters, who are switching their self-declared alliegance from Independent to Republican (or from Dem to Republican in the case of hawkish dixiecrats and a certain percentage of "security moms") could very easily be overwhelmingly larger than the MARGINAL percentage increase of Dems and "new voters" mobilized to vote against Bush, who were not already accounted for in the last election (new voters in 2000 having since replaced deceased older cohorts.)

Any emperical data to back that up (I've seen plenty to the contrary) or is it just ass smoke?

[deckerd]

6. You've seen plenty of anecdotal data here on DU

But the actual registration figures, while huge for Dems, result
in marginal increases in Dem registration that are STILL smaller,
numbers-wise, than the shift in SELF-identification amongst
registered independents and conservative Dems that the anti-sauce
pollsters (and Rasmussen in his raw numbers) claim to be picking up.

That's before you factor in any uptick in GOP GOTV efforts amongst
evangelicals which would offset the old-news movement of moderate
Republicans to vote for Kerry or stay home, a trend which has already
played out in the absence of media momentum for Kerry.

[HFishbine]

8. More Specuation without substatiation

The stuff I've seen on DU has been sourced (see below). You, on the other hand, offer nothing but barely intelligible conjecture -- pure ass smoke.

-----

Democrats signing up more new voters in key US states

NEW YORK (AFP) - Democrats have far outpaced Republicans in efforts to register new voters in two key US states, according to the New York Times, which conducted a county-by-county analysis of registration data in Ohio and Florida.


more: http://www.brunei-online.com/bb/mon/sep27w31.htm

[deckerd]

14. The graph above shows a parabolic uptick in * support

Meaning Bush disapproval has reached a CIELING and instead of
trying to keep him in stasis, complacent Dems allowed Bush
to gain the momentum amongst undecideds for the first time in
office.

It is you and other Pollyannas who are blowing smoke out your ass.

The numbers of new registrations are insufficient to outweigh relatively minor shifts amongst soft voters. It is those shifts you need to worry about, not base turnout. If you think we should concentrate on base turnout it implies the base is not energised.

My analysis assumes the base IS energized. Base turnout was already huge in 2000 preventing a landslide electoral defeat for Gore had the numbers stayed the same.

[HFishbine]

16. Tiresome

You keep making assertions with no proof:



- The numbers of new registrations are insufficient to outweigh relatively minor shifts amongst soft voters.

- Base turnout was already huge in 2000

When you start providing something other than your fabricated rhetoric, we can continue this discussion. If not, please come back on Nov. 3rd.

[deckerd]

19. Apparently you feel pollkatz approval index of all polls is insufficient

evidence to establish a trend, especially when scared voters are
marching in impressive -- and frightening lockstep compared to pre-911.

Of course, we could be seeing the results of 5 corrupted polling firms
telling us what the media wants us to hear, but the "uncorrupted" polls
and the "sauced" polls do not appear to be off on their own, showing
some alternate trend the way the state polls sometimes do. They merely
hug the bottom of the trend line like Gallup hugs the top.

[BillZBubb]

20. Good post. Congratulations for asking tough questions.

I agree that Bush* has seen an uptick in his favorables lately. In all the other spikes before there was a single event that produced it. Then it faded soon thereafter.

This spike is a little different. I think it is due to the convergence of three factors: 1) The relentless attack on Kerry as being a weak-kneed flip-flopper has become "accepted" as fact by too many people. 2) Bush*, Cheney, and EVERY Repug surrogate has been hammering away at how much danger we face and that only Bush* is up to the task of protecting us. 3) The school massacre in Russia.

Points 1 and 2 are total bullshit, but by constant repetition the Repugs have made them stick. The school massacre conjured up a worst case scenario for those with children. It fed right into the Repug propaganda.

So, I think this little Bush* surge has more staying power than previous ones. On the other hand, other big events--economic/iraq related,etc--will happen before the election turning people's attention to those issues. Then the Chimperor will slip back.

We are in a real fight here, but it's not going to be anything like 1968. If people pay attention, we'll win.

[deckerd]

21. Thanks for the thoughtful reply! I'm hoping for a late Kerry spike. n/t

[Awsi Dooger]

22. Excellent post; frantic registration is dwarfed by a shift in preference

Math is never a specialty on DU so don't get frustrated. When a thread mentions a 250% increase in new registrations, somehow we interpret that as worth 4 or 5 points and therefore nullifying any supposed Bush lead in the polls.

As I emphasized in a post recently, in high populus states like Ohio and Florida, even the most optimistic registration net advantage for Democrats will only be worth 1/4 to 1/2 percent, not even close to a significant tilt in preference among undecideds. Given the razor margins from 2000 every edge is welcome, but the DUers who are counting on that registration surge are on the same wishful thinking mirage as the ones who believe we can win the election state by state even if Bush takes the popular vote by 1 or 2 percent. In short, no fucking chance.

I was certain Bush's approval rating had bottomed at the apex of the prison scandal when it hit the 43% range. The Republican base is in that area. Still, I was praying it would remain below 50% throughout. The relentless flip-flop theme and the terrorism-as-priority uptick have spiked Bush's numbers and wounded Kerry, at least temporarily.

As an issue, terrorism polls as more than twice the priority for voters as Iraq according to CNN -- 35% to 15% -- yet Kerry insists on focusing on Iraq. We needed economy/terrorism and have received Vietnam/Iraq.

The soft voters who have bought Republican the last two months are primarily women, split among security moms and young "grown ups" as you put it. We probably need 50% of them to return. But retrieving them from an "likeable" presidential incumbent who is on message and with fear as an ally is a greater challenge than Kerry has faced post Vietnam, not exactly equivalent to overtaking William Weld and maintaining your own seat in a Democratic state like Massachusetts.

[deckerd]

25. I hear ya; in fact I didnt realize reg numbers were that small compared to

percentage of undecideds!

[ComerPerro]

23. The most intersting part here

is that Bush's approval never can seem to spike as high as the last time some disaster happened.


But, Thank you for posting this. I have always known, even on 9/11/2001, that Bush's only claim to legitimacy would be 9/11 and any other disaster for which he was responsible but would allow him the illusion of being a leader

[usregimechange]

24. I think the regression line has been slanting downward. nt