Is the "trend line fallacy" the most used on this forum?

[I spent the winter writing songs about getting better]

It's not an "official" logical fallacy, but it should be. It basically works like this:

A variable has gone from level A to level B in x period of time. Therefore one assumes it must continue to change in A-B for x periods of time ad infinitum, and eventually will reach a certain point.

We see this quite frequently in arguments about states trending ("This state voted x% one year and y% the next election so it'll obviously start voting Republican/Democratic by *year*) and are seeing it now in the polls, especially the tracking polls (I must admit, the McCain supporters clinging to a 0.07% gain made me very amused and happy.) And yes, it happens on both sides.

[Lunar]

Not the "neighboring state fallacy?" haha

Um, I don't think it's "the most used" even on the daily tracker threads.  People screwing up correlation/causation occur more often, and possibly people setting false expectations ("Obama SHOULD be up this high, at 51%/whatever, so this is actually good news") as well.

[I spent the winter writing songs about getting better]

and possibly people setting false expectations ("Obama SHOULD be up this high, at 51%/whatever, so this is actually good news")

That's pretty much just J. J.

[Filuwaúrdjan]

Probably. Especially in the case of Pennsylvania. For some reason.

[dead0man]

Ya know you just killed the last argument the Global Warming enthusiasts had.

[Lunar]

Ya know you just killed the last argument the Global Warming enthusiasts had.

It actually works if you have established significant causation and the main independent variable can be logically proven to continue its trend (China and India aren't slowing down).

Who the electorate is going to vote for, and polling in general, is a lot more complex however

[Torie]

Probably. Especially in the case of Pennsylvania. For some reason.

A trend from what to what?

[Torie]

It's not an "official" logical fallacy, but it should be. It basically works like this:

A variable has gone from level A to level B in x period of time. Therefore one assumes it must continue to change in A-B for x periods of time ad infinitum, and eventually will reach a certain point.

We see this quite frequently in arguments about states trending ("This state voted x% one year and y% the next election so it'll obviously start voting Republican/Democratic by *year*) and are seeing it now in the polls, especially the tracking polls (I must admit, the McCain supporters clinging to a 0.07% gain made me very amused and happy.) And yes, it happens on both sides.

Sometimes the trend is your friend, and sometimes not. It is a judgment call. When does one extrapolate the line? When it's right! 

[Ban my account ffs!]

I think the

"Crap poll a and crap poll b have shown state X to be competitive.  OMG STATE X IS COMPETITIVE!!!!!"  Then crap poll c and crap poll d show one candidate to be far ahead in state X.  "OMG STATE X IS COMPETITIVE!!!!!!"

argument gets a little old, especially when it comes to my home state.

[afleitch]

I have no problem with longer term trend lines - ie over a matter of months not weeks. If you go on Pollster and remove the crappy polling companies and run a longer term tend line you notice only very gradual changes.

[Jake]

Smash uses this way too much.

[Alcon]

I personally think many (most?) "trends" are over by the time they show up, but that's just me.

[Queen Mum Inks.LWC]

No - it's the straw man fallacy

[muon2]

Trend lines should be suggestive, not definitive. When I see a trend line, there are two factors to consider. One is the statistical reliability of the trend, and the other is the underlying cause of the trend.

One occurrence of the "fallacy" is due to a lack of statistical reliability i the trend line. We as humans want to put data into a pattern. We are naturally good at seeing lines from a disparate set of points. This gets enhanced when someone mindlessly runs a regression analysis to create a guiding line. Too often no one looks at the confidence associated with the line, which may be quite different than the confidence associated with individual points.

If there is some statistical basis to accept the trend I look to see if there is an underlying cause to explain the trend. It's hard to justify projecting a trend forward if the underlying cause is unknown, and this is the second occurrence of the "fallacy". As one example the cause of the trend might not last long enough to continue the trend.  So, if the trend seems to be linked to an underlying cause, then I would ask whether or not that cause reasonably projects forward. If it does, and the statistics warrants it, I might be comfortable extrapolating the trend towards a future conclusion.

[J. J.]

and possibly people setting false expectations ("Obama SHOULD be up this high, at 51%/whatever, so this is actually good news")

That's pretty much just J. J.

I think I've said the converse as well.  And that isn't a "trend line" fallacy.  I'm actually saying that we should factor out some things in looking for trends.

[I spent the winter writing songs about getting better]

and possibly people setting false expectations ("Obama SHOULD be up this high, at 51%/whatever, so this is actually good news")

That's pretty much just J. J.

I think I've said the converse as well.  And that isn't a "trend line" fallacy.  I'm actually saying that we should factor out some things in looking for trends.

McCain up by 0.07 isn't enough to find a trend.

Amusingly arguing it is is the exact opposite of what you argued with jfern on the coin flip thing.

[I spent the winter writing songs about getting better]

This is the worst use of it I have EVER seen:

In 1950, the poverty rate was around 30%.  From then until the late 1960s, the poverty rate decreased pretty much linearly at a rate of 1 percentage point per year to about 12%.  In the late 1960s, we created a bunch of new social programs and LBJ launched the War on Poverty.  At that point, poverty suddenly stopped decreasing, and it is still at the 1960s levels today.  With the track that we were on, poverty would have likely been eradicated in America by 1980 if the government had never intervened in the first place.  

[Coolface Sock #42069]

Not necessarily a fallacy or the most widely-used one. But it's definitely wrong and annoying. If you're going to talk about trends, you'd better have some type of statistical/actuarial background, because if a variable has been at A for a long time and changes to B, the most likely outcome is for the variable to go back to A than to continue to go past B and further away from A.

[Figueira]

I think it comes from the misleading nature of the word "trend."

[Kingpoleon]

It is trending to be more and more common.

Considering the past, it has increased only exponentially and never has decreased, it is doubtful that this will ever stop.

[I spent the winter writing songs about getting better]

A decade later still very relevant.

[Rules for me, but not for thee]

As the ads for investments always used to say:

"Past performance is no guarantee of future results"

[Politician]

Yep. Just look at how Atlas talks about states like IA, WI, MN, ME, RI, etc.

[Gracile]

Trends alone are not the end-all and be-all for predicting election results. However, observable trends over time give a good insight into a region's inclination to vote a certain way. It's certainly more reasonable to follow trendlines than to assume that an area will vote a certain way based on past performance. That kind of reasoning often ignores underlying factors, such as demographic change, that affects shifts in regional partisanship.