Polls acting like gas prices?

One of the things I have noticed over the years is that gas prices come down very slowly, but when they go back up, the jump is usually substantial.

4.20, 4,18, 4.16, 4,11, 4.07, 4.01, 3.96. 3.95, 3.94, 3,92, 3,89, 3.87, 3.84, 3.83, 3.81.... 4.79

I have noticed the same thing with much of the Presidential polling. Every time there is a lull, it seems that gravity slowly pulls the race back into the margin of error, then something happens and "boom" - you see massive immediate polling changes in favor of Hillary Clinton. The post convention "bounce" - the post debate "bounce" - plus a handful of other various scandal "bounces" surrounding some major revelation about Donald Trump.

The question becomes, what is the actual "norm" here. Is it a close race with both candidates leading in national polls, and a handful of states within a point or two... or is it a three to four point lead for Hillary Clinton, and the "firewall" states that seem just a few points out of reach.

I know the prognosticator pundits do not like the tracking polls, and they certainly don't like the LA Times polls. But those polls have been much more consistent and have not shown the wild changes that most of your traditional media polls have been showing. The main difference between these tracking polls and the traditional polling is that these tracking polls generally control their demographics, and in the case of the LA Times, they actually control the sample.

In other words, no matter what is happening in the news... many of these tracking polls will sample the same demographic percentages, and the LA Times literally poll the same sample of people over and over. This makes the assumption that the overall demographics of who is going to vote is pretty much set, and any changes to the polls suggests changes within those demographics (improvements within Party, with the opposition Party or better results with independents). In the case of the LA Times, it literally requires someone to change their minds in order to change the ultimate results.

On the flip side, the more traditional pollsters will allow the demographics to move from poll to poll. This suggests that movement in a poll may reflect that the demographics have changes as much as there are changes "within" the demographics. To put this in perspective, when adjusted for demographics (my projection spreadsheet that averages polling within each demographic), all of the post debate movement has moved that number by less than a point.

Now I am not suggesting which is right or which is wrong. My personal opinion is that while weighing for a projected sample will more accurately project overall movement, I do understand that there may be movement in demographics as well. My issue is the proven phenomenon that someone who hasn't changed their mind (about voting and who they are voting for) may still be less likely to answer a survey in the wake of bad press or other negative happening for their candidate (which would explain the large polling bounces that generally don't hold). People like Nate Silver have written about this, while still suggesting that the actual bumps are real.

Bottom line: As of today, Trump still leads in the LA Times tracking poll, the People's Pundit tracking poll, the UPI/CVoter tracking poll. In fact, these three have hardly moved since the debate. Clinton leads by one (down from three a couple of days ago) in the new Rasmussen tracking poll.

So while conventional wisdom once again wants to make this a Clinton election to lose, the overall sum of the whole makes me wonder if what we have seen is another media driven "bounce" that will ultimately fall prey to gravity. As I stated a couple of weeks ago when Nate Silver showed the race virtually even... we are always just a "moment in time"... and the only constant in the election has been change.