The one-point Clinton-Trump race overall is well within the survey’s margin of sampling error. Combining the last seven nights - across which results have been very stable - the results flip to 46-45 percent, Clinton-Trump, with .4 percentage point gap. Again, it is not a significant difference.
In short, it's not time to panic. It could be one erroneous survey. Which ABC point out they have a track record in...
Although the election is close at this point, vote preference results a week out are not necessarily predictive of the final result. Mitt Romney was one point up against Barack Obama in comparable tracking poll results in 2012, for example, and John Kerry was one point up against. George Bush a week out in 2004.
It's certainly true that over the course of six previous elections, the poll a week out has predicted two results incorrectly.
People have also pointed out that internal polling tends to be more well-funded, and therefore accurate, than public polling.
What real movement looks like in a Presidential campaign vs. tracking polls. https://t.co/oJDq1ec59G