AP/NORC, Aug. 15-19, 1058 adults (change from June)
Approve 36 (-2)
Disapprove 62 (+2)
R: 79/20
D: 5/95
I: 31/65
Strongly approve 21 (+2)
Strongly disapprove 46 (+3)
This is Trump's worst showing in this poll since the shutdown.
Possible outlier? Change first appears as outliers. The strong disapproval has changed less than the margin of error, and it is now 10% above total approval. But note well that much of the not-so-strong disapproval now seems significantly higher. People go from neither disapproval nor disapproval or from slight approval to slight disapproval unless one sudden event destroys all credibility in a large constituency.
100-DIS gives an estimated 38% vote share for the President, which is just worse than for Hoover in 1932.
Just looking at the topline figures, one thing that jumps out is that the poll's sample is only 63% non-Hispanic white (compared to 71% in the CNN 2016 exit poll, and 72% in the 2018 exit poll). Furthermore, 44% are Democrats or lean-D independents, while 35% are Republicans or lean-R independents. It's not unusual for Democrats to have a slight edge here, but it's not normal for it to be as high as 9%.
So, the sample looks to be too D-friendly, but I'd be interested to see the crosstabs for this.