They were one of the most accurate pollsters in 2012. But the Trump years have not been kind to them.
I’m pretty sure their Clinton +7/Feingold +5 (Oct. 31 - Nov. 1, 2016), Clinton +12/Feingold +6 (Oct. 18-19, 2016), and Walker +1 (Oct. 28-30, 2014) WI polls don’t fall into the Trump years, and they underestimated Republicans in other states as well in 2014, often by substantial margins. When you consistently overestimate D strength, it’s no surprise that you’ll end up being the 'gold standard' in the one election in which Democrats actually do outperform expectations across the board (2012). That said, they were pretty good in 2010 and actually overstated R strength a little in the Upper Midwest that year, it’s just that their performance since 2013 (which is also when they started aligning themselves more and more with D clients, if I’m not mistaken) has been less than stellar.
People would do well to pay less attention to polling in general, regardless of the state/topline. It’s tempting not to (I admittedly got a little excited about those Evers numbers as well), but at the end of the day, betting on fundamentals is probably the better choice. My prediction is that this seat will end up being 1-2%-points more R than the PA race again (in line with recent Senate + House Popular vote results in those two states).
