NBC/Marist underestimated Obama's vote in 2012 in NH and Nevada

[molova]

Caution when reading a recent NBC/Marist poll claiming that Trump is at striking distance of Hillary.
However close it may be, the closeness may be off by 3% or 4%, judging by NBC/Marist's tendency to underestimate Obama's vote in 2012.

In Nevada, the last pre-election NBC/Marist poll had Obama ahead by 3%. He won by 6.7%
In New Hampshire, they had Obama up by 2%. He won by 5.6%

http://www.realclearpolitics.com/epolls/2012/president/nv/nevada_romney_vs_obama-1908.html
http://www.realclearpolitics.com/epolls/2012/president/nh/new_hampshire_romney_vs_obama-2030.html

[Loki Liesmith]

1. Doesn't matter

Look at the averages

[molova]

2. The averages were off in 2012 too (2% in NH and 2.8% in NV)

Did you look at the averages before advising that I look at them?

[Loki Liesmith]

3. I actually look at the medians from all polled states every hour

My code downloads the newest polls on huff post and computes the median on a ten day interval. Bounces around less than an average.

However, the point is obsessing about bias in one poll is pointless. Every poll has bias and they all largely wash away in the aggregate.