I recently co-wrote an analysis of national level polling error in the 2016 election over at The Crosstab. We situate error in national level polls (i.e. estimating popular vote) in the context of historical polling error in the U.S. since the 1980 election, and compare both 2016 error and historical U.S. error relative to national polling for general elections in the UK dating back to 1979. We conclude that national polling has improved and has continued in a strong, high-quality trajectory based on results from the 2016 U.S. election (though the key word here is national–and not state–level polling). Below is one of the key graphs charting polling error for all U.S. elections, all UK elections, and the 2016 U.S. election at each point one year out from Election Day. You can read more about this and the rest of the report here.
Agadjanian Politics 