Paul Grossinger writes on Inc.com about the last US presidential election:
1) In 28 states doing exit polling, 23 states had a margin favoring Trump, against 5 for Clinton. That's 82% of states going one way in a highly divided nation.
2) Among Clinton's 5, only one was above 5% discrepancy, New York. Trump had 10 and a full 13 above 4%. The typical margin of error in exit polling is sub-3%, dating back at least 4 elections.
3) Clinton suffered a significantly above the margin of error discrepancy in these states: Ohio (8.4%), North Carolina (5.9%), Pennsylvania (5.6%), and Wisconsin (4.8%). She also suffered a within-but-close-to-the edge margin in Florida (2.6%) These were the states most critical to winning the election.
4) Of the critical swing states for Clinton, only Michigan (0.3%) did not show a significant discrepancy in favor of Trump.
I don't want to veer away from the data into the realm of speculation, as the purpose of this article is purely to convey the details of the factual discrepancy of the exit polls against the vote results, and to highlight that an 82/18% split in favor, and a 100% split in swing states, is not normal in a closely contested election.
Frankly one can find arguments like this for almost any prior US election: that a systematic (in terms of [swing] states) discrepancy between exit polls and actual results point to potential systemic problems (fraud or at least some systematic error/bias) in the election process itself. (In Grossinger's case he then starts to enumerate potential vulnerabilities of electronic voting machines... and basically asks the reader to put the pieces together, wink, wink.)
As other examples: there was a paper by Steven Freeman that strongly suggested that Kerry won the 2004 election relying at least mostly on the discrepancies argument, saying there was no explanation for them:
in key state after key state, counts were showing very different numbers than the polls predicted; and the differentials were all in the same direction
There was another one by RC Barragan & A Geijsel concluding that Sanders was affected by massive fraud in his primary contest with Clinton in 2016; this paper has a section titled "anomalies exist between exit polls and final results".
Is there merit to this argument that discrepancies between exit polls and vote results point to problems with the election process (rather than say, the exit polls) in US elections?