In 2000, we had a split between the popular vote and the electoral vote. The state that made the difference, Florida, had a difference of only 0.5 points between its vote and the national margin. In 2004, the difference in margin between the national vote and the key state of Ohio was, likewise, only about 0.5 of a point. So, then, doesn't that mean that a 2-point split between the state that is likely to decide the election and national vote is very unlikely? Not exactly.
The reason most of us don't think a 2-point split is possible is because our election dataset is so small. Expanding our dataset, we can examine the last electoral vote and popular vote split before 2000. Grover Cleveland won the 1888 popular vote by about 0.8 of a percentage point, but lost the pivotal state of New York and, thus, the electoral vote to Benjamin Harrison. What was the margin in New York? 1.1 points. Thus, the difference in margin between the tipping-point state of New York and the popular vote was 1.9 percentage points – or nearly 2 points.
Math indicates that another 1888 is quite possible.