Readtangle.com has a great article on this topic, including a reminder that 2012 was so close that for a hot-second it appeared Romney might lose while having popular vote.
In 2020 they make an interesting point:
---
"The argument that a national election would create more chaos than we have now is, perhaps, the single best argument for preserving the electoral college. Many of these arguments were written before this year’s election, but imagine the case that could be made after the last month. Imagine a world where Donald Trump, contesting the results of the election, did not have to go through the states to overturn results but instead had a federal government body, or official, that could impose its will on his behalf.
Imagine, as others have, a case where the national popular vote was within 100,000 or even 10,000 votes. Imagine the chaos of a full, 50-state recount in an election as divisive as the one we just witnessed. What would happen to the country? How could an incumbent leverage their power to change the outcome? How could the Supreme Court review challenges from a dozen, let alone 40 or 50 states, all with different election laws? What would we do in the weeks or months with no clear outcome?
But no perfect plan to fix it exists. The simplest solution would be to change nothing about our elections except the fact that, when the votes are all tallied and reported, we add them up and pick a popular vote winner. But then we’d need universal laws across the states to formalize the elections. This would require both federal rule and federal oversight. And if the federal government is deciding how states run elections, and then overseeing those elections, the ruling party — the executive branch — suddenly has more control over the outcome than a challenger. This is already an issue in the form of gerrymandering, but we’d essentially be nationalizing that problem. It’d also put us back to square one where we were 220 years ago: we’d face the issues of separating the powers of the federal government and our state governments."
---
The arguments and the history.
www.readtangle.com
That is a compendium of bad arguments. That's not surprising, since all arguments in favor of the electoral college are bad. In response:
1. Note that all other democracies manage to have nationwide votes without the problems identified here. Are we just so bad at running elections that we can't even aspire to Portugal or Brazil?
2. "Trump wouldn't have to go through the states." Going through the states is almost always easier. That's why special interests and the wealthy love to "leave it to the states." State governments are easier to buy and demagogue. It's very unlikely that the nonsense from the Georgia Board of Elections would happen at a federal agency.
We would have to change the constitution to actually abolish it, and while we are at it, we would want to remove election administration from the president's executive powers and put it in a more neutral body.
3. The EC almost certainly encourages far more election hijinks. For one thing, states have an incentive to suppress votes or fiddle with their voting rolls for statewide elections like Senate and Governor. These laws are not motivated by presidential elections alone. So getting rid of the EC wouldn't change the legislatures' behavior.
But it would change the cheaters' behavior. The Electoral College basically shows cheaters exactly where to launch their attacks. Cheating in an election is usually foolish, because the odds of your cheating making a difference is very low and disproportionate to the penalties you'd suffer if caught. If the votes for each candidate are in the range of 70M+, it's really hard to envision a way of cheating that would flip enough votes to make it worthwhile. But with the EC, everyone knows where to cheat. It's not just PA; it's also specific places in PA. It's like a road map.
This is common sense but also born out by experience. Electoral cheating is far, far more common on the local level than state level. Remember NC 9th, and the absentee ballot fraud? Yeah, that was a House race. It wasn't a Senate race. I don't remember if there was a Senate race that year, but there's a reason the cheating was done for the House. 2,000 votes can be the difference, whereas that margin very, very rarely is the margin for a statewide race. You see these sorts of hijinks in local races from time to time.
4. The idea that the EC prevents urban areas from deciding elections is the most pernicious myth out there. The article describes how HRC's margin of victory was provided by Chicago and LA alone. That's one way of looking at it. Another way of looking at it is that her margin of victory was provided by winning 52% of the vote in St. Louis County instead of 45%, or winning New Mexico by 10 points. Another way of looking at it is that her margin would have been even bigger except for Alabama and Mississippi. Singling out those two places as if they and they alone were THE difference is nonsense.
Now, it is the case that campaigns would put a lot of focus on cities. That's because most Americans live in cities! If anything, we should want cities to have more power over rural areas, since cities are far, far more important to the nation. If Canada invaded us and took North Dakota, Montana and Idaho, we'd manage. But if they took Manhattan, we'd be screwed. If they took SF, we'd be screwed.
But anyway, campaigns will focus on where there are gettable votes. Everyone's votes are the same. If the rural party will struggle because cities are dominant, then maybe that party should change its platform to be more attractive to cities! They could do that.
Besides, cities already dominate. Why is Illinois blue? Because Chicago. Why is New York blue? Because NYC (or at least that was true recently). 75% of Nevadans live in the Las Vegas area.
5. We actually would almost never need national recounts. The closest popular vote election of our lifetime, 2000, was still outside the margins that are normally considered recountable. I don't know of any states that allow for a mandatory recount for margins of a half percent. And the bigger the electorate, the less likely recounts will make a difference. That half percent was 500,000 votes. That is miles and miles above what could possibly be the result of voting mistakes. This is basic statistics. If there are 10 tabulation errors in a town of 100, it's imaginable that 8 would be for one candidate and 2 for the other. So the recount could change the result by 2%. But in a population of 100M, an equivalent error rate would probably make almost no difference. Even if there were 10M tabulation errors -- which is itself very hard to imagine -- what are the odds that they would split 80/20? Very, very low. In a voting pool that large, the errors would almost certainly cancel out to a very large degree. We'd probably end up with a net change in four digits, maybe five (i.e. less than 100K) -- i.e. 1/10 of a percent.