Above the fold, here's the simplified message.
Even if there is no funny business (e.g., Diebold & etc), it is simply not possible to count each ballot perfectly and unequivocally.
Imagine a hyper-close presidential election under direct popular vote. Think Florida in 2000 being played out from coast-to-coast.
Don't we need a way to very closely mirror the popular vote while also containing inevitable problems to a manageable size?
More detailed consideration below.
There are difficulties with 100% accuracy.
As I see it, they fall into two areas. One is inferring voter intent. One is voter registration accuracy.
The ultimate in verifiable balloting is a piece of paper that the voter marks. Without some paper trail, the election is basically unauditable. And if the voter doesn't actually mark the paper, then you've introduced the possibility that the printed record is (nefariously or not) not exactly what the voter intended (due perhaps to the voter's own failure to accurately proofread, but it's inaccurate nonetheless). The problem with a voter-marked piece of paper is that some small percentage of voters will fail to mark their ballot as instructed. True, most of the time voter intent is nonetheless clear. But that's the problem: most of the time.
Then there are registration issues. Voters are in the military temporarily, away at school temporarily, or maybe they've actually moved and just want to keep their registration awhile longer in the "swing state" where they used to live. How to know for certain? Particularly before the election? Otherwise the ballot has already been counted or never delivered. Or perhaps they submitted a registration form within the deadline, left something out or parts are ambiguous/unreadable, it's not fixed prior to Election Day. Whether they're permitted to vote or not, the door is wide open for one side to dispute how the voter was handled. You can likely think of other borderline situations.
Point is, it only takes 1 marginal situation per 10,000 voters and even a state of 2,000,000 voters has an indeterminate outcome if the difference is under 200. And 1 in 10,000 should probably be more like 1 in 5,000 or maybe less.
Florida 2000 times twenty??
I don't think I need to elaborate. I think everyone here knows what Florida was like for a month after the Nov 2000 election. Consider that happening in every county in every state. Fifty Katherine Harris's??? Every single precinct clerk who didn't remember to seal the equipment post-election (even if nothing actually happened)? The microscopic examination of every absentee ballot arriving post-Election Day but being counted with a proper postmark date (which assumes you can even read the postmark)? You know the potential for mischief. We're seen it in "small" doses. Now think nationwide.
A major caveat
Given that I see 100% accuracy as unattainable and given that a razor-thin popular vote is inevitable (agreed that it may only be inevitable in the same way that it's inevitable that a >5 mile wide comet or asteroid is going to coming barreling into our planet someday; but many things in our lives are hugely unlikely, until they happen), I find myself compelled to consider solutions no matter how difficult or unlikely.
This leads me to gerrymandering. I accept that "to the victors go the spoils" is a perfectly acceptable way to handle things so long as the other guy is going to have that option available when he wins. But I believe that we're in agreement that this is actually somewhat like nuclear weapons - if we could put them beyond the reach of everyone, it'd be a good thing. It'd be a good thing for our state legislatures, a good thing for Congress, and a good thing for voters. What's more, it'd be really helpful to the problem at hand. So let's, for the sake of argument, presume that we've wiped out gerrymandering everywhere.
Containment and consistency
As the Electoral College currently operates, it accomplishes one goal (limiting contested outcomes to a single state) but it increases the risk that an inconclusive outcome in one state will put the national outcome in question. The winner-take-all nature of the EC (in 48 states & DC currently) creates large prizes like FL (29 EVs) and OH (20 EVs). For instance, in order to keep FL from being determinative of the outcome, a candidate must win 53% (270/509) of the other EVs. This "lumpiness" of EVs also enhances the possibilities for the EC outcome to conflict with the popular vote outcome.
In contrast, the ME/NE method of allocating EVs reduces the largest "bundle" of EVs to 3, vastly reducing the likelihood that any one "bundle" will be determinative of the national outcome. And it also greatly reduces the scope of any contested outcome from statewide to a single congressional district (in most cases; there would still be situations where 2 or 3 EVs could be contested on a statewide basis). So long as gerrymandering has been eliminated and far more congressional districts are competitive and reflective of entire state in which they are located, the resulting EVs should much more closely track the popular vote than under the current system (which is a very good thing). And it also retains a defacto minimum standard although at a somewhat reduced level (i.e., independents and 3rd parties are excluded unless they can achieve 1st place in at least one congressional district rather than one state) which I take to be a suitable compromise.
So, absent gerrymandering and using the ME/NE allocation, we have an EC methodology which far more closely tracks the popular vote while greatly increasing containment of inevitable inconclusive outcomes. Not utopia certainly, but that's not an option.
So what is one to do in the real world?
Sadly, we don't live in a world without gerrymandering. That being the case, what is one to do?
First, it seems to me that as beneficial as the NE/ME method of allocating EVs can potentially be, at present it only magnifies the potential for damage to democracy. It really saddens me to write that.
Second, recognizing how horrendous a razor-thin popular vote outcome without the EC could be, even with the disadvantages of winner-take-all we should keep the EC because of its one brilliant feature of containing inconclusive outcomes. I can guarantee that if the EC is ever dumped, it won't be back (well maybe; after a catastrophic contested election it may make a comeback, sort of like closing the barn door after the horse is gone).
Finally, we need to double down on efforts to abolish gerrymandering complete with a stake through the heart and a garlic clove in the mouth. Florida has Repub supermajorities in both chamber of the state legislature and a gross gerrymander in congressional districts and legislative districts, but Fair Districts got a state constitutional amendment passed by citizen initiative. Whether the courts reject inevitable Repub efforts to ignore the new rules remains to be seen, but even in this most difficult of circumstances an avenue for progress was found.
So that's it for me. Beddy-bye time. And I'm booked solid through the a.m. tomorrow. But I'll be checking in come afternoon in the EDT.