OK, I did some work on this. Let me try to explain exactly what I did, and keep in mind this isn’t my thing generally. What I wanted to know was basically how many of the biggest metro areas in the US (starting with the biggest and going down from there) would a candidate need to campaign in before they hit the areas that half of the people who voted in November live in. I was also curious how all those areas voted. I knew they’d mostly be blue, but I wondered which weren't, and what the overall totals looked like for the individual areas and the areas put together in one big group.
So I made two spreadsheets and did a bunch of data entry. The first has the 30 biggest metro areas, broken up into counties and totaled. I looked at the official definition of each metro area the government uses (dept of the interior maybe? can’t remember) in terms of what counties are included. Some of the counties I ended up tossing because they seemed to be more rural — ones that had way less than 100K votes and don’t border the county where the main city is. Very easy to do on Politico’s big results map. Seemed to make more sense for what I was doing to include them with the rest of the country, and it saved me some time. The numbers for them weren’t big enough to make a big difference anyway. The second spreadsheet was of the 100 biggest counties, but after comparing the two sheets at this point, they were basically the same results, so I concentrated on the other one.
So for all 30 metro areas (NYC to KC), I went county by county and entered the votes for Hillary, for Trump, and then the total number of votes for both plus any votes for Johnson, Stein or any other third party candidate that were getting around 1% or more. Most McMullin in the states he was on the ballot. Apparently “none of these candidates” is an option in Nevada, and that got a few percentage points. Could’ve just done Clinton + Trump + all other votes, but I was more curious about the more prominent third parties. And again, the numbers of votes aside from those was too small to make much of any difference.
So the national vote total for Clinton, Trump and the main third parties was right about 130 million. So 65 million would be the magic number I was looking for where we could say you’d campaign in front of half the voting population’s metro areas. I didn’t get to that after 30 cities. I believe I had accounted for about 50 million votes after KC. I didn’t want to do more data entry at that point. So instead, I took the vote totals for each city I did, and compared them to their 2015 population estimate to get a rough percentage of how many voted compared to population. Then I figured out the average voting turnout average for those cities (39%), and applied it to each city starting with the 31st biggest to estimate their voting total. The Cleveland area is just over 2 million, 39% of that is about 800K, so I then applied that (803,715 to be exact) to the total vote number that was just over 50 million after KC. Added those up, and didn’t get to half the total national vote after the fortieth biggest metro area (Jacksonville), so I kept going. Checked after the 50th biggest (Buffalo), and found I had just gotten past that 65 million threshold. I was at 65,225,569.
So this isn’t scientific, but I think you’d definitely have to campaign all over the place in a popular vote scenario. You’d definitely have to go to way more places than just bussing around the Rust Belt and Florida, plus a few other states. And keep in mind that those metro areas aren’t all blue voters either. They largely are in the urban center, but not entirely on the whole. For all those counties (I think it’s about 200 or so) Clinton won 60% to Trump’s 37%. Trump won a group of everywhere else in the nation 54% to 40%. Also worth keeping in mind IMO is that if a candidate did only campaign in big cities, it would be a huge opportunity for the other candidate to exploit with everyone else pretty easily. A lot of people in metro areas would look down on not going outside the big cities as well I think.
You can have a look at the spreadsheet if you like. The part I did after the 30th city isn’t on there, I did it on the side because I wasn’t entering the counties and was only estimating their vote totals. I think this link should work.