Remember that from that 75% is supposed to be things like insurance policies and such, and I suspect that currently drivers aren't bothering to properly insure their cars (using it for a business rather than personal use). In a driverless car case, someone is going to have to pay for that insurance, whether the manufacturer, operator or rider, that cost is going to come from someplace, and will probably be significant. I'm not sure that you could achieve 50% savings, and maybe not even 25%.
"All it takes is half of Manhattan's car-owning households to switch for Manhattan to hit a 7-fold reduction in car ownership from present-day America." But what does that prove? Manhattan isn't a great barometer of anything. Even if they did drop to 14% or so, that's not really moving the needle much either way, is it?
The fleet provider (e.g. Uber) would surely self-insure. Insurance for self driving cars will be extraordinarily cheap (at least an order of magnitude less than driver based insurance).
10x is admittedly bullish. I expect 4 to 7x within 20 years, i.e. 3 to 4 product cycles. But 10x is more reasonable than it seems at face value.
Let's use Manhattan to illustrate. Another HN user (wadenick) kindly provided [1] some references I'd like to point to. Car ownership across the United States stands at 81% of households [2]. In Manhattan it's 23% [3]. Manhattan has roughly fewer than 1/3 the number of cars as the rest of America. This is partly why there are only 39 gas stations in Manhattan (down from 60 in 2004) [4]. Repair shops, too, are fewer and further in between. Many are owned by taxi operators for their fleets' use only.
Uber pays 75% of its earnings to its drivers. How much cheaper could a driverless be? 50% less? How would that tilt our 23% figure? All it takes is half of Manhattan's car-owning households to switch for Manhattan to hit a 7-fold reduction in car ownership from present-day America.
Everywhere isn't Manhattan. Density matters. Less-dense locales will need to accept some combination of higher wait times, higher prices or lower rates of fleet reduction. That said, the scaling economics are nearly identical. Fleets will be cheaper than individually-owned self-driving cars because they can more efficiently use their down time. This, in turn, makes individually car ownership more expensive (by degrading retail-servicing infrastructure and supply chains). That, in turn, makes fleet usage more desirable and so on.
The degree to which this happens may vary. But at its lower bound, I'd posit 4x and at its higher bound over 10. Given that income and urban dwelling are correlated, and continue to increase in their correlation, that bodes badly for the edge-case rural families demanding high degrees of simultaneity at a price even remotely approaching what it costs to maintain a small garage of cars today.
[1] https://news.ycombinator.com/item?id=10837732
[2] http://web.archive.org/web/20140209114811/http://data.worldb...
[3] http://www.nycedc.com/blog-entry/new-yorkers-and-cars
[4] http://www.wnyc.org/story/say-goodbye-manhattans-gas-station...