Booking.com experimented with it, conversion dropped immensely. Everything they do there is A/B-tested. I wish I could share statistics, but those are documented internally so you'll just have to take this anecdotal evidence for what it is: something a guy on the internet posted.
That said, the reason no large e-commerce websites use it means that it doesn't work, currently. As with all innovation, things won't be effective until users get used to it.
Furthermore, with infinite scrolling come several issues:
- It showed (at Booking) that first-choice picks convert the best. Having people scroll down to completely irrelevant (note: their sorting algorithm was equally well-tested) doesn't make any sense: the farther down people got, the less they would convert;
- In the case of product listings (sorted by price, alphabetically, etc.) people might want to jump to a certain page directly by use of pagination. Infinite scrolling typically doesn't offer such an option for the user, forcing them to scroll, wait, scroll, wait, ..., until they get the page they wanted;
- There really doesn't seem to be a compelling reason (that I know of) to implement infinite scrolling, other than the novelty-factor. People in general understand pagination, if making money is your goal I would not look further into infinite scrolling, and certainly not until the bigger companies start doing it.
In other words, the article that Matt Obee linked to is pretty much spot-on. I'm just confirming separately tested results from a large commercial website, where it had been tested against many millions of users from all over the world.