A past data set was collected from UCLA courses in Spring 2010, and Amazon
at that time was found to be almost uniformly lower than those of the UCLA
bookstore's. Now in 2018, the UCLA bookstore is about even with Amazon on
the vast majority of titles, and there is no statistical difference in the
sample data.
The most expensive book required for the course was generally used.
The reason why we advocate for using raw amount differences instead of
percent differences is that a 20% savings on a $10 book is minor relative
to a 20% savings on a $100 book, meaning a small and largely insignificant
price difference on low-priced books would balance numerically (but not in a
practical sense) a moderate but important price difference on more expensive
books. So while this tends to result in a bit less sensitivity in detecting
some effect, we believe the absolute difference compares prices in a
more meaningful way.
Used prices contain the shipping cost but do not contain tax. The used
prices are a more nuanced comparison, since these are all 3rd party sellers.
Amazon is often more a marketplace than a retail site at this point, and
many people buy from 3rd party sellers on Amazon now without realizing it.
The relationship Amazon has with 3rd party sellers is also challenging.
Given the frequently changing dynamics in this space, we don't think any
analysis here will be very reliable for long term insights since products
from these sellers changes frequently in quantity and price. For this
reason, we focus only on new books sold directly by Amazon in our
comparison. In a future round of data collection, it may be interesting to
explore whether the dynamics have changed in the used market.