DEGENERATE and PARTIAL, where buyers did not receive all information about a market,
hashes turned grey in the case where a buyer did not receive any information about a
particular seller. Note that in the treatment NO there was no such graphical representation at
all as sellers did not have labels. This easy and obvious color coding was chosen in order to
make the quite complex history information easy to read. (Screenshots are shown in the
Appendix.)
3 Results
In reporting the results we focus on three key variables: the average demand (or trust rate),
that is, the share of first movers that purchased the experienced good; the average traded
quality (or honor rate), that is, the share of second movers providing good quality (conditional
on first movers having bought the experience good); and the share of mutually beneficial
trades (or performance rate), that is the share of matches that resulted in (buy, good quality),
our efficiency measure.
Table 1 reports these averages (and standard deviations) for the four different treatments and
shows test statistics for a number of binary treatment comparisons (where the unit of
observation is always one entire matching group).
In NO, where there are no incentives for reputation building, the trust and honor rate are very
low and, as a consequence, so is the performance rate. Only in 5% of all matches do subjects
reach the mutually beneficial trade outcome. Introducing labeling of sellers in
DEGENERATE increases the trust rate by half and more than triples the honor rate. This is
not surprising: When customers can identify firms average quality improves drastically. As a
result the number of mutually beneficial trades is nearly quadrupled.
Increasing network density in PARTIAL, by allowing each buyer not only to know his own
experience but also the experience of a neighbor, improves market performance even further.
However, this change is solely due to more trust in the market, average quality is statistically
not distinguishable from average quality in DEGENERATE. Finally, when moving from
PARTIAL to FULL we observe a further increase in the performance rate. This time the
improvement is due to higher average quality while demand stays constant.