Harvesting Ratings

Harvesting Ratings with Johannes Johnen
Download Paper
Web Appendix
Evidence suggests lower prices lead to better ratings, but better ratings induce firms to charge higher prices in the future. We model that consumers are only willing to make the effort to rate a seller if this seller provides a sufficient value-for-money. Using this model, we explore how firms use prices to impact their own ratings. We show that firms harvest ratings: they offer lower prices in early periods to trigger consumers to leave a good rating in order to earn larger profits in the future. Because especially low-quality firms harvest ratings, harvesting makes ratings less-informative about quality. Based on this mechanism, (i) we argue that rating harvesting causes rating inflation; (ii) we show that a marketplace that facilitates ratings (e.g. through reminders, one-click ratings etc.) may get more ratings, but also less-informative ratings; (iii) a marketplace that screens the quality of sellers makes ratings less-informative if the screening is insufficient. Counter to the conventional wisdom that consumers benefit from ratings via the information they transmit, we show that consumers prefer somewhat, but never fully informative ratings. Nonetheless consumers prefer more-informative ratings than average sellers. We apply these results to characterise when a two-sided platform wants to facilitate ratings, and argue that efforts of major platforms to facilitate ratings did not just lead to less-informative ratings, but also shifted surplus from consumers to sellers.

This paper was previously titled “Ratings and Reciprocity”.

Some fun figures to explore

You may play around with the parameters.

Note: Price mimicking relates to low-quality firms obtaining no ratings. Hence, more price mimicking, reflected by δ*, translates to more informative ratings.

Observe how the informativeness of ratings (δ*) influence payoffs for low-quality firms.

Explore how platform design (e) affects the informativeness of ratings (δ*).

Find out how screening (γ) changes the informativeness of ratings (δ*).