177
deterred from entering because it holds a 16% shareholding in a rival unless
the decision to enter was very marginal.
354,355
7.37 We agree that the model is helpful in illustrating the mechanisms at work as a
result of the Transaction and the difference in incentives resulting from a 16%
investment as opposed to a full merger. We note that some of the model’s
specific assumptions, and the calibration of the model, do not accurately
reflect the market (such as the online restaurant platforms being treated as
symmetric, the model not explaining current market outcomes, assuming that
all existing players in the market are active in all local markets and unclear
choice parameterisation).
356,357
As such, we place limited weight on the
results produced by the model, but agree with the logic of the model and
consider it is directionally helpful in illustrating that the effects of making a
minority investment will be different to a full merger.
7.38 We also consider the model provides a static assessment of the financial
implications of the investment in isolation. In particular, it does not account for
the possibility that the investment is strategic in the sense that it gives
Amazon a position from which it can continue to make further investment in
Deliveroo or use its investment to generate a strategic partnership. We
discuss our assessment of this in the following section. However,
notwithstanding the limitations of the model, such that we place limited weight
on the exact results, we do think it illustrates the limited cannibalisation effects
354
An increase in Amazon’s shareholding in Deliveroo would be expected to affect Amazon in two ways. First,
entry into an area by Amazon would divert sales from Deliveroo to Amazon (ie cannibalisation), thus reducing the
return it receives from Deliveroo. If Amazon had a larger shareholding in Deliveroo, this decreasing return would
likely result in Amazon entering fewer areas, Second, if Amazon did enter, the unilateral effects of the merger
would be greater if Amazon had a larger shareholding: a proportion of sales lost by Amazon would be recaptured
by Deliveroo and with a larger shareholding Amazon would likely have an incentive to compete less intensely and
prices rise.
355
In response to the Revised Provisional Findings, Domino’s submitted that entry can take place at different
scale and is not a binary ‘enter or not’ choice, and therefore the 16% shareholding may lead to Amazon entering
in fewer areas (see paragraphs 85 to 90 of Domino’s response to the Revised Provisional Findings, 10 July
2020). However, as described in this paragraph, the Parties’ model considered exactly this point (ie the extent of
the shareholding on the number of areas that Amazon enters and how it competes once it enters – including the
overall effect on price). Given we assessed this as a national market (see paragraph 5.127), and the precise
locations of any Amazon entry would be relatively speculative, we have not assessed Amazon’s entry decisions
in specific local areas. However, we note the general principle holds that if Amazon had a strong incentive to
enter a particular area then it is unlikely a 16% shareholding would materially reduce Amazon’s incentive to enter
that area.
356
The model assumes that firms are all similar along key dimensions, in technical terms symmetric. For
instance, the players in the model are identical with respect to their costs, quality and coverage of areas they are
active. These assumptions may bias the outcomes of the model and we are therefore concerned about the
reliability of the exact numerical results of the model.
357
The model is numerically simulated to illustrate different possible outcomes. We are concerned that, first, this
numerical simulation is based on a model that does not approximate actual market shares. However, responses
of firms may depend on their market share and therefore the model does not take this into account. Second,
evidence suggests that not all players are active in all markets (we observe markets with one, two or three
players). Third, while the Parties claim that they are using a wide and reasonable range in the numerical
simulation for the inputs into the model, we are concerned that those numerical values are not the most realistic
ones. For example, the Parties assume that the highest willingness to pay for an online delivery is a given
amount, however, it is not clear that this reflects the actual maximum willingness to pay by consumers. Overall,
we are therefore concerned about the reliability of the exact numerical results of the model.