Contrary to the conventional wisdom, the buyer's expected revenue and the surplus need not decrease with collusion, and the ex-ante surplus increases with the amount of information revealed in equilibrium. This is because when communication is cheap, bidders cannot directly collude on higher prices. Rather, communication leads to a competition between fewer, but more aggressive bidders, which entails more allocative efficiency and a decrease in the total wasteful entry cost (emphasis added).
The proposition indeed seems to challenge conventional wisdom about cartelization in public procurement markets and, consequently, the assumptions of the model deserve some close consideration. It is also relevant to stress that, throughout the paper, the author herself waters down this claim and its implications--which may have warranted a revision of the abstract (?).
In my view, the example used as the rationale for the paper (the well known Boeing-Airbus struggle over a $40bn USAF aircraft contract between 2008-11) might have primed the author in some of her considerations, which may downplay the relevance of the fact that reputation-based messages or threats are usually of scant significance in public procurement markets, unless very specific circumstances concur--which, mainly, boil down to the (pre-)existence of a very closed oligopoly, the presence of (very high) bidding costs, and a situation where all members of the colluding oligopoly communicate public messages before the tender [conditions that will be introduced slowly and progressively into the model, but which are not clearly spelled out from the beginning].
The paper eventually acknowledges these restrictions in its technical construction (where the preexistence of a cartel is clearly indicated as a condition for the 'cheap-talk' game; see pages 10 and 12), but it may not be equally clear in the narrative--where there is no clear indication as to the very rare occurrence of all of the conditions for the game equilibria to hold.
Moreover, the author itself stresses that, even within the model,
Keeping participation secret from bidders could then be a way to prevent collusion. This is consistent with the competition authorities' recommendation according to which sellers should not be allowed to communicate about their intention to participate. Though, this solution is practically hard to implement in the case of public procurements, for which bid preparation may last several years (p. 17, emphasis added).
This also bears stressing, as only a very limited number of procurement projects require the preparation of bids over a number of years.
Also, it needs to be underlined that the 'cheap talk' strategy is only relevant if there are significant (discrete) participation costs that the bidder can (completely) avoid if (as a result of the signalling derived from the 'cheap talk phase') it decides to stay out of the procurement. However, and very relevant here, significant participation costs are incurred upfront where bidders are invited/accepted to complex procurement procedures (such as competitive dialogues in the EU) and which are necessary to even reach the phase where bidders have a relatively clear idea of the 'private' valuation or production costs--and that, in my view, would require incorporating the role of 'sunk costs' into the decision-making processes analysed in the game.
Finally, it is also worth emphasising that, despite the design of the model as a repeated game (or, at least, its requiring some prior experience by the bidders in tendering against each other), dynamic (welfare) effects are not taken into account. In that regard, it would still be required to assess whether the potential allocative efficiency derived from a decrease in total wasteful participation costs is not outweighed by negative dynamic effects, such as market exit, reduction of innovation or, eventually, the evolution of the 'soft cartel' necessary for the model to work into a full-fledged hard cartel (where cheap talk is no longer cheap, but an all encompassing bid rigging strategy--something that the author will reckon as a sort of an after-thought in her concluding remarks).
In my view, the very extreme (and sometimes artificial and disconnected from the way particularly complex procurement processes work) assumptions of the model may reduce the validity (or, rectius, the practical relevance) of some of the findings of the paper--or, at least, of the normative propositions that could be derived therefrom. To her merit, it should be stressed that the author herself acknowledges that by indicating that
We do not think [these] results advocate authorizing communication between sellers in public procurements, though. Clearly, if communication is not cheap, sellers can either collude on higher prices, which indeed increases public spending, or on bid rotation schemes which are generally inefficient. Rather, we think these results emphasize how di fferent can be the outcome implications of cheap-talk and binding communication(emphasis added).
After reading the paper, and particularly because cheap talk is twinned to high participation costs (which eventually force bidders to put their money where their mouth is), it is hard to actually appreciate the differences between such 'cheap talk' and other forms of 'binding' communication within cartel rings. It follows that it is also difficult to extract any normative recommendation from the paper (and this should be stressed, as a superficial reading of the title, abstract and introduction may provide the contrary impression). Its scope is much narrower than it could seem at first glance and no changes on the existing rules and enforcement priorities agains bid rigging should be derived from the findings of the paper. Much ado about nothing? Nonetheless, in purely intellectual terms, it is interesting to read, particularly for game theory scholars.