State aid and tacit collusion

Christoph Bertsch Claudio Calcagno

Mark Le Quement

European University Institute, Economics Department

via della Piazzuola 43, 50133 Firenze, Italy
Ph. +39 055 4685 927; Fax: +39 055 4685 902

JEL classi…cation: D43, G21, K21, L41

Keywords: competition policy, State aid, collusion, banking.

30 September 2010

Abstract

Both literature and policy debate on State aid (or government subsidies) have focused on the trade-o¤ between

the potential ine¢ ciencies caused by state intervention (ine¢ cient allocation of resources, moral hazard) and

the potential gains from intervention (whether related to the resolution of market failures or to the achievement

of some dimension of social equity).

By contrast, the debate has ignored another important negative e¤ect of State aid: governments, by setting

up aid schemes to ailing …rms, may increase the likelihood of (tacit) collusion in an industry characterised by

idiosyncratic shocks.

Speci…cally, we show that, in a repeated-game setting, a systematic bailout regime increases the expected

pro…ts from cooperation and simultaneously raises the probability that competitors will still be in business to

carry out punishment against cheaters.

We are thankful to Franklin Allen, Elena Carletti, Piero Gottardi, Stan Maes, Nicola Pesaresi, Jan-Peter Siedlarek and Eliana
Viviano for their helpful comments. We also wish to thank participants at the EARIE 2010 conference and seminar participants
at EUI for helpful discussions. Correspondence: Christoph Bertsch, Department of Economics, EUI, Florence, Italy. E-mail:
[email protected]

1
 Despite the generality of the model and of its key insight, we study this problem through an application to

the banking sector, as it has recently been subject of much attention within the context of the economic and

…nancial crises.

2
1 Introduction and motivation

The …nancial crisis and its aftermath have been a¤ecting the global economy since 2007. This has involved the

bankruptcies of a large number of global …rms, including major …nancial institutions. The rapid deterioration of

the …nancial and economic situation has prompted massive interventions from Treasuries (and Central Banks) the

world over. State intervention, and in particular State aid, has been playing a major role, with governments rushing

to rescue not only a large fraction of their …nancial industries, but also other key industries like the automotive one.

Apart from the …scal burden imposed, State aid entails some thorny legal considerations, especially in the European

Union (EU), where Art. 107 of the EU Treaty explicitly forbids measures that confer (through public resources)

economic advantages to selected entities, a¤ecting trade. The European Commission (EC) was therefore under

pressure to soften its stance. In October 2008, owing to the gravity of the situation, it published a communication

on the application of State aid rules to …nancial institutions.1

To understand the magnitude of the aid involved, it is worth noting that in the EU the number of newly

introduced state aid measures is steadily increasing. The number of cases where aid was legally granted grew from

202 in 2003 to 636 in 2007 and to 964 in 2009.2 Moreover, during the crisis, EU Member States strongly supported

the …nancial sector, with e300 billion in capital injections and almost e3 trillion of guarantees.3

From a theoretical perspective, most of the debate on State aid (or government subsidies) has focused on

the trade-o¤ between ine¢ ciencies caused by intervention (ine¢ cient allocation of resources, moral hazard) and

potential gains following from aid measures. Arguments in favour of State aid range from equity considerations to

the potential resolution of existing market failures. In the …nancial sector, for example, intervention (including that

of a Central Bank) is often justi…ed on the grounds of gains in the stability of the …nancial system. Bankruptcies of

individual banks may trigger contagion e¤ects across the sector (through the interbank and asset markets), and may

also harm consumers directly through the loss of private deposits (subject to national deposit insurance schemes).
1 O¢ cial Journal C 270, 25.10.2008, pages 8–14. The accompanying press release explained that aid would be approved if: "Non-

discriminatory access in order to protect the functioning of the Single Market by making sure that eligibility for a support scheme is
not based on nationality; State commitments to be limited in time in such a way that it is ensured that support can be provided as
long as it is necessary to cope with the current turmoil in …nancial markets but will be reviewed and adjusted or terminated as soon
as improved market conditions so permit; State support to be clearly de…ned and limited in scope to what is necessary to address the
acute crisis in …nancial markets while excluding unjusti…ed bene…ts for shareholders of …nancial institutions at the taxpayer’s expense;
[a]n appropriate contribution of the private sector by way of an adequate remuneration for the introduction of general support schemes
(such as a guarantee scheme) and the coverage by the private sector of at least a signi…cant part of the cost of assistance granted;
[s]u¢ cient behavioural rules for bene…ciaries that prevent an abuse of state support, like for example expansion and aggressive market
strategies on the back of a state guarantee; [a]n appropriate follow-up by structural adjustment measures for the …nancial sector as a
whole and/or by restructuring individual …nancial institutions that had to rely on state intervention."
2 Source: EC, DG Competition, http://ec.europa.eu/competition/state_aid/studies_reports/measures.html.
3 The numbers include only measures approved by the EC by September 2009 and hence do not include the granted

and planned capital injections after the European stress tests in 2010 and Basel III. Source: EC, DG Competition,
http://ec.europa.eu/competition/recovery/…nancial_sector.html.

3
On the other hand, State aid can distort the competitive process (prompting a misallocation of resources), as well

as create moral hazard: if …rms expect that the Government will intervene to help them in case of failure (or in

adverse circumstances more generally), these may have the incentive to take excessive risks.

The economic e¤ects studied have always impinged on …rms’unilateral behaviour. In this article, in contrast,

we develop a simple (in…nite-horizon) model that sheds light on a result that to our knowledge has not been stressed

before: a government policy aimed at systematically rescuing …rms in the presence of negative idiosyncratic shocks

facilitates (tacit) collusion. Collusion is easier to sustain because the punishment threat faced by a …rm is more

e¢ cient, which is in turn due to the increased survival probability of …rms. Indeed, when …rms are guaranteed

to be in business in the next periods, expected future cooperative pro…ts increase, so that foregoing such pro…ts

implies an increased opportunity cost. Furthermore, the guaranteed presence of competitors in the future makes the

expected punishment harsher than in an environment where competitors may exit the market due to an exogenous

shock, which would leave the deviant …rm unpunished. Our results impinge on the provision of systematic bailouts;

the EC’s o¢ cial policy of "one-time-last-time" aid would thus undermine our mechanism.4 However, the EC has

departed from this principle a number of times (especially so in the latest …nancial crisis) where ex post this would

have been detrimental for the economy; we discuss this in Section 3 in greater detail.

Although the model and its implications are general and can match many industries, we study this problem

through an application to the banking sector, as it has recently received much attention due to large bailouts.

Some words of caution are in order. The key result of the paper - namely that State aid may facilitate tacit

coordination - cannot be taken at face value in the real world. This is ultimately an empirical question, which

regulators, policy-makers and courts would need to assess on a case-by-case basis. There are several industry

features that may facilitate coordination, such as stable demand, homogeneous products, little innovation, market

transparency.5 It is clearly arguable whether the banking industry actually meets these criteria. Our point, however,

is that for given industry characteristics, an explicit (or implicit) commitment to bail …rms out upon the realisation

of shocks facilitates coordination.

This article proceeds as follows: in Section 2 we brie‡y consider the relevant literature; in Section 3, we present

the model. Section 4 shows and discusses the results, including a welfare analysis. Section 5 concludes and suggests

some extensions.

4 See European Commission (2004).

5 Clear and sound economic reasoning (presented in a suitable manner for policy-making) on the economics of tacit collusion can be
found in Ivaldi et al (2003).

4
2 Related literature

There is not a vast literature on the economics of State aid as such, possibly re‡ecting the richness of the literature

on subsidies and trade.6 State aid is typically criticised by economists, as it leads to a variety of ine¢ ciencies. Besley

et al (1999) discuss two broad classes of e¤ects: externalities arising from aid (strategic trade policy, tax competi-

tion and economic geography considerations) and ine¢ cient competition between governments. Dewatripont and

Seabright (2006) go beyond intergovernmental issues and build a model where local politicians invest in wasteful

projects purely to show their diligence and win votes. Collie (2000) instead proposes an economic explanation of

why individual states may have an incentive to subsidise …rms with the aim of reducing oligopolistic distortions.

He shows that a multilateral institution responsible for prohibiting subsidies can increase welfare.

Friederiszick et al (2008) review the e¢ ciency rationales for aid (tackling market failures such as externalities,

public goods, asymmetric information and lack of coordination) as well as equity considerations. They also point

towards cross-border (positive) externalities in the case of EU State aid. Their paper then highlights the potential

costs of State aid (beyond the direct cost of intervention) such as anti-competitive e¤ects, “picking wrong winners”

and international spillover e¤ects. Among the potential distortions of competition, they list the support of ine¢ cient

production; the distortion of dynamic (inter-temporal) incentives; the potential increase in market power; and the

distortion of production and location decisions across EU countries. Finally, they propose an actual e¤ects-based

framework to assess whether particular State aid measures should be approved. Martin and Valbonesi (2008)

develop a model of the impact of State aid on market structure and performance in an integrating market (i.e. a

common market with increasing trade ‡ows) and …nd that in equilibrium governments grant State aid, reducing

common market welfare. However, they only focus on non-cooperative equilibria.

Finally, Hainz and Hakenes (2009) compare the e¢ ciency properties of …ve options to grant State aid to …rms

(some of which would also feed through to the banking system). The most e¢ cient option is shown to depend on

the tax distortion and the informational cost needed to select the "good" …rms.

In the case of banking, it may be bene…cial to sacri…ce some level of competition in the interest of …nancial

stability.7 However, this relationship is rather complex and both empirical and theoretical results are far from being

clear-cut.8 This debate is nevertheless beyond the scope of our paper, since we only take banking as an example and
6 An extensive review of the role and the e¤ects of State aid can be found in Nitsche and Heidhues (2006). For an equally policy-

oriented approach based on economic theory, the reader is also directed to OFT (2004) and Buelens et al (2007).
7 See Carletti (2008) for an excellent survey on this trade-o¤.
8 Keeley (1990) was the …rst one to …nd a positive empirical relationship between more competition and more risk-taking but later

studies came to mixed or even opposite results. The earlier theoretical literature mostly gave rationales for Keeley’s …ndings but also
discussed why the opposite may be the case (most prominently Boyd and De Nicoló (2005)). For a theoretical discussion see Allen and
Gale (2004). A more recent extensive review is given by Vives (2010).

5
our model is not sector-speci…c. Rather, we simply point towards some work on imperfect competition in banking

in dynamic setups. Hellmann et al (2000) show that in a dynamic model deposit rate controls can be superior

to costly regulation.9 They explore the interaction between …nancial liberalisation and prudential regulation and

…nd that capital requirements (though costly) can prevent banks from excessive risk-taking in a static model. By

contrast, this positive e¤ect comes at a very high cost in a dynamic model because of the business-stealing e¤ect

that induces banks to opt for riskier strategies than their competitors and o¤er better interest rates to depositors in

order to collect more deposits. Perotti and Suarez (2002) investigate the impact of competition on banks’portfolio

risk choices. In particular, they examine the relationship between the optimal portfolio risk and banking regulation

(merger policy and market entry regulation) in an oligopoly context. The main mechanism in their model is the

strategic substitutability between portfolio decisions of duopolistic banks. In particular, a given duopolist has

an incentive to invest in the prudent asset if the competitor chooses a risky strategy (since she can expect large

monopoly rents if the competitor fails).

In industrial economics, models of tacit collusion have been heavily applied, but not in the context of State aid.

We take a standard framework of tacit collusion for repeated oligopoly interaction (originally due to Stigler (1964))

and analyse the e¤ect of State aid in an application to the banking industry.

3 Model

We develop a model on the relation between State aid and tacit collusion through an application to the banking

industry. The mechanism we describe is not speci…c to the …nancial industry and can be readily transposed to

other industries.

The starting block of our model is Freixas and Rochet’s (2008) extension of the Klein-Monti model to Cournot

oligopoly. The original monopolistic model features a single bank facing an upward-sloping demand for deposit and

a downward-sloping demand for loans (as developed by Klein (1971) and Monti (1972)).

To shed light on the impact of State aid on tacit collusion and consumer welfare, our model will make sim-

pli…cations with respect to Freixas and Rochet (2008). In our economy, the banking industry is characterised

by a duopoly competing in the deposit market over an in…nite horizon. We consider time in a discrete fashion.

Production (management) costs are normalised to nil. Banks simultaneously set interest rates r1 and r2 ; where
9 Hellmann et al (2000) do not model imperfect competition explicitly but make assumptions on derivatives of the deposit demand

function. See Repullo (2004) for an explicit modelling of imperfect competition that delivers some additional insights.

6
r1 ; r2 > 0. The (linear) demand function is given by:

Q(r1 ; r2 ) = min Q; max fr1 ; r2 g : (1)

The environment is stochastic in the sense that there is a single asset (project) in which a bank invests its funds.

This asset (project) is subject to the following idiosyncratic shock: it yields net return RH (with probability p) and

RL (with probability 1 p).10 We assume that RH < Q (this simply normalises static competitive pro…ts to nil).

For simplicity, and without loss of generality, we normalise RL to 1. This means that all funds are lost in the

presence of a bad shock and nothing is returned to depositors.

Banks have discount factor 0 < < 1 and the timeline of our game, for each period t 2 [1; 1), is given in Figure

1.11

Figure 1: Timeline
At each time t:

Banks stay (pay

Consumers choose gross interest) or exit
their bank (c. k.) Time

r1, r2 are set Shocks realise RH (p) Individuals receive capital plus
(common and RL (1-p) interest and consume it (p) or
knowledge) (c. k.) lose everything (1-p)

Every period, bank i = 1; 2 maximises pro…ts ((Ri ri ) Q(ri ; rj )) by choosing ri , taking into account the

possibility to tacitly collude for high enough discount factors (as in Stigler (1964)).12 We assume that the market

(deposits and pro…ts) is equally shared by the banks when they set the same interest rate. Importantly, when a bank

receives a bad shock (RL ), it is forced out of the market (as in Perotti and Suarez (2002)). It goes bankrupt because

it cannot repay depositors and its authorisation to operate is not renewed. We make the following assumption:
1 0 In another industry one could argue that …rms can be hit with a certain probability in each period by a cost shock or a shock on

their production technology that forces them to leave the market.

1 1 Discount factors over in…nite-horizon games are often interpreted as the probability that the game will actually be played in a given

period, so as to implicitly relax the in…nite-horizon interpretation. Note that this is di¤erent from the adverse shocks that we introduce
(with probability p) as these are idiosyncratic.
1 2 In general the game has multiple equilibria but we restrict ourselves to the best equilibrium in terms of pro…t maximization, i.e.

banks collude for su¢ ciently high discount factors and both charge the monopoly interest rate.

7
Condition 1 RH > 4 1 p p .

This condition is necessary to ensure positive levels of consumer welfare.13 Banks will set interest rates only

taking the good state of the world (RH ) into account. Let us adopt the superscript M for monopoly (or collusion)

and C for competition, and denote pro…ts with . If a bank is alone in the market, pro…t maximisation leads to
RH RH RH 2
rM = 2 , Q(rM ) = 2 and M
= 2 . Competition, by contrast, leads to rC = RH , and Q(rM ) = RH and
C 1 RH 2
= 0: Let us de…ne W M = 2 2 as the consumer surplus (or welfare) at the monopoly (or collusive) price
2
RH
level and W C = 2 as the consumer surplus under perfect competition.

The next element in our economy is a national deposit insurance scheme (NDIS), covering 100% of deposits.

Hence, when a bank goes bankrupt, depositors are returned their initial investments (without interest). The NDIS

is funded by ‡at-rate taxes. We use the subscript "C" for collusion and "NC" for "no collusion". The size of the
RH RH
insurance scheme is: C1 = 4 whenever one collusive duopolist bank fails; C2 = 2 whenever a monopolist
RH
bank fails or both banks in a collusive duopoly fail; N C1 = 2 whenever banks compete and one receives a bad

shock; N C2 = RH whenever banks compete and both receive a bad shock.

Notice that the introduction of the deposit insurance scheme is not only crucial within the context of a banking

model, but it also generates two important features. First, it enables us to isolate the e¤ect of State aid on collective

competitive behavior when we compare consumer welfare under a State aid and a No State aid regime. This follows

from the fact that depositors get their deposits fully refunded under both regimes whenever a bank fails, while the

…nancing of the scheme is done by a non-distortionary tax in both regimes. Second, the introduction of deposit

insurance makes our model readily comparable to applications to other industries where the failure of a …rm does

not cause an immediate loss in wealth to its customers.

Figure 2 summarises, graphically, the above discussion. Notice that as the demand schedule is upward sloping,

the various areas in the graph (pro…t , consumer surplus and deadweight loss, or DWL) are inverted (horizontally)

with respect to a traditional diagrammatic analysis of linear demands. For simplicity, we have only depicted the

potential loss from bankruptcy ( C2 ) in the case of monopoly. Notice the analogy to a model of duopolistic

competition in a non-banking industry with a downward sloping demand curve.

Finally, we de…ne State aid (also …nanced through lump-sum taxes) as the following, cumulative actions. First,

the State renews the bank’s authorisation to operate even when the bank has gone bankrupt. Second, the State
1 3 Under (1 p)
reasonable parameter assumptions we have RH > 4 p
, e.g. for p = 0:9 and RH = 0:6 we have 0:6 > 0:4 (recall RL = 1,
so RH = 0:6 is not particularly high).

8
 Figure 2: A simple model of demand for deposits

incurs a sunk cost per bank rescue. We do not explicitly model how this cost arises but it could be justi…ed,

for instance, as the cost of resources devoted by the …nancial regulator to examine the books of a failing bank and

facilitate its rescue. In a non-…nancial industry this cost could arise from expensive restructuring. Notice that we

do not impose any assumption on . Instead, in the welfare analysis section, we determine upper bounds on its

value such that the aid policy is welfare-enhancing to consumers.

As mentioned earlier, the EC is committed to a "one-time-last-time" approach to aid. But in practice there have

been several exceptions to this rule, most notably during the …nancial crisis. Moreover, in the …nancial industry

policy makers fear potentially strong negative externalities in the short-run (contagion e¤ects, adverse impact on

real economy) if they do not grant aid to an insolvent bank. For this reason policy makers consider bailouts as

being necessary ex post.14 Consequently banks that are very interconnected or "too-big-to-fail" can expect to be

rescued because policy makers can hardly make a credible commitment not to bail banks out. We therefore think

that our approximation of systematic State aid is adequate for this type of implicit or explicit guarantees in the
1 4 See also Lyons (2010) for a discussion of this issue.

9
…nancial sector.15

Finally, as we noted above, the NDIS is independent of whether there is a State aid policy (however, in total

terms, the NDIS will be costlier in the presence of a State aid regime since banks are always rescued and can fail

several times, whereas in absence of State aid there can be two bankruptcies at most in the economy). The NDIS

is a banking-speci…c concept, but as the NDIS is exactly matched in size by a lump-sum tax, our framework is

general and goes beyond the banking sector application.

4 Results

In this section we …rst demonstrate that the introduction of State aid decreases the minimal discount factor for

which tacit collusion is possible. This implies that there is a range of values of the discount factor for which the

implementation of State aid causes a change from a competitive outcome to a collusive one. Next, we examine the

implications of this result on the welfare e¤ects of a State aid policy. We demonstrate that for this range of values

of the discount factor, State aid always a¤ects consumer welfare negatively, while it has a positive e¤ect for all

other values of the discount factor if the intrinsic cost of rescuing ( ) is not too high.

4.1 State aid and collusion

Let us posit a collusive regime in which …rms adopt a simple trigger strategy by charging the collusive price as long

as no one in the cartel defects and reverting to the competitive price (forever) as soon as one …rm has defected

from the tacit collusive agreement.

The …rst step is to derive the critical discount factors above which tacit collusion is sustainable under each

policy regime. In the case of no State aid, the collusive pro…ts need to embed the probability that a …rm becomes

a monopolist (i.e. whenever the competitor receives a negative shock) as well as the probability that the …rm itself

goes bankrupt. The pro…ts in the period of deviation also need to account for the probability that the deviating

…rm goes bankrupt, since the shock occurs after market conduct is chosen. Likewise, punishment pro…ts (from the

following period onwards) need to encompass the possibility that the deviating …rm will actually be a monopolist

for some time, as well as the possibility that the deviating …rm goes out of business.
1 5 The existence of implicit guarantees is of course hard to prove empirically. Nevertheless there is some indirect evidence. Rating

agencies publish "external support" ratings which re‡ect their expectations on the likelihood of a bailout. See Gropp et al (2010) for a
paper that uses this data to estimate how bank risk-taking behaviour is a¤ected by the presence of a guarantee.

10
 Without State aid the incentive compatibility constraint (ICC) for tacit collusion - which we fully derive in

Appendix A.1 - looks as follows:

Present value of exp ected pro…ts under collusion

z }| {
X1 M
t 1
p2t 1
+ pt (1 pt 1
) M
(2)
t=1
2
1
X
t 1
p M + p2t 1 C
+ pt (1 pt 1
) M
:
|{z}
t=2
Exp ected instantaneous pro…t from deviating | {z }
Present value of exp ected pro…t under punishm ent

C
From now onwards, we use the superscripts "NA" for "no aid" and "A" for "aid". Noting that = 0, tacit
NA 1
collusion can be sustained for all = 2p2 .

With State aid, by contrast, a bank that has received a bad shock is rescued at no cost to it. There is no pro…t

(nor loss) in the period of failure. Noting once again that the competitive pro…t is zero every period, we set up the

following ICC:

1
X M 1
X
t 1 M t 1 t 1 C t 1
p + (1 p) 0 p + (p + (1 p) 0): (3)
t=1
2 t=2

A
So tacit collusion can be sustained for all = 21 . This is the traditional result obtained in a supergame where

symmetric duopolists compete on price. The only di¤erence is that both the collusive pro…t and the deviation

pro…t have to be scaled by the probability of receiving a good shock; however the actual probability p cancels out.

Figure 3 provides a graphical representation of this result. When banks place little value on the future (left

half of the chart) tacit collusion cannot be sustained, regardless of whether there is State aid. When banks care
1
much about the future for a given likelihood of a good shock ( 2p2 , i.e. the area at the top-right corner) tacit

collusion can be sustained in either regime. Finally, for intermediate values of the discount factor ( 21 < 1
2p2 ,

i.e. the area at the bottom-right) only the competitive outcome is sustainable in the absence of an aid policy, while

tacit collusion is made possible by a State aid policy.

11
 Figure 3: Probability of good shock and critical discount factor
p 1.0

Collusion
0.9
in both regimes
0.8

0.7

0.6
Competition
Competition in
0.5 without aid
both regimes
0.4
BUT
0.3
Collusion with aid
0.2

0.1

0.0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Delta

The results are summarised in Proposition 1:

Proposition 1 State aid and collusion

In the presence of idiosyncratic shocks, a State aid policy that keeps banks (…rms) in business after a shock

facilitates tacit collusion; that is, tacit collusion can be sustained for lower values of discount factors.

1 A NA 1
Proof. The proof of is trivial, as 2 = < = 2p2 ; 8 p < 1.

4.2 Welfare impact of State aid

In the last section we examined the relation between competitive behaviour and the exogenous discount factor for

the two regimes, State aid and no State aid. We derived two critical discount factors that de…ned three regions

in the [p ] parameter space. We next perform a comparison of consumer welfare under the two regimes and

summarise our results in Proposition 2 at the end of the section.

12
4.2.1 No State aid

1
Under collusion ( 2p2 ): The exercise here is to capture all possible scenarios, namely collusive duopoly and

monopoly, considering all the possible combinations of events (good and bad shocks). Each scenario will occur with

a certain probability and will entail a certain level of consumer welfare (including the tax levied). We provide a

full derivation in Appendix A.2. Expected consumer welfare under a high enough discount factor to sustain tacit

collusion and in the absence of State aid is computed to be:

RH (1 p) p4 RH RH p
2 RH 2(1 p)
E(WCN A ) = + : (4)
2 (1 p2 ) 2 (1 p)

Note that Condition 1 ensures positive consumer welfare levels.16

1
Under no collusion ( < 2p2 ): Here, we carry out the analogous exercise; however, due to the lower discount

factor, there can only be either a monopoly (if the competitor has received a bad shock) or a competitive duopoly.

In either case there can be good or bad shocks (in the latter case it could be that only one bank received the bad

shock in the last period). The expected consumer welfare in the absence of aid, under a low enough discount factor

to guarantee competition and in the presence of both banks on the market, is given as follows:17

2 p
p (RH ) RH 2 RH 2(1 p)
E(WNNCA ) = 2
+ : (5)
4 (1 p ) 2 (1 p)

4.2.2 State aid

We proceed in the same fashion as in subsection 4.2.1. However, here, banks never exit. When consumers’

deposits are lost due to the bad shocks, the Government refunds them the original capital, as well as paying the

direct rescuing costs . To close the model, we need to compute the total expected stream of aid and set up a

corresponding lump-sum tax (which includes the …nancing of both the NDIS and the direct rescuing costs ) on

consumers, thus reducing their welfare.

1
Under collusion ( 2 ): In the presence of State aid, the expected consumer welfare under a high enough

discount factor to sustain tacit collusion is:

p
RH 4 RH (1 p) 2(1 p)
E(WCA ) = : (6)
2 (1 ) 1
1 6 To (1 p) R =2 RH =2
see this notice that the terms in brackets are negative since RH > 4 p
. Further 1 H p2 < 1 p
whenever p < 1. Eventually
RH =2 p R =2
for expected welfare to be positive we need to have 1 p2
p)
[(1 R ] > 1 H p [ p2 RH 2(1
4 H
p)] which is ensured since 1
1 p
p2
< 2.
1 7 See in Appendix A.2 for the derivation. Again, Condition 1 ensures positive consumer welfare levels.

13
Under no collusion ( < 12 ): In the presence of State aid, the expected consumer welfare under a discount factor

low enough to guarantee competition (when both banks are in the market) is:

RH (pRH (1 p)) 2(1 p)

E(WNAC ) = : (7)
2 (1 ) 1

4.3 Summary and discussion of the results

Having derived these four benchmark levels of consumer welfare corresponding to the two possible policy regimes

and the two possible competitive regimes, it is instructive to plot them jointly in a diagram. We do so in Figure 4.

The horizontal axis corresponds to the discount factor . Three relevant regions can be identi…ed. In the left region

( < 21 ) banks compete with each other regardless of whether there is State aid. In the right region ( 1
2p2 ) tacit

collusion can be sustained in either regime. However, in the central region, tacit collusion can only be sustained

in the presence of State aid. In the left and in the right regions, State aid can be consumer welfare-enhancing and

this depends on whether the vertical distance between the broken and the solid line (for any given discount factor)

exceeds the (expected tax bill due to the) direct costs of rescuing banks . In the central region, by contrast, State

aid decreases consumer welfare.

Figure 4 abstracts from the cost of rescuing a bank. In what follows, we compute the upper limits on such

that the expected consumer welfare under State aid is larger than under no State aid, in the cases where public

intervention does not a¤ect the competitive state of the economy. Call bN C this limit value of for the case of no

collusion in both policy regimes, i.e. for < 12 :

E(WNAC ) E(WNNCA ) ()

1 RH (pRH 2(1 p))

bN C = E(WNNCA ) (8)
2(1 p) 2 (1 )

In Appendix A.3 we show that bN C 0. We now compute the upper limit bC corresponding to the case of collusion
1
in both regimes, i.e. for 2p2 :

E(WCA ) E(WCN A ) ()
!
p
1 RH 4 RH (1 p)
bC = E(WCN A ) (9)
2(1 p) 2 (1 )

Again we show in Appendix A.3 that bC 0.

Proposition 2 summarises the results on the welfare e¤ects of State aid.

14
 Figure 4: E¤ect of State aid on consumer welfare, by discount factor

Proposition 2 State aid and consumer welfare

(i) In the range of discount factors such that there is competition in absence of State aid but tacit collusion with
1 1
State aid, such a policy reduces consumer welfare. That is, E(WCA ) < E(WNNCA ); 8 2 < 2p2 . This is true even

in the absence of direct rescuing costs.

(ii) In an environment where duopolistic banks (…rms) compete regardless of whether there is State aid ( < 12 ),

such a policy reduces consumer welfare if and only if the direct costs of rescuing a bank (…rm) exceed bC .

(iii) In an environment where duopolistic banks (…rms) can sustain tacit collusion regardless of whether there is
1
State aid ( 2p2 ), such a policy reduces consumer welfare if and only if the direct costs of rescuing a bank (…rm)

exceed bN C .

Proof. The proof of (i) is in Appendix A.4. For the proof of (ii) and (iii) see above inequalities.

Proposition 2 introduces a simple dichotomy between two scenarios that summarises the e¤ect of State Aid on

consumer welfare. In one scenario (cases ii) and iii) above), State aid does not in‡uence the competitive state of

the economy, and is welfare-improving as long as its intrinsic cost is not too large. Indeed, in such a case, the

15
only e¤ect of State aid is to bene…cially preserve the existence of the market, which is advantageous to both …rms

and consumers. In the other scenario (case i)), where State aid a¤ects the competitive state of the economy by

triggering collusion, we …nd that its overall e¤ect is always negative. This means that in such a scenario, the

adverse collusion-creating e¤ect of State aid dominates its bene…cial market-preserving e¤ect. The dominance of

the …rst force over the second force does not appear like an a priori necessity. We attribute this feature to the fact

that this scenario corresponds to relatively low values of the discount factor. For such values of the discount factor,

the adverse price e¤ect of collusion thus dominates the long run positive bene…ts related to the preservation of the

market.

5 Conclusion

The literature on State aid and the related policy debate have typically focused on the adverse e¢ ciency e¤ects

of such policies (misallocation of resources, moral hazard) and on countervailing arguments typically (though not

exclusively) based on social policy.

Here we developed a very simple in…nite-horizon model that sheds light on a separate result that to our knowledge

had not been stressed before: a government policy aimed at systematically rescuing …rms in the presence of negative

idiosyncratic shocks facilitates (tacit) collusion. This is because expected future co-operative pro…ts increase (since

…rms are ensured to be in business in the subsequent periods); and because the guaranteed presence of competitors

in the next periods makes the (expected) punishment phase harsher than under an environment where competitors

may exit the market due to an exogenous shock, which would leave the deviant …rm unpunished.

Examining the implications of this result for the welfare e¤ects of State aid policy, our main result is the

identi…cation of a range of discount factors for which an aid policy is always detrimental for consumer welfare because

of its collusion-facilitating e¤ect. Yet, in the real world, this link would actually have to be shown empirically and

regulators, policy-makers and courts would need to assess this on a case-by-case basis.

The analysis in this paper has focused on the basics of the mechanism to highlight the key insight. However,

this paper can set the stage for interesting extensions. One possible direction is to devise more complex aid

policies (stochastic, with repayments, limited to the last failing …rm) or to consider a more complex competitive

setup (introducing asymmetries, entry, a richer menu of contracts). Another possibility would be to embed a true

banking model within our framework, for instance by endogenising portfolio choice and modelling an interbank

(wholesale) market.

16
References

[1] Allen, F., & Gale, D. (2004). Competition and …nancial stability. Journal of Money, Credit and Banking, 36(3),

453-480.

[2] Besley, T., Seabright, P., Rockett, K., & Sorensen, P. B. (1999). The e¤ects and policy implications of State

aids to industry: an economic analysis. Economic Policy, 14(28), 15-53.

[3] Boyd, J. H., & De Nicoló, G. (2005). The theory of bank risk taking and competition revisited. Journal of

Finance, 60(3), 1329-1343.

[4] Buelens, C., Garnier, G., Meiklejohn, R., & Johnson, M. (2007). The economic analysis of state aid: some

open questions. European Economy Economic Papers, 286.

[5] Carletti, E. (2008). Competition and regulation in banking. In A. V. A. Boot, & A. V. Thakor (Eds.), Handbook

of Financial Intermediation and Banking (pp. 449-482). Amsterdam: Elsevier North Holland.

[6] Collie, D. R. (2000). State aid in the European Union: the prohibition of subsidies in an integrated market.

International Journal of Industrial Organization, 18(6), 867-884.

[7] Dewatripont, M., & Seabright, P. (2006). "Wasteful" public spending and State aid control. Journal of the

European Economic Association, 4(2-3), 513-522.

[8] European Commission (2004). Communication from the Commission - Community guidelines on State aid for

rescuing and restructuring …rms in di¢ culty, O¢ cial Journal C 244, 1.10.2004, 2-17.

[9] European Commission (2008). Communication on the application of State aid rules to measures taken in

relation to …nancial institutions in the context of the current global …nancial crisis. O¢ cial Journal C 270,

25.10.2008, 8-14.

[10] Freixas, X., & Rochet, J.-C. (2008). Microeconomics of banking. Cambridge, MA.: MIT Press.

[11] Friedriszick, H. W., Roeller, L.-H., & Verouden, V. (2008). European State aid control: an economic framework.

In P. Buccirossi (Ed.), Handbook of Antitrust Economics (pp. 625-670). Cambridge, MA: MIT Press.

[12] Gropp, R., Hakenes, H., & Schnabel, I. (2010). Competition, risk-shifting, and public bail-out policies. Working

paper series of the Max Planck Institute for collective goods, 2010/05.

[13] Hainz, C., & Hakenes, H. (2009). The politician and his banker. Ifo Working Paper, 71.

17
[14] Hellmann, T. F., Murdock, K. C., & Stiglitz, J. E. (2000). Liberalization, moral hazard in banking, and

prudential regulation: are capital requirements enough? American Economic Review, 90(1), 147-165.

[15] Ivaldi, M., Jullien, B., Rey, P., Seabright, P., & Tirole, J. (2003). The Economics of tacit collusion. IDEI

Working Paper, 186, Final Report for DG Competition, European Commission.

[16] Keeley, M. C. (1990). Deposit insurance, risk and market power in banking. American Economic Review, 80(5),

1183-1200.

[17] Klein, M. A. (1971). A theory of the banking …rm. Journal of Money, Credit and Banking, 3(2), 205-218.

[18] Lyons, B. (2009). Competition policy, bailouts and the economic crisis. CCP Working Paper, 09-4.

[19] Martin, S., & Valbonesi, P. (2008). Equilibrium State aid in integrating markets. B.E. Journal of Economic

Analysis & Policy, 8(1) Topics, 1-37.

[20] Monti, M. (1972). Deposit, credit, and interest rate determination under alternative bank objectives. In G. P.

Szegö, & K. Shell (Eds.), Mathematical methods in investment and …nance (pp. 430-454). Amsterdam: North

Holland.

[21] Nitsche, R., & Heidhues, P. (2006). Study on methods to analyse the impact of State aid on competition.

European Economy Economic Papers, 244.

[22] OFT (2004). Public subsidies - A report by the O¢ ce of Fair Trading. OFT 750.

[23] Perotti, E. C., & Suarez, J. (2002). Last bank standing: what do I gain if you fail? European Economic Review,

46, 1599-1622.

[24] Repullo, R. (2004). Capital requirements, market power, and risk-taking in banking. Journal of Financial

Intermediation, 13, 156-182.

[25] Stigler, G. J. (1964). A theory of oligopoly. Journal of Political Economy, 72(1), 44-61.

[26] Vives, X. (2010). Competition and stability in banking. CEPR Policy Insight, 50.

18
A Appendix

A.1 Derivation of the critical discount factors

A.1.1 No State aid

We start by deriving the expected collusive pro…t (LHS of the ICC). At each period t, a bank gets collusive pro…ts
t M t M
2 with probability p1 (t); monopoly pro…ts with probability p2 (t); and 0 with probability p3 (t). The

respective probabilities can be written as follows:18

p1 (t) = pt (pt 1
) = p2t 1

1 pt 1
p2 (t) = pt (1 p) + p(1 p) + p2 (1 p) + ::: + pt 2
(1 p) = pt (1 p) = pt (1 pt 1
)
1 p
p3 (t) = 1 p2t 1
pt (1 pt 1
)=1 pt :

This yields:

1
X M
t 1
LHS = p2t 1
+ pt (1 pt 1
) M

t=1
2
p p p M
= +
2(1 p2 ) 1 p (1 p2 )
p p M
= :
1 p 2(1 p2 )

Next, we turn to the right-hand side of ICC, i.e. the immediate deviation pro…t (obtained with probability p since

the shock occurs after the interest rate decision) plus the expected punishment stream from the following period
M
onwards. The former pro…t is simply p . As for the latter, there are four possible events at each time t 2:

1. Both banks are in the market at the beginning of the period and the deviating bank has a good shock in that

period: p4 (t) = p2(t 1)

p:

2. Both banks are in the market at the beginning of the period and the deviating bank has a bad shock in that

period: p5 (t) = p2(t 1)

(1 p):

3. The deviating bank will be in the market at time t 2 and earn monopoly pro…t alone:
t 1
p6 (t) = pt (1 p) + p(1 p) + p2 (1 p) + ::: + pt 2
(1 p) = pt (1 p) 1 1p p = pt (1 pt 1
):
1 8 Notably the probability of the competitor being in the market in period t can be computed as pt 1. The fact that the competitor
might have to leave the market in period t does not a¤ect the pro…ts of the other bank in period t.

19
 4. The deviating bank will not be in the market: p7 (t) = 1 p2(t 1)
pt (1 pt 1
).

C
However, it is only p6 (t) that is associated to a non-zero payo¤ (p4 (t) and p5 (t) are associated to = 0 and

p7 (t) to previous exit). Thus, adding over t:

1
X
M t 1 t
RHS = p + p (1 pt 1
) M

t=2

M p2 M p3 M
= p +
1 p 1 p2

M
Constructing the overall ICC by comparing LHS against RHS (i.e. (2)), solving for and noticing that falls

through, one gets that tacit collusion is sustainable if:

1 1 M p p2 M
p 1+ p
1 p 2(1 p2 ) 1 p 1 p2
| {z } | {z }
exp ected pro…t of collusion exp ected pro…t of deviation
1
i:e: :
2p2

A.1.2 State aid

With State aid a bank that has received a bad shock is rescued and allowed to operate in the following period.

There is no pro…t (nor actual loss) in the period of failure. Setting up the ICC and solving for the critical discount

factor, we obtain the traditional supergame result in a symmetric price-setting duopoly:

X M X X X
t 1 t 1 M t 1 C t 1
p + (1 p) 0 p + (1 p) 0 + p |{z} + (1 p) 0
t=1
2 t=1 t=2 t=2
| {z } =0
| {z }
exp ected pro…t of collusion exp ected pro…t of deviation
M
p M
i:e: p :
2(1 )

1
That is, collusion is sustainable for 2.

A.2 Derivation of the consumer welfare equations

A.2.1 No State aid

1
Under collusion ( 2p2 ):

From a consumer welfare perspective, there are six possible states at time t: (I) there is a collusive duopoly

20
and the deposits are returned with interest by both banks; (II) there is a collusive duopoly and all deposits are

lost because of the bad shocks to both duopolists; (III) there is a collusive duopoly and only one bank receives a

bad shock; (IV) there is a monopoly and the deposits are returned with interest; (V) there is a monopoly and the

deposits are lost because of a bad shock; (VI) there is no market at all (banks have exited and consumer welfare is

nil).

pI (t) = p2t

pII (t) = p2(t 1)

(1 p)2

pIII (t) = 2 p2(t 1)

p(1 p) = 2 p2t 1
(1 p)

pIV (t) = 2 pt 1
(1 p) + p(1 p) + p2 (1 p) + ::: + pt 2
(1 p) p

= 2pt (1 pt 1
)

pV (t) = 2 pt 1
(1 p) + p(1 p) + p2 (1 p) + ::: + pt 2
(1 p) (1 p)

= 2pt 1
(1 p)(1 pt 1
)

pV I (t) = 1 (pI (t) + pII (t) + pIII (t) + pIV (t) + pV (t)) :

The next step is to associate consumer welfare values to each state:

2
1 RH
pI (t) : WC =
2 2
RH
pII (t) : C2 =
2
2
WM 1 RH RH
pIII (t) : + C1 =
2 4 2 4
2
1 RH
pIV (t) : WM =
2 2
RH
pV (t) : C2 =
2
pV I (t) : 0:

Next, we simply sum up these welfare levels (adjusted by the probabilities) over time, accounting for the discount

factors. Notice that the sum of the losses is the same as the total size of the NDIS and thus the expected present

21
discounted value of total taxes in the economy, which thus enter as negative terms ( < 0):

1
X
t 1 WM
E(WCN A ) = p2t W C + p2(t 1)
(1 p)2 C2 + 2 p2t 1
(1 p) + C1 +
t=1
2
X1
t 1
+ 2pt (1 pt 1
) W M + 2pt 1
(1 p)(1 pt 1
) C2
t=1
RH =2 h p i R =2 h p
H
i
= (1 p) RH + RH 2(1 p) :
1 p2 4 1 p 2

1
Under no collusion ( < 2p2 ):

From a consumer welfare perspective, there are again six possible states at time t: (I) there is a competitive

duopoly and the deposits are returned with interest by both banks; (II) there is a competitive duopoly and all

deposits are lost because of the bad shocks to both duopolists; (III) there is a competitive duopoly and only one

bank receives a bad shock; (IV) there is a monopoly and the deposits are returned with interest; (V) there is a

monopoly and the deposits are lost because of a bad shock; (VI) there is no market at all (banks have exited and

consumer welfare is nil). These events occur, respectively, with the same probabilities pI (t) through pV I (t) that

we discussed above; it is just that "collusive duopoly" has to be replaced with "competitive duopoly". However,

the welfare levels associated to each probability are di¤erent:

2
RH
pI (t) : WC =
2
pII (t) : N C2 = RH
C 2
W RH RH
pIII (t) : + N C1 =
2 4 2
2
1 RH
pIV (t) : WM =
2 2
RH
pV (t) : C2 =
2
pV I (t) : 0:

We sum again these probability-adjusted welfare levels over time, to obtain:

1
X
t 1 WM
E(WNNCA ) = p2t W C + p2(t 1)
(1 p)2 N C2 + 2 p2t 1
(1 p) + N C1
t=1
2
X1
t 1
+ 2pt (1 pt 1
) W M + 2pt 1
(1 p)(1 pt 1
) C2
t=1
RH =2 p RH =2 h p i
= RH + RH 2(1 p) :
1 p2 2 1 p 2

22
A.2.2 State aid

We proceed in the same fashion as before. However, here, banks never exit. When consumers’deposits are lost due

to the bad shocks, the Government refunds them the original capital, as well as paying the direct rescuing costs .

1
Under collusion ( 2 ):

There are three scenarios that can characterise the economy at any period t: (I) both banks have a good shock

and return deposits with interest (which occurs with probability pbI (t) = p2 ); (II) only one bank receives a bad

shock (probability pbII (t) = 2p(1 p)); (III) both banks receive a bad shock (b
pIII (t) = (1 p)2 ). This is true every

period and each probability is associated to the following welfare levels:

2
1 RH
pbI (t) : WM =
2 2
2
WM 1 RH RH
pbII (t) : + C1 =
2 4 2 4
RH
pbIII (t) : C2 2 = 2 :
2

We can therefore sum this stream of expected payo¤s and then subtract the present discounted value of the total

tax bill (NDIS and direct aid):

12
X
t 1 WM
E(WCA ) = fp2 W M + 2p(1 p) + C1 + (1 p)2 ( C2 2 )g
t=1
2
RH =2 p 2(1 p)
= [ RH + p 1] :
1 4 1

Under no collusion ( < 21 ):

We proceed exactly as in the case of collusion. The probabilities are the same as those derived above but the

associated welfare levels are di¤erent:

2
RH
pbI (t) : WC =
2
WC RH2
RH
pbII (t) : + N C1 =
2 4 2
pbIII (t) : N C2 2 = RH 2 :

23
Summing up over time:

1
X
t 1 WC
E(WNAC ) = p2 W C + 2p(1 p) + N C1 + (1 p)2 ( N C2 2 )
t=1
2
RH =2 2(1 p)
= [pRH 2(1 p)] :
1 1

A.3 On the direct cost of rescuing

In this section we show the positivity of bN C and bC . As for the threshold bN C :

1 RH =2 RH =2 h p i RH =2 h p i
bN C = [pRH 2(1 p)] R H RH 2(1 p) 0
2(1 p) 1 1 p2 2 1 p 2

or
1 1=2 1=2 1 1
pRH + 2(1 p) + 0:
1 1 p2 1 p 1 1 p
| {z } | {z }
>0 <0

It can be shown that the …rst term is larger in absolute terms. As a result the threshold is positive. To see this

remember (Condition 1) that RH > 4 1 p p . Consequently we have that pRH > 2(1 p). Further notice that

1 1=2 1=2 1 1
> +
1 1 p2 1 p 1 1 p

which gives us the result that bN C is positive. As for the threshold bC :

1 RH =2 h p i RH =2 h p i RH =2 h p i
bC = RH (1 p) (1 p) RH RH 2(1 p) 0
2(1 p) 1 4 1 p2 4 1 p 2

or

p 1 1 1 1 1 2
RH + + (1 p) + 0:
4 1 1 p2 1 p 1 1 p2 1 p
| {z }
>0

For the same argument as before we have p4 RH > (1 p). Moreover

1 1 1 1 1 2
+ > +
1 1 p2 1 p 1 1 p2 1 p
1
> 0
1 p

24
which is true since p < 1, hence bC is positive.

A.4 Proof of (i) in Proposition 2

Proof. By contradiction. Suppose that

RH =2 h p i 2(1 p)
E(WCA ) = RH + p 1
1 4 1
RH =2 h p i R =2 h p
H
i
E(WNNCA ) = RH + RH 2(1 p)
1 p2 2 1 p 2

or
p 1 hp i 1 hp i
RH (1 p) RH + RH 2(1 p) : (10)
4 1 p2 2 1 p 2

Note that RHS in equation (10) is continuous and decreasing in if p4 RH + p 1 > 0, which holds by assumption.19

To see this take the derivatives:

@ 1 (1 p2 ) + (1 )p2 p2 1
= = <0
@ 1 p2 (1 p2 )2 (1 p2 )2

@ 1 (1 p) + (1 )p p 1
= = < 0:
@ 1 p (1 p)2 (1 p)2

Yielding:
@RHS p2 1 h p i p 1 hp i
= R H + RH + 2p 2 < 0:
@ (1 p2 )2 2 (1 p)2 2
| {z }| {z } | {z }| {z }
<0 >0 <0 >0

Moreover:
1 1
< :
1 p2 1 p

Consequently the RHS of equation (10) is biggest for the smallest value of in the given range, which is = 21 .

After plugging in we arrive at:

1=2 h p i 1=2 p
2
R H + [ RH 2(1 p)]
1 p =2 2 1 p=2 2
1 hp i 1 hp i
= R H + RH (1 p) :
1 p2 =2 4 1 p=2 4
1 9 Recall that RH > 4 1 p p .

25
Thus we have:

1 hp i 1 hp i p
R H + RH + p 1 RH + p 1
1 p2 =2 4 1 p=2 4 4
p2
(1 p) ( 12 4 )
RH 4 p2
:
p ( p1 4 )

1 p2
Leading to a contradiction since 2
1
4
p2
< 1 has to hold since p 2 [0; 1] and because of the assumption that
p 4

RH > 4 (1 p p) . We thus reach the result stated in Proposition 2(i): E(WCA ) < E(WNNCA ) ; 8 p 2 (0; 1) ^ 1
2 < 1
2p2 .

26