The Economics of Financial Intermediation: Overview Papers
Lecture 1
Spulber – Market Microstructure and Intermediation
• Market microstructure: study of intermediation and the institutions of exchange • Intermediation activities of firms are determined by the type of information
imperfections that are present
o Immediacy: solves randomness of supply and demand o Brokering: solves uncertainty about willingness to pay
o Guarantees for product quality: solves unobservability of characteristics of buyers or sellers
o Monitoring: solves cost of observing actions of sellers and buyers • Main function of intermediaries: clearing the market (pricing should match
purchases to sales) • Intermediation
o Three types of agents in an economy:
§ Consumers, market-taking firms and market-making firms (intermediaries that create and operate markets)
§ Presence of intermediaries modifies the familiar circular flow of economic activity
o Intermediation activity in the US accounts for about a quarter of GDP o Companies combine manufacturing with merchant activities, operating
markets for goods and services and factors of production • Price setting and market clearing
o Price setting can be costly and rigid for firms (menu costs)
o Profit maximizing firm sets prices to equate its marginal revenue to its marginal factor cost
• Providing liquidity + immediacy
o Avoid problem of double coincidence of wants
o Intermediaries provide immediacy by holding inventories and cash – pricesetting intermediaries will adjust prices to maintain inventories
o Market maker must deal with informed and uninformed traders, dealing with informed traders will result in a loss
• Matching and searching
o Avoid the cost of decentralized search
o Intermediaries must compete with decentralized search (eg car market) § Disadvantage of decentralized search: buyer has incentive to
understate willingness to pay and seller of overstating opportunity costs (asymmetric info)
§ Buyers can choose between using intermediaries to trade at a known price vs the uncertainty of decentralized market
§ Search costs: consumers and suppliers continue to face search costs from visiting multiple intermediaries (when higher discount rate or
search cost, this lowers number of consumers and raises number of firms because then firms can ask more to avoid search)
• Guaranteeing and monitoring
o Collecting and supplying of info, along with other services o Lemons model for cars
o Introducing monopoly intermediary to market with adverse selection improves efficiency
o Warranties and contract terms
o Delegated monitoring: like a contractor when you are building a house, you only need to monitor the contractor
Spulber – Adverse selection in financial markets
• Intermediation:
o when there are returns to scale in producing and distributing information about market demand and customer requirements
o when there is asymmetric information (benefits are then reduced by information rent though)
o asymmetric info about borrower characteristics: banks use interest rates as a screening device to separate high risk from low risk borrowers
• Informed vs uninformed traders o Uninformed: trade for liquidity
o Return of dealing with both are affected by competition between intermediaries
• Insiders, liquidity traders and specialists
o Classic adverse selection problems when the market maker cannot distinguish between informed and uninformed traders – then pooling contract based on average valuations becomes problematic (the market maker no longer faces the average trader and this is no longer economically feasible), then profitability depends on returns to providing liquidity services o Model:
§ Expected loss from dealing with an insider: H(p,w) with p=ask, w=bid and v=valuation (uniformly distributed on the unit interval, so
E(v)=0.5)
§ Liquidity trader has D(p)=1-p and S(w)=w. Expected revenue from dealing with a liquidity trader is J(p,w)
§ Normalize number of traders to 1. (1-alpha) is proportion of liquidity traders
§ Proportion of informed traders cannot exceed 2/3 to still make a profit
o Bayesian updating by specialists
§ Insider decisions provide information to the specialist about the insider’s knowledge
§ Case of perfect competition: profits to zero and not necessarily demand=supply
§ The greater alpha, the bigger the bid-ask spread for a history of trades. The greater the proportion of insiders then, the more info will
be conveyed by trades and this will then narrow the spread. If alpha too large, markets can break down again
o Competition between specialists
§ No free entry for market makers (only n registered market makers) § Every market maker is committed to trade any quantity q
§ Value of the asset can be 1 or -1 with equal probability
Lecture 2
Snider and Youle – The fix is in: detecting portfolio driven manipulation of the Libor
• Libor manipulation: admissions of manipulation by RBS, UBS and Barclays
• Libor calculation: truncated average of the reported borrowing costs of a panel of large banks
• Manipulation incentive: this paper focuses on panel bank portfolio exposure to the libor (not on reputation incentives)
o Bank profits depend on actual fix of the rate but they face misreporting costs o The model predicts a particular form of “bunching” in the intraday
distribution of libor quotes
o This paper statistically distinguishes “too much” bunching of a given bank’s quotes around the pivotal quotes
§ Primary assumption: bank borrowing costs should be correlated through similarities in the banks themselves
o In reputational theory of misreporting, banks don’t care about overall fix of the rate (and this also allows for a maintaining-stability-in-the-public-interest type justification)
o Potential solution: anonymously submitting: this would solve the reputational issue but would exacerbate the portfolio incentives by decreasing detection costs
• Libor
o Rate at which banks in London offer unsecured Eurodollar deposits
o In financial crisis, submissions were out of line with what one would expect from CDS spreads (should have been tightly correlated) or from Eurodollar bid rate (this should behave like bid-ask spread with libor)
o Misreporting for trading reasons (unlike that for reputation reasons) was initiated by individual traders
• Model
o Each bank chooses their quote q and has an actual cost c drawn from some joint distribution H
o Graph
Abrantes-Metz and Sokol – The lessons from libor for detection and deterrence of cartel wrongdoing
• Anomolies with libor:
o Libor did not move for over a year until the day before the financial crisis of 2009
o submitted quotes moved simultaneously
o libor was unresponsive to changing market conditions
• Econometric screens: a tool to both improve detection of potential price fixing cartel behaviour and police illegal firm behaviour
• Rules of screens:
o One size does not fit all
o If you put garbage in, you get garbage out • Requirements of screens:
o Understanding of the market o Theory on the nature of cheating
o Theory on how the nature of cheating will affect market outcomes o Design of a statistic that captures the key factors
o Empirical or theoretical support for the screen o Appropriate benchmark to compare evidence • New antitrust paradox part I and II
o First paradox: lack of sound economic analysis in antitrust law
o Second paradox: DOJ’s Antitrust Division’s criminal enforcement differs from the rest of the Antitrust Division in its lack of econometric screening
• How screens fit into antitrust enforcement
o Leniency program: destabilizes a cartel through defection of a cartel member who reports the cartel activity (current detection rates of around 20 percent) o Screens complement leniency program: they draw enforcers’ attention to
anomalous behaviour. Leniency is more likely to fail to detect some of the most successful cartels (cause self-selected)
o A company may run its own cartel screen to ensure that its own compliance systems are effective
Lecture 3
Bufton and Rivero – Clearstream: general court confirms commission decision
• Crime: refusing to supply clearing and settlement services to one of its customers (Euroclear) and applying discriminatory prices to that customer
• Clearing: process by which the obligations of the buyer and the seller are established • Settlement: transfer of securities from the seller to the buyer, transfer of funds from
buyer to seller and corresponding entries in the securities accounts
• Primary clearing and settlement: only the financial institution that keeps the securities in custody can do that
• CSD:
o Entity which holds and administers securities and enables securities transactions to be processed through book entry for trades of those
securities that have been deposited with it and which it holds in final custody o Clearstream is Germany’s only CSD
• The 2004 decision:
o Clearstream was said to hold a dominant position in the market for the provision of primary settlement and clearing
o Use of secondary clearing and settlement through an intermediary could not be a substitute because not same level of service
• Market relevance:
o EU Commission has urged market players involved in EU cross-border securities to promote cheaper and more efficient services
o In 2006, Code of Conduct on clearing and settlement: aims to enhance transparency and increase competition (monitoring group of the code of conduct – MOG)
• The judgment:
o Distinction between primary and secondary settlement and clearing services: § primary is when securities are kept in final custody by the same entity
which does clearing and settlement and when a change occurs on securities account
§ secondary is performed by intermediaries on their own books as a result of internalisation of mirror operation
o Clearstream contested definition of the market: they thought it should be looked at from a perspective of end customers
o This was rejected: when defining markets the starting point is the need expressed by those requiring the product or service in question (Euroclear in this case)
§ In that sense, Clearstream held a dominant position as an unavoidable trading partner
• The abuse
o Refused access o Charged higher fees • Refused access
o Clearstream argued it was the complexity of the connection, as well as technical problems of Euroclear
o False cause they provided access to other customers in a matter of months o CFI underlined responsibility of dominant position not to allow their conduct
to impair genuine competition • Higher fees
o Clearstream denied cause Euroclear received different service packages
Van Cayseele and Wuyts – Cost efficiency in the European securities settlement and depository industry
• EU settlement and custody institutions operate in an efficient way?
o Estimate translog cost function and constant elasticity of substitution – quadratic cost function
o Clearly economies of scale o Economies of scope as well • Introduction
o CSD: investors either hold account through this or indirectly through a financial institution who then has an account at the CSD
o Horizontal consolidation has taken place: Euroclear has become the CSD for a number of EU countries
o Question is: should consolidation continue?
o Trade-off between efficient infrastructure for domestic trades and efficient infrastructure for cross-border trades
o Other functions CSD: keeping track of stock, communicating general meeting… Should this be done by same entity?
o Through estimating cost functions, we will research economies of scope and scale
• The EU settlement and depository industry
o Trading of a security: transfer of the security + transfer of funds (should be efficient and secure)
o Types of activities: § Clearing
§ Settlement (majority is immobilised/dematerialised)
o Custodian: holds a so-called “omnibus account” with the CSD, can settle trades on their own books
o When counterparties are not with same custodian, CSD takes care of the transaction between omnibus accounts of custodians
o DVP: delivery vs payment – the accepted standard for settlement in financial markets (means security is only transferred when payment is received) o Separation between asset and issuance servicing activity and settlement
services, hypothetical situations:
§ All the securities are deposited in a single CSD and all the securities accounts are held by monopoly local agent bank (he can then settle all transactions). The CSD does the dividends and gives them to the local agent bank
§ Fully integrate asset and issuance servicing activities with the
settlement function; Everything settled on accounts of CSD; dividend policy can be implemented directly (Finnish model)
§ Intermediate model: some investors have account with the CSD, others with local agent bank
• In this model, when investors are local agent banks who all have account with large global custodian (monopolist) and CSD raises prices for this monopolist, it is not necessarily a bad thing. Positive network externalities may outweigh
disadvantages o Positive network externalities
§ Two-way networks:
• any new participant in the network creates 2N goods • for N clients, the size of the product space is N*(N-1)
• Then solution for the problem becomes: either fully integrated CSD (then CSD gets all network externalities), either fully separated (bank “)
• Analytical solution:
o Asset servicing is one way network: M*N (with M companies and N households)
o Question is now: does the combination of the one-way with the two-way network increase efficiency? Yes because:
§ Efficiency:
• Keeping the account of an investor for asset servicing can also be used without extra cost for settlement
• Can avoid errors § Network externalities:
• Size of the product space increases both in N and M for both activities: efforts to attract an additional investor yield benefits on both the investor and emittent side of the market for a fully integrated CSD – it then becomes a platform
o Disadvantages of combining both activities: § X-inefficiency
§ Abuse of market power • The model
o Variables
§ In addition to cost efficiency, we also look at synergetic effects now § TC: total operating expenses including depreciation
§ Clients: number of clients, comes closest to N. N^2 (or actually N*(N-1)) is the size of the grid, and this is a cost driver (rather than the number of transactions)
§ Sechold: reflects depository and issuance services. We use the
number of securities rather than the value of securities, comes closest to M.
§ Input variables: two prices. Lc and Oc: labour costs and other costs. o Translog cost function
§ Ln(TC)=f(ln(Clients), ln(Sechold); ln(Lc), ln(Oc)) § Second order Taylor expansion: see paper
§ Share equations: basically derive model to ln(Lc) and ln(Oc) § Economies of scale: derive to ln(Q); MC/AC
§ Economies of scope: ln(TC(clients,
sechold))<ln(TC(clients,0))+ln(TC(0,sechold)
§ Translog cost function is not stable in view of verifying the scope properly (because of taking an ln of zero). Depending on values of coefficients, it will be impossible to conclude in favour of economies of scope or economies of scope will be forced upon the function (flip-flop property of translog cost function)
§ Possible solution to flip flop: restricted version (restrict second order output effects to be zero – consequences: constant returns to scale and impose diseconomies of scope but then the statistically
significant negativity of the interaction term coefficient points towards economies of scope!)
o Quadratic cost function (CES-Q function)
§ Estimating a constant elasticity of substitution solves the flip flop § Model: see paper
§ Parameter r is an indication of the substitution possibilities for a CSD – this has to be estimated using the maximum likelihood method
§ Then you can easily test for economies of scope (see paper)
Lecture 4
Carletti et al. – Bank Mergers, Competition and Liquidity
• Model of the impact of mergers on:
o Loan competition (impacts the next two)
o Reserve holdings (increase through internalization effect or decrease through diversification effect)
o Aggregate liquidity (merger leads to more precise estimates of liquidity needs)
• Reserves
o Needed to cope with uncertainty about depositors’ time of withdrawals (stochastic as a fraction)
o When the premature withdrawals exceed the reserve holdings, a cost is incurred to obtain from the interbank market the liquidity needed o This is why bank’s demand for reserves depends on its uncertainty about
withdrawals and on the relative cost of refinancing (ratio of the cost of borrowing on the interbank market in case of liquidity shortage to the cost of raising more deposits and holding more reserves initially)
o When there is excess aggregate demand on the interbank market, the central bank must intervene
• Merger effects
o Impact on reserve management:
§ modifies the uncertainty of deposit withdrawals (diversification) § creates internal money market (internalization)
§ ceteris paribus, internalisation increases the marginal value of each unit of reserves (cause it can now be used to meet withdrawals from both banks)
§ internalization effect dominates when relative cost of refinancing is low; diversification dominates when relative cost of refinancing is high § in any case, merged banks benefit from scope economies in their
liquidity management by raising deposits in two imperfectly correlated deposit markets
o Impact on market power and loan market competition:
§ Loan rates increase when market power effect dominates
§ Loan rates decrease when cost efficiency effect (from reoptimisation of reserve holdings and a potential reduction in lending costs) prevails o Impact on interbank market and aggregate liquidity
• Higher aggregate liquidity supply and lower expected aggregate liquidity needs when banks increase their reserve holdings
• Opposite when reserve holdings are reduced § Asymmetry channel
• Merger inducing greater asymmetry among banks increases the variance of liquidity demand, and increases expected aggregate liquidity needs
• Opposite when smaller asymmetry
§ Total impact depends on interaction between the reserve and the asymmetry channel
• Insights of the model:
o Mergers between large banks lead to a “polarization” of the banking system and lead to higher aggregate liquidity needs than mergers involving small banks (asymmetry)
o Mergers are more likely, ceteris paribus, to increase aggregate liquidity needs in developing countries than in industrial ones, since they induce lower individual reserve holdings in less efficient markets, where banks face higher refinancing costs
o Effects of mergers on loan competition and aggregate liquidity tend to be complementary in industrial countries but not in developing ones
o Impact of mergers on reserve holdings and aggregate liquidity may depend on the phase of the business cycle (in upturns, mergers may affect reserves more negatively than in downturns)
• The model:
o Three date (T=0,1,2) economy with three classes of risk-neutral agents: N banks, entrepreneurs and individuals
o Competition in the loan market:
§ each bank faces a linear demand for loans, gamma represents the substitutability of loans (gamma>0)
o Deposits, individual liquidity shocks and reserve holdings: § Banks raise deposits in N distinct regions
§ Depositors are offered demandable contracts (just initial investment at T=1 and a net rate r_D at date 2)
§ Deposits are subject to liquidity shocks: a fraction delta_i of
depositors develops a preference for early consumption. This fraction is stochastic and distributed between 0 and 1 (uncertainty at each bank AND in the aggregate)
§ At T=1, each bank faces a demand for liquidity: x_i=delta_i*D_i and uses R_i to satisfy this
§ At T=0, each bank faces a liquidity risk given by phi_i and liquidity shortage given by omega_i
o Interbank refinancing and aggregate liquidity
§ At T=1, interbank market opens where banks can lend or borrow depending on whether x_i< or > than R_i
§ Both banks and central bank operate in this money market § Banks can borrow at r_IB and lend at rate r_IL
§ Given presence of aggregate uncertainty, there can be aggregate excess or shortage of private liquidity
§ Large phi and large omega, are the aggregate private liquidity risk and shortage
§ Those are indicators of the frequency and the size of central bank operations
• The Status Quo
o All banks are identical o Assumptions:
§ No bank runs: illiquidity of loans and r_d>0 guarantee that depositors will not withdraw prematurely unless hit by liquidity shocks
§ Loan market sufficiently profitable s.t. banks can operate in the interbank market
o Date 0 maximization problem
§ Pick r_i and R_i so as to maximize the expected profit
§ Profit function is sum of profits from loans, losses from deposits and profits/losses in the interbank market
§ Bank’s financing cost: the losses in interbank market combined with losses from deposits
§ Convenience: r_IL=0 and r_IB=r_I, this simplifies the equation
§ In determining R_i, a bank then makes the trade-off between the cost of satisfying the expected liquidity needs through interbank market vs raising more deposits and keeping them ---- depends on uncertainty and cost to borrow liquidity
o Proposition 1: the symmetric status quo equilibrium § See paper (without proof)
o Implications of proposition 1:
§ Optimal reserve-deposit ratio: k_sq (see paper)
§ In status quo equilibrium: each bank has liquidity risk phi and liquidity needs omega (see paper)
• Logical: liquidity risk and expected liquidity needs are increasing in deposit rate and decreasing in refinancing cost
Van Cayseele – Financial consolidation and liquidity: prudential regulation and/or competition policy?
• Introduction
o Consolidation within EU: has led to very concentrated banking market – concerns of market power of banks vs depositors and borrowers + concerns for liquidity
o Ambiguity: mergers can also reduce loan rates because of efficiency gains o Cost efficient vs revenue enhancing mergers, kinds of mergers:
§ Revenue enhancing:
• Increase market power
• Improve liquidity management by enhancing the revenue basis (we incorporate cost savings resulting from a better pooling of resources within a group)
§ Cost efficient:
• Improve on operational cost efficiencies
o Assumption: sources of market power are equally sources of increased liquidity risk
o Localized loan competition with liquidity shocks originating in the real part of the economy
• Survey of the literature
o Mergers, banking competition and antitrust issues
§ Appropriate to focus on mergers in banking markets where competition is localized (tastes are Kaldorean rather than Chamberlinean)
o Banks as liquidity providers and prudential regulation
§ If liquidity demands on both sides of the balance are to some extent negatively correlated, important synergies can be exploited from a single buffer stock of liquid assets
§ Firms who experience a boom need cash instantaneously: in order to fund the loans when they realize, banks will secure core deposits to meet part of the loan demand. Once the loans realize, additional funding will be obtained from an LOLR or through the interbank market
o Bank mergers, competition and liquidity
§ Consequences of mergers on competition and liquidity management § Shock either perfectly correlates in a positive or a negative way
(whereas CHS consider no correlation or negative correlation between the post merger liquidity risks)
o This paper
§ Focus in this paper is on the management of the liabilities side of the balance sheet whereas the liquidity needs come from the asset side of the balance (loans)
§ Banks will first use core deposits to cover loans and then seek additional funding from the LOLR
§ Single LOLR who has monopoly power to banks who seek additional funding and monopsony power vs banks who have accumulated liquidity they don’t use
§ Balance sheet in case of depression/ in case of boom (see paper) § Rates paid for deposits are given exogenously here and in CHS § Impact of merger on market power: unilateral effect
o Similarities and differences between CHS model and this one: important! (see paper)
§ IMPORTANT: in this paper, we have the possibility to endogenously determine the size of a cost advantage associated with a particular merger (CHS distinguish between operating cost and flat rate
borrowing in IB market). We focus on LOLR lending at increasing rates § Decrease resp. increase in loan rate is indication of reduced resp.
increased market power
§ Expected Liquidity Shortage to measure liquidity demand • Model of spatial competition
o Assumptions:
§ Economy with 6 banks, 2 distinct regions
§ Within each region, firms a priori could open a credit line with each of the 3 banks against which they can borrow (up to a fixed limit of 1) § Every firm will at most consider 2 banks to borrow from
§ Banks in Northern region: N, SW, SE § Banks in Southern region: S, NW, NE
§ Any firm located at x derives indirect utility v-(r_n+tx) when borrowing from N, and v-(r_se+t*(1-x)) when borrowing from SE § Reservation value v is sufficiently high so as to keep the market
covered at all times
§ Whenever a liquidity shock hits a region, all credit lines are used to borrow against: perfect positive correlation for the banks within a region, perfect negative correlation of two regions against each other § Probability theta that North is hit, (1-theta) for south
§ Rate paid to LOLR is increasing function of y, cost of operation is r(y)*y and r(y)=r*y
§ LOLR monopsony power so banks have to leave their liquidity at bank and carries no interest L
o Timeline
§ Banks announce loan rates and secure term deposits § Firm decide on identity of bank to open credit line § State of the economy realizes
§ Firms in booming region borrow against open credit lines
§ Banks in booming region seek additional funding with LOLR while banks in depressed region transfer funds to LOLR
o Profit function
§ North (see paper) § South (see paper) o Policy measures:
§ Stability criterion: expected shortage of liquidity (ELS)
§ Loan rates criterion: post merger loan rates compared to pre merger levels (SSNIP test – Small but Significant and Non-transitory Increase in Price – usually fixed at 5 per cent)
• The model: analysis and results
o Nash Equilibrium in pure strategies: Status-Quo o Types of merger:
§ Intraregional: within § Interregional: between
§ Within with operational efficiencies § Giga merger
• Status quo model:
o Derive profit function to find r and T in status quo for north and south o Calculate average loan rate
o ELS and P are minimal when theta=0.5 (symmetric case)
o Consider symmetric case from now on, these are the benchmark values against which the different mergers will be compared
o Important to note:
§ Increases in t always reduce consumer surplus § Increases in r charged by LOLR always reduce ELS
§ t does not affect ELS and r by LOLR does not affect consumer surplus § Term deposit rate r influences both ELS and consumer surplus • Intraregional mergers/within
o Eg N and SE merge: they will charge r_m from now on and the region becomes a duopoly
o Profit function of merged firm: just fill in the general one o Consequences:
§ Loan rate of both the merged entity and the outsider increase § Term deposits increase
§ ELS decreases
o Good for prudential regulation, bad for competition o Merged entity violates SSNIP5 when t>3c/37+6r/37 o Profits increase for all banks, so this merger will happen • Interregional mergers/between
o Eg N and S merge: only cost side is affected, revenues are simply twice those of the status quo
o Consequences
§ Loan rates of both merged entity and outsiders decrease § Term deposit holdings decrease
§ ELS increases
o Good for competition, bad for prudential regulation o Profits increase for all banks, so this merger will happen • Intraregional mergers with operational cost efficiencies
o Eg N and SE merge: reduce operational cost to zero o Profit function: fill in with c=0
o Consequences:
§ Depends on magnitude of cost reduction whether loan rates go up or down
§ t<c (loans are not too differentiated products), then decrease in loan rates and increase in term deposits ---- win-win situation!
o Win win if t<c
o Profits increase for all banks, so this merger will happen • Giga mergers
o Eg all banks except SW and NE o Consequences:
§ Loan rates can increase or decrease § ELS can increase or decrease
• The model: policy implications
o For within and between, always policy trade-off
o For within with cost efficiencies and for giga, it depends on the values that the parameters take
o Types of merger review process
§ Polyarchy (“” only by one instance) – this will clear everything, so we consider one stop shop principle instead (meaning either competition policy, either prudential regulation dictates whether or not the operation is cleared)
§ Competition law/general rule vs exemption approach/prudential regulation
§ Look at table in paper on which organisations clear which types of mergers
§ Know all the conditions • Extensions
o Bank runs can be included by making the assumption that not all loans will be paid back and that merged banks might be bailed out (moral hazard)
o Interregional mergers could then trigger contagion ---- so an interregional merger may then not lead to the desired liquidity management
improvements
• Market structure: LOLR-behavior
o We can endogenize the function of the price of the funds that the LOLR lends o His revenues enter his objective function as a profit
o Inflation enters into the objective function as a penalty for too much money creation
o When we used the r(y)=r*y, we can think of which r the LOLR would set? o Simple objective function for the symmetric shock case (see paper) o Consequences:
§ When the LOLR chooses r as to maximise his profits, taking inflation into account, the market will propose 49% of within mergers and 51 between
§ When LOLR determines r as Stackelberg leader, then complex problem
• Market structure: increased contagion possibilities
o Incomplete market structure as opposed to the disconnected version: two regions brought together
o Shock moves along the circle, each time different starting point and
extending 180 degrees in clockwise sense – any bank can be in a cluster with any other one
• Regions as nations
o Each having the same or a different policy mix
o Then you need to follow Sah-Stiglitz hierarchy principle
o Particular case: between type of merger when the different countries are differentially affected by liquidity demand
§ Like a bank leaves the safe territory of the home market to compete in foreign, more dangerous environment (higher liquidity shortages)
Focarelli and Panetta – Are mergers beneficial to consumers? Evidence from the market for bank deposits
• Trade-off between efficiency effect and market power effect – they will have an opposite effect on deposit rates for consumers (not loan rates in this case!)
• We take deposits instead of loans because those are not influenced by asymmetric information (borrower risk) – we take household deposits and current accounts (homogenous products)
• Competition is at local level with barriers to entry (good, cause then we can observe how mergers really change the market structure)
• Bank deposits are highly standardized product
• This paper considers a longer period after the merger cause improvements in efficiency may only emerge after some time (cost-cutting takes time, reluctance to lay off staff, merging disparate workforces), typically for banks:
o Difficulties in refocusing lender policies o Rationalizing branches
o Integrating data processing systems and operations
o Training the personnel of the target to market new owner’s products o Culture clashes (relationship banking, soft information)
• In-depth analysis of the relation between market structure and the pricing effects of M&As
• Results:
o Short run consolidation increases market power and lowers deposit rate by about 16 basis points
o Long run deposit rates go up by 13 basis points compared to premerger level o Compare price changes induced by in-market mergers with those induced by
out-of-market mergers (those had indeed no short term change and only long term increase)
o Alternative hypothesis: the increase in price is just because of deterioration in quality (not true, rate changes are not explained by proxies for service quality – rates rise only for those banks that are successful in reducing costs after merger)
• Market power and Efficiency effects o Efficiency
§ Cost-saving technologies § Spread fixed costs more § Economies of scope § Managerial efficiency
o Dynamic analyses: conclusion is that in-market consolidation generates substantial market power (problem with this is that the analysis is too short-term cause market power can be exercised immediately after the deal whereas value gains take a long time – take three years as gestation period) o Separation between long run and short run consequences
§ Investigate over nine years
Lecture 5
Abrantes-Metz - The Power of Screens to Trigger Investigations
• Screen: statistical test designed to identify industries where competition problems exist and to further identify which firms and individuals are involved in a conspiracy or manipulation
• Two types of screens:
o One that searches for improbable events
o One that uses the concept of a control group to identify anomalous patterns • Four characteristics of a good screen
o Minimize false positives and false negatives o Easy to implement
o Costly to disguise o Empirical support • Six uses of screens
o Establish likelihood of a conspiracy o Decide whether to apply for leniency o Class certification stage
o Internal monitoring for firms o Estimating damages
o Motions to dismiss • Stock options backdating
o SEC implemented requirement that stock options grants should be reported in two business days
o Countrywide had a knack of granting stock options to executives shortly before it released favourable news to the market
o Court case: arguments for and against (cherry-picking days, on-schedule/off-schedule grants)
• NASDAQ dealers conspiracy
o Changes in prices have a lower bound, the tick size o NASDAQ stocks were trading only at even eights
o Inside spread: lowest selling offer – highest buying offer
Christie and Schultz: Why do NASDAQ market makers avoid odd-eighth quotes?
• Several hundreds of firms act as dealers, 2-50 market makers per stock (they do not have exclusive franchise and are not as closely regulated)
• Spreads of one-eighth are virtually non-existent for a majority of the sample • All sorts of comparisons between NYSE/AMEX spreads and NASDAQ spreads • Potential other explanations
o Negotiation hypothesis: coarse price increments to reduce cost of
negotiation (unlikely cause large trades should then be less likely to occur on odd eighths but they are the ones that do)
o Economic determinants of spreads: logistic regression and maximum likelihood estimates (most significant variable is whether odd eighths were already used in January – whether there had been a practice established) • Tacit collusion among market makers: infinitely repeated game, round numbers used
as focal points
Lie – On the timing of CEO Stock Option Awards
• Abnormal stock returns are negative before unscheduled executive option awards and positive afterward
• Three factor model of Fama and French: o Overall market factor
o Factors related to firm size o Book-to-market equity • Preferred timing
o After anticipated future stock price decrease
o After recent price decrease that they believe to be unwarranted o Before anticipated stock price increase
• When focus on scheduled awards to remove possibility that results are attributable to opportunistic timing of the awards, they interpret it to suggest that executives opportunistically time the release of information around fixed option awards (but this paper focuses on unscheduled awards, so timing of awards rather than of releasing information)
• Stock return pattern for unscheduled awards is strong and striking (before: -3% and after 2%)
• Scheduled awards less strong (before: -1% and after 1%) • Ex post facto timing hypothesis: option backdating
• Insiders can forecast long-term marketwide movements but not short-term (so those should be retro-actively timed)
• Compensation committee is supposed to decide over size and timing of stock option grants but in reality, just ratifies what executives want
• Retroactive timing occurs in practice because:
o Effective and simple way of boosting value of awards
o Stock option plans are vague as to how grant date should be determined o Difficult for outsiders to uncover such practices
• Trends for unscheduled awards have become more distinct over time: executives have become more effective over time in timing the awards to their advantage • Assumption: executives have superior forecasting ability for future firm-specific price
changes but not for future marketwide movements: then it becomes clear it is impossible that they did not time at least some of the awards ex post facto
Abrantes-Metz.. – Libor manipulation?
• Unjustifiably low borrowing costs for the calculation of the daily LIBOR
• For the 7-month period that starts in 2007, intraday variance of individual bank quotes is not significantly different from zero
• Why would banks do this? Net borrowers would profit from lower rates, net lenders from higher rates
• Even when they would not influence the published rate in general, they may have a motivation to adjust the submitted quote relative to its peers (for public reputation, artificial narrowing of the cross-sectional variation in quotes; or for signalling pricing intentions to each other)
• Other reason: banks may use Libor as an hedge against rate fluctuations elsewhere • In summary: some of these reasons would lead to a change of Libor altogether,
others just to the range of submitted quotes narrowed down • Reputation reason:
o Reputation for effectiveness has been proven to be correlated with superior financial outcomes
o This is why institutions would manipulate • Screening methods (see Abrantes-Metz before):
o Search for improbable events (casino) o Use control group (concrete)
• Screening techniques used for Libor manipulation are combination of: o Price-fixing conspiracy
§ Some collusive markers: higher prices, reductions in price variance, lower responsiveness of prices to costs, declines in imports of substitutes..
§ Price indices o Bid-rigging conspiracy
§ Bidding patterns that are highly improbable under competitive bidding – bids should be independent (after controlling for common legit info)
§ Correlation between bids can be extremely high when there is collusion (Tenessee valley conductor cables)
§ Divergence between bids and costs in competitive markets can also be a give-away
o Market manipulation
§ Involves fewer members (sometimes single firm) § Does not focus on maintaining a fixed price level per se § Screening processes should be more individualized § Manipulation: causation / price artificiality
§ Manipulations induce noise in the market and distort market expectations about future prices
• Aggregate level: relationship between Libor and other major benchmarks (H1) • Micro-level: pattern of individual quotes (and compute pairwise correlations as in
bid-rigging screening) (H2)
• Finally: compare to CDS spreads as proxies for borrowing costs (H3)
• H1: not significantly different when you look at the historical difference/correlation between Federal funds rate and Libor vs the period considered
• H2: “deciding group of 8”, null hypothesis is rejected
• H3: consistency with CDS spreads, null hypothesis is rejected
Lecture 6
Diamond - Banks and Liquidity Creation: a simple exposition of the Diamond-Dybvig model
• Mismatch of liquidity: loans cannot be sold quickly at a high price whereas deposits can be withdrawn at any time (problems when there is a bank run)
• This model explains why banks choose to issue deposits that are more liquid than their assets and why banks are subject to runs
• Illiquid asset: asset of which the proceeds from physical liquidation on some date is less than the present value of the future payoff (the lower the fraction, the more illiquid)
• Demand for liquidity
o Consumers: they are uncertain about their date of consumption, can be either T=1 or T=2; has probability t of being type 1 and 1-t of being type 2 (no aggregate uncertainty: e.g. t=0.25).
o Investor's expected utility: t*U(r_1) + (1-t)*U(r_2)
o Risk-averse utility function: U(c)=-1/c, then constant 1 is added to yield positive number. An illiquid asset has r_1=1 and r_2=2 and an illiquid asset has r_1=1.28 and r_2=1.813. Expected utility from holding liquid asset is then 0.391> utility from holding illiquid asset 0.375 (all investors want liquid asset) o Non risk-averse utility function: U(c)=c, then for illiquid asset expected
payoff: 1.75 and liquid asset: 1.68 (now all investors want illiquid asset) • Entrepreneurial demand for liquidity:
o May have a sudden need to fund a very high return project at date 1
o With probability t, he will be able to fund the high return project that returns phi per unit invested
o Values asset as: t*r_1*phi + (1-t)*r_2 if phi>r_2/r_1 and as t*r_1 + (1-t)*r_2 if phi<=r_2/r_1
o if phi=2.5 and r_2=2, then entrepreneur prefers liquid asset • Bank liquidity creation
o Bank can provide more liquid asset by offering demand deposits
o Bank has 100 deposits which it invests in illiquid assets. At t=1, the portfolio is worth 100. 25 depositors want their money (1.28 each), so the bank has to withdraw 25*1.28=32 assets. 68 will remain, which will yield 136 at date 2. Divide this among 75 investors remaining: 1.813
• The optimal amount of liquidity
o optimal levels of r_1 and r_2 will maximize the ex ante expected utility of each investor at date 0
o Interior optimum: U'(r_1)=R*U'(r_2), marginal utility is in line with the marginal cost of liquidity (eg U(c)=-1/c, then U'(c)=-1/c^2)
• Even more illiquid asset
o Now returns are 1-x if liquidated at T=1
o Bank now invests in enough short term assets to finance all of the date 1 withdrawals. Bank puts a fraction (t*r_1) of assets into short-term assets and the rest into illiquid (then same payoffs as before, called asset management of liquidity)
• Investors holding the assets directly
o cannot perform as well as the bank, cannot achieve r_1>1
o this is because an investor needs all or none of his liquidity while the bank knows that a fraction t of depositors will need liquidity at T=1
• Bank runs
o unverifiable private information which investor is of which type at T=1 o multiple equilibria: more than one self-fulfilling prophecy about who
withdraws at T=1 (bad equilibrium: all withdraw at T=1 because they all expect each other to do the same)
o Type 2 depositors will want to withdraw at date 1 if r_2(f)<r_1 o Bank runs then force banks to call in loans early
o Tipping point for a run is implying that r_1>=r_2, in the example f=0.5625 • Suspension of convertibility
o suspension is unpopular
o Problematic when there is aggregate uncertainty about t, then it may withhold some type 1 ppl from withdrawing which is costly
• Deposit insurance
o promise to pay the amount promised by the bank no matter how many depositors withdraw
o Governments have taxation authority o Good when aggregate uncertainty about t
o may cause incentive problems if bank regulation is poorly structured
Diamond and Dybvig – Bank Runs, Deposit Insurance and Liquidity
• Transformation service of banks: transform illiquid assets into liquid liabilities • Three important points:
o Banks issuing demand deposits can provide better risk sharing among people who need to consume at random times
o Undesirable equilibrium: bank run
o Even healthy banks can fail because of bank runs
• The results imply that real damage from bank runs is primarily from the direct damage occurring when recalling loans interrupts production
• Each agent has a state-dependent utility function • For all r_1>1, banks are susceptible to bank runs
• Bank run: worse allocations for all agents than without the bank. Everyone receives a risky return with mean 1
• Suspension of convertibility (when t is known)
o Simple variation on the demand deposit contract which gives banks a defense against runs: suspension of allowing withdrawal of deposits
o Then you remove the incentive of type 2 agents to withdraw early o Unique Nash equilibrium with f=t
• Stochastic withdrawals (t is unknown)
o Proposition 1: bank contracts (which must obey the sequential service constraint) cannot achieve optimal risk sharing when t is stochastic and has a nondegenerate distribution; proof that no bank contract can attain the full-information optimal risk sharing
o Government deposit insurance: guarantees that the promised return will be paid to all who withdraw, if this is a real guarantee, there are taxes in the form of inflation through the money creation. It can tax any agent that withdraws at T=1 (those with low values of f_j). Potential benefit from government is that a government can tax after agents have withdrawn o Proposition 2: demand deposit contracts with government deposit insurance
achieve the unconstrained optimum as a unique Nash equilibrium if the government imposes an optimal tax to finance the deposit insurance (never pays to participate in a bank run)
Shin – Reflections on Northern Rock: the Bank Run that Heralded the Global Financial Crisis
• Only after the central bank had announced its intervention to support the bank, retail depositors started queuing outside the branch offices (because that meant they failed to find a buyer)
• Northern Rock had a heavy reliance on nonretail funding (it was this nonretail short term funding that dried up) and virtually no subprime lending
• The bank run was an event in the aftermath of the liquidity crisis (retail deposit funding is perhaps the most stable form of funding available to a bank)
• Northern Rock started expanding its mortgage assets but its retail deposits could not follow. The percentage of retail funding of total liabilities had fallen to 23% and only a small proportion consisted of traditional branch-based deposits. Bulk of retail deposits were postal and telephone accounts (allowed them to expand beyond regional base but were more vulnerable to withdrawals)
• Gap in funding was made up by securitized notes, interbank deposits, “covered bonds” (long term liabilities written against segregated mortgage assets)…
• Special purpose entities were consolidated on Northern Rock’s main balance sheet (unlike US vehicles), notes issued by Granite; floating rate “controlled amortization notes”. Notes issued by Granite had longer maturity than similar type in the US. This is why the securitized notes issued by NR do not appear culpable for the run – at least not in a direct way
• In order to see what caused the run, we compare the composition of NR’s liabilities before and after the run
o Largest difference: liability to Bank of England after the run o Covered bonds increased
o Securitized notes only fell slightly
o Largest fall: retail deposits and wholesale liabilities
§ Wholesale liabilities: nonretail funding that doesn’t fall under covered bonds or securitized notes – it was said to be a mix of short and medium term funding with diversification. The short-term wholesale funding shared many similarities with the short-term funding raised by off-balance-sheet vehicles such as the “conduits” and the
“structured investment vehicles” -> initial run on NR was the nonrenewal of NR’s short-term and medium-term paper
§ Retail deposits: largest fall came from post account deposits, offshore deposits, telephone and internet deposits
• Classic models of bank runs describe a pattern of coordination failure. With NR, the withdrawal of credit hit the whole market, not a subset of institutions. Also, classic models assume individual rational investors whereas NR bank run was enacted by sophisticated institutional investors (often face constraints from certain rules) -à the run on NR can rather be seen as a “tightening of constraints on the creditors of NR”, rather than a coordination failure among them
o Leveraged financial firms: market conditions are pivotal in determining their leverage. When conditions turn hostile, sharp pullback in leverage (creditors demand a higher equity cushion to shield them from losses)
o Gauge of overall leverage and funding conditions is to look at the implicit maximum leverage possible in collateralized borrowing transactions such as repurchase agreements – repos (borrower sells a security today for a price below the market price on the understanding that it will buy back the security at a pre-agreed price, difference is haircut – this determines maximum possible leverage, through the reciprocal of the haircut ratio)
§ Increase in haircuts entails very substantial reductions in leverage, creating hard choices (sell assets / raise more equity)
• Vulnerability and leverage at NR
o Leverage is defined in principle by assets/equity (but you have to take right measure for equity)
§ Common equity: stake held by owners with voting power
§ Shareholder equity: common equity + preferred shares (no voting power but senior to shares of common equity in case of liquidation – like a perpetual bond)
§ Total equity: shareholder equity + subordinated debt (senior to common and preferred equity but junior to other types of debt) o Key is that the haircut is the equity stake held by the controlling party
(borrower should have sufficient ownership stake in the assets that it
controls so there is no moral hazard) -à so take common equity to calculate leverage
o NR was faced with giant margin call, where lenders demanded higher
haircuts. Usual way to meet this is to sell assets and raise the cash. But assets of NR were illiquid long-term mortgages
• Economic role of Short-Term debt
o What NR shows, is that creditors can be subject to external financial constraints and may have to take actions that are outside the immediate principal-agent relationship with the bank
o Eg: bank 1 has borrowed from bank 2. Bank 2 suffers credit loans on its other loans but bank 1 has good creditworthiness. Bank 2 has to reduce its overall lending to achieve its micro-prudential objective of reducing its risk exposure. But then bank 1 also feels like it had a run
o Maturity mismatch is double-edged: from the point of view of incentive effects, fragile balance sheet is desirable. However, spillover effects from outside the principal agent reglationships can generate countervailing inefficiencies
• Implications for financial regulation
o Cornerstone of financial regulation has always been capital requirements o Key determinant of the size of the regulatory capital buffer should be the
riskiness of a bank’s assets (Basel)
o Two other types of policy proposals in this paper:
§ Liquidity regulation (would reduce spillovers): bank can survive a run if
• It has sufficiently stable (illiquid) liabilities such as long-term debt
§ Limit on the raw leverage ratio, rather than risk-weighted assets: prevent build-up of leverage during good times. For both lender and borrower, the leverage constraint binds during boom times so that the imperative to reduce leverage is less strong in the bust
• Critique on this policy: does not take riskiness of assets into account (counter: is not intended to replace Basel, just to supplement it)
Lecture 8
Cheng and Van Cayseele – State aid and moral hazard in banking: the case of China in the late nineties
• Panzar and Rosse test: used in most literature on competition in banking, this assume profit maximization of banks – may not be realistic, since banks also have other objectives such as being “too big to fail”. This paper introduces a reduced form model where banks can pursue other goals and looks into moral hazard issues that may plague banks receiving state aid
• Problems with state aid:
o Removing level playing field for institutions that have (not) received state aid o Inducing future moral hazard behavior (risk-taking)
• Countering state aid:
o Refuse state aid (not often)
o Impose remedies after rescue (aim at restoring viability of institution and compensating distortions in the competitive playing field)
• Moral Hazard: behavior that deviates from profit maximization (eg wanting to become too big to fail)
• When trying to isolate this moral hazard, we need to rely on the occurrence of a true natural experiment in the data (often state aid is endogenous)
• Reduced form model:
o Panzar-Rosse test indicates: sum of revenue elasticity with regard to input price changes (H statistic) must be non positive when a monopolist
maximizes his profits
o Rejection of test can occur due to departure from monopoly (then this sum becomes positive), under the Panzar-Rosse test this means that market structure is different than a monopoly (eg oligopoly, perfect competition…) o Our model will reject this test for other reasons (moral hazard). The
theoretical value for H statistic under different assumptions is
straightforward to derive (relevant departures from profit maximization: revenue maximization, output maximization)
§ Revenue maximization: revenues from the granting of loans (TBTF), H statistic will be zero since changes in input prices will not affect the choice for the output level
§ Output maximization: many loans, many clients. Also TBTF, also typical for state-owned enterprises with a political goal of serving as
many clients as possible. Then banks will pick the output level for which price still covers costs. H will exceed one.
• Chinese Banking System and State Aid
o State-owned banks vs joint-stock commercial banks § Two-tiered banking system:
• Big Four specialized banks; policy-lending conduits for the government; lacked incentives to compete
• 11 joint-stock commercial banks in response to need for financing projects from the non-state-owned sector (they were also in a monopoly situation in their specific sub-markets)
o Non Performing Loans (NPLs) had gradually become a serious issue for the state-owned banks because state-owned enterprises (SOEs) had little incentives to pay. As a group, the big four banks had a negative net worth and were insolvent by Western accounting standards
§ Chinese government injected 33B into the four banks
§ Chinese government established 4 asset management corporations (AMCs) to purchase NPLs from the banks
§ They also stepped in for commercial banks but this did not guarantee continuation (bankruptcy of Hainan Development Bank)
o This is why state-owned and joint-stock commercial banks can serve as reference groups in identifying the impact of state aid on bank behavioural changes
• Empirical results and model
o Empirical model: we take the baseline model from Panzar-Rosse: see paper o Sum(beta) is the H statistic to be tested: hypothesis of a profit maximizer
cannot be rejected when H is non positive; null hypothesis of revenue maximizer cannot be rejected when H is zero; null hypothesis of output maximizer cannot be rejected when H larger than one
o Conclusions
§ State-owned banks gained significantly more revenue than commercial banks
§ SO banks paid sign more to get funding than their counterparts § Commercial banks have higher price of physical capex
o Including interaction term between post 1999 and state-owned allows us to distinguish the difference-in-differences for state-owned and joint-stock commercial banks before and after the state-aid period
§ Conclusions:
• Both commercial and SO banks enjoyed higher revenue after state aid but more for SO banks but difference shrunk after state aid
• When controlling variables are added, interaction term becomes insignificant
• Without controlling variables but with interactions between input prices and dummies,…
• Impact of State aid on behavioral changes in Chinese banking industry o Calculation of H statistic shows:
§ One cannot reject that before 1999, Chinese banks were output-maximized
§ Same for revenue maximized after 99
§ Inclusion of controlling variables seems to take away the effect of state aid
§ Objectives of SO banks have not changed after state aid (even more output maximization)
§ Objectives of commercial banks for revenue max go down
§ Pre and post 1998 (when commercial banks were alerted by that one bankruptcy): significantly different behavior for SO (output to revenue max) and commercial (output to profit max)
o Commercial banks received little state aid and altered away from output-max to profit-max much faster: indicating the importance of a “first time last time principle” in fighting moral hazard
