RDP 1999-05: Trends in the Australian Banking System: Implications for Financial System Stability and Monetary Policy 4. Consolidation: Efficiency and System Stability

In this section of the paper, we focus on the broad implications of consolidation for the stability and efficiency of a financial system. We leave a discussion of the implications of financial conglomeration and the rise of the competitive fringe to Section 5.

The debate regarding the impact of mergers between banks has been a longstanding one, both in Australia and around the world. This debate has typically emphasised the effect of mergers on efficiency. We begin the section with a discussion of efficiency in Section 4.1. In Section 4.2 we introduce a broad framework to analyse policy questions relevant to both efficiency and system stability, recognising that there may be a trade-off between these two objectives. In Section 4.3 we use this framework to analyse the impact of consolidation on system stability.

4.1 Consolidation and Efficiency

Many studies find a positive and significant relationship between market concentration and measures of bank profitability. There are two alternative hypotheses that might explain this result – with different implications for economic efficiency.[10]

The Structure-Conduct-Performance (SCP) hypothesis is that a more concentrated market permits banks to behave in non-competitive ways so as to boost their performance (usually in terms of profits).[11] If this hypothesis is true, consolidation will lead to higher prices for consumers and a reduction in economic efficiency.

The Efficient-Structure (ES) hypothesis is that some banks have greater efficiency, and hence, profitability. These banks increase their market share, either by gradually forcing out less efficient banks or by merger and acquisition. According to the ES hypothesis, it may be that some banks are inherently more efficient than others, perhaps through idiosyncratic management ability. Alternatively, it may be that there are economies of scale or scope which allow larger banks to force out smaller banks.[12] In either case, the ES hypothesis implies that greater concentration will be accompanied by a mixture of higher profits and lower prices (and/or better services) for customers and hence, greater economic efficiency over time.

On balance, a review of the evidence from a multitude of studies does not strongly support one hypothesis in favour of the other (Berger and Humphrey 1997). A consistent finding is that although some consolidations improve cost efficiency, others worsen the performance of the combined institutions. The net effect across all institutions is no significant gain in cost performance. In addition, these studies find that cost efficiency is a better explanator of financial institution profitability than market power, but together these two effects explain only a small proportion of variation in performance across institutions.

Recent studies distinguish between cost efficiency and profit efficiency. Cost efficiency improves when costs per unit of output fall for given output quantities and input prices. Profit efficiency incorporates cost efficiency but is more general because it also includes cases where profits increase in response to changes in the output mix. Akhavein, Berger and Humphrey (1997) show that some mergers between large banks in the US have led to improved profit efficiency. This occurred through diversification benefits. Other things being equal, diversification should reduce risk. However, the response of the merged bank has typically been to shift the production mix towards higher risk products, that is, away from securities towards loans (Berger 1998). As a consequence, profits increased but without a substantial reduction in the overall risk of the bank.

Most of the empirical studies of mergers and banking efficiency are based on foreign markets.[13] The nature of the Australian banking industry makes it difficult to apply the overseas evidence to the Australian situation. There are a number of points worth mentioning in this regard. First, the Australian market is increasingly becoming a national market (see below). The largest banks already have extensive geographic and product diversification, therefore, further consolidation will produce smaller diversification benefits than suggested by overseas studies. Second, on the other hand, this high degree of geographic diversification may imply greater duplication of branch networks, so that mergers in Australia could generate greater cost savings through branch closures. Third, many of these foreign markets remain relatively unconcentrated compared with the Australian market. Fourth, many studies conclude that substantial economies of scale exist, but only up to a relatively small size (Berger, Hunter and Timme 1993). While there is a wide variation in the exact size of this cut-off point, the largest Australian banks are clearly above this point.[14]

There are two additional arguments to consider with regards to the impact of consolidation on efficiency. These are recent trends which may imply greater contestability of the market, and the implications of new technology. The potential for both of these developments to increase efficiency may not yet have been captured by existing studies.

4.1.1 Contestability and the competitive fringe

To determine the degree of contestability, it is necessary to first define the extent of the market. It is then possible to examine the ease with which either existing smaller suppliers, and/or new entrants can provide effective competition to the large Australian banks. A strong competitive fringe, or even the potential for this fringe to develop, may ensure competitive behaviour in a market dominated by a few large suppliers.

The first distinction to be made is the difference between the retail and wholesale markets. There are a number of reasons to believe that the Australian banking wholesale market is strongly contestable. First, the market has a sizeable competitive fringe of both domestic and foreign banks and non-banks providing a wide range of wholesale products and services. Second, the wholesale market is a national one and does not require an extensive branch network in order to conduct business. Finally, there are few barriers to entry, particularly for larger foreign-owned banks experienced in the provision and development of new and sophisticated products.

The degree of contestability in the Australian retail banking market is arguably lower than in the market for wholesale banking. However, this depends in part on the precise definition of the retail market in terms of the degree to which retail banking products are viewed by customers as being ‘bundled’, and the extent of geographic boundaries.

The Australian Competition and Consumer Commission (ACCC) has defined the product dimension of the market to be ‘the cluster or bundle of services provided by banks to their retail customers’ (ACCC 1996, p. 15). The argument was that there are no close substitutes for a cluster of services which include loans, deposits and payments. While there is undoubtedly some convenience value to bundled services, there is evidence that a sizeable proportion of consumers use unbundled banking products (Wallis 1997). Furthermore, although demand for transaction accounts is relatively insensitive to price (Wallis 1997, p. 437), this is not the case for home loans where price differentials have more substantial wealth effects (Wallis 1997, p. 438). The success of mortgage managers and other providers of non-bundled products (including cash management trusts, credit card services, etc) suggests that fringe providers are able to apply competitive pressure to banks. This is especially true when banks cross-subsidise or otherwise overprice certain products. Many of these developments are the result of technological innovations which may not yet have been fully reflected in existing studies of the competitive effects of consolidation. Increased contestability across many retail product lines is likely to remain a continuing trend for some time.

In 1995, the ACCC also determined that the relevant market for retail banking was state based, rather than national. More recent developments suggest that for many products the relevant market is becoming increasingly national. Improvements in technology have reduced the cost of data transmission, and hence, the cost of delivering many financial products. In addition, there is no reason why previously state-based banks cannot establish branch networks in other states.[15]

4.1.2 New technologies

Recent advances in technology may imply that there are economies of scale even for larger banks, thereby creating pressures for further consolidation. However, if these advances have not yet been fully exploited, they may not have shown up in existing studies of cost efficiencies. While it may be true that the most recent technological advances imply cost efficiencies from scale, it is also possible that technological advances in the future may work in the opposite direction. To demonstrate this point we draw parallels between developments in technology and market structure in the banking and steel-making industries. The steel-making industry has already progressed through three distinct phases of technological innovations – the latest phase is helping to reverse an earlier trend towards consolidation.

The mass production of steel began with small decentralised production facilities located near to where inputs were mined (Ashton 1969). This was necessary because of the high costs of transporting these inputs. In a similar fashion, until recently, back-office operations in banking were located in individual branches because of the high cost of communicating and storing data.

Falling transportation costs and advances in production technology led the steel industry to move towards large centralised production facilities that were able to take advantage of economies of scale. Similar advances in communication and computing technologies have recently allowed many back-office operations in banking to be undertaken centrally in order to take advantage of economies of scale.[16]

It is not clear that further consolidation in banking is necessary to take advantage of these economies of scale. It may be feasible in the future for banks to contract out some of these services to a single large provider that can take full advantage of economies of scale.[17] However, there are at least two constraints on this practice becoming widespread: the issue of efficient access (including pricing) which might be difficult to establish; and the problem of proprietary rights to information gained by the firm running such a system.

More recently in the steel industry, new technology (embodied in mini-mills) has allowed a substantial portion of production to move back to smaller decentralised facilities (Barnett and Crandall 1986). These facilities can benefit from being closer to customers and more responsive to their requirements. Similarly, in the financial sector the fixed costs of providing risk-management services may have fallen considerably over the previous decade.[18] In particular, many products, which even two decades ago were difficult to price, have now become standardised,[19] and there appears to be a greater availability of highly trained personnel in the field of finance. A fall in the fixed costs of head-office risk-management operations would make it easier for smaller banks to enter the market for these services. Whether this type of change eventuates in the case of back-office operations in banking is purely speculative. However, given the trend of rapid advances in computing technology, this prospect is not implausible.

So far in this section of the paper, we have focused most of our attention on the impact of consolidation in terms of the domestic banking market without considering the international market for banking services. However, there is an argument that consolidation domestically is necessary in order for banks to become large enough to compete successfully in the global market for financial services. In part, this reflects the need for substantial capital investment which presumably will keep banks at the forefront of international best practice in terms of the optimal use of computing and communications technology and the development of sophisticated financial products. Whatever the merits of this argument, it still needs to be weighed against the potential costs of increased domestic concentration. It is also worth noting that there is scope in global markets for niche players to provide specialised products, and more generally for smaller players to take advantage of their ability to maintain closer relationships with their customers.

4.2 A Framework for the Analysis of System Stability and Efficiency

Having discussed the impact of consolidation on efficiency, we digress in order to introduce a general model which outlines the policy-maker's decision process regarding financial system efficiency and stability. The model also provides a formal definition of system stability. Many aspects of the model are clarified later in Section 4.3 when we apply it to the question of how consolidation may affect system stability.

The essence of the model is to outline the objectives of the policy-maker, to emphasise the role of expectations, and to identify in broad terms the relevance of the policy instruments. In our model, the policy-maker cares about two things: the macroeconomic losses that could originate from disturbances to the financial system, and the efficiency of the financial system. That is, the policy-maker's objective is to enhance financial system stability and financial system efficiency – recognising that in some cases there may be a trade-off between the two.

For simplicity, we assume that financial disturbances are associated with the failure of financial institutions, and that the macroeconomic losses are a function of the number of institutions that fail in a particular episode. We assume that the policy-maker faces an uncertain world, but knows the macroeconomic costs that could arise given various financial disturbances. The policy-maker's task is to choose a set of policies that maximise utility subject to a set of constraints about how the financial system works. The set of policies might include: the maximum degree of market concentration; conditions of entry; or the terms and conditions for central bank liquidity support.

While intentionally simple, the model draws out a number of issues, including the impact that consolidation and conglomeration might have on system stability, the relevance of contagion and the implications of central bank support for institutions experiencing difficulties.[20]

More formally, the problem of the policy-maker is to choose a course of action x so as to maximise expected utility subject to a collection of constraints:

The constraints (which we do not spell out here) describe the trade-off between stability and efficiency.

The policy-maker's utility function u(L, E) depends negatively on the macroeconomic loss L, and positively on the efficiency of the system E. For simplicity we assume that the measure of efficiency E is independent of the state of the world i.

The measure of loss L represents the lost output that follows from the failure of some proportion Pi of the financial system and hence, a reduction in the extent of intermediation. The greater the proportion of the financial system which fails, the greater will be the loss L.

The state of the world, i, can be fully described by the proportion of institutions that have failed, Pi. The policy-maker is assumed to know the impact of their actions on the probability density function, fx(i), which is defined over the states of the world i. The state of the world i is revealed after the policy-maker has determined a course of action.

We assume that the relationship between the proportion of the system that fails and the macroeconomic loss L is represented by the function L(Pi), where L′ (∘) ≥ 0 and L″ (∘) > 0. In words, the complete failure of the system results in a loss which is more than ten times greater than the loss from the failure of one-tenth of the system.

The assumption regarding the shape of L(Pi) follows from a reasonable assumption about the macroeconomic consequences of a reduction in financial intermediation and the costs of resolving financial crises.[21] Provided that contagion is contained, smaller failures are relatively easy to resolve rapidly without substantial disruption to the process of intermediation. It is assumed that when a relatively small number of financial firms fail, the remaining assets of these firms can be sold, or the firms can be restructured and sold rapidly because they represent only a small proportion of the market. Losses to depositors and creditors may be relatively minor and meanwhile the process of intermediation in the healthy portion of the system continues largely unaffected. However, in the case of the failure of a substantial proportion of the system, resolution becomes problematic and disruption of intermediation becomes extreme. Invariably, governments cannot sell the assets of the failed institutions (particularly large failed institutions) without severe consequences for the asset values of the healthy institutions. Restructuring failed institutions may require the government to cover a large amount of non-performing loans. Even after restructuring (and perhaps a break up of larger failed institutions) it may be difficult to find buyers willing to pay a reasonable price during the crisis for such a sizeable portion of the financial system.

The range of policy options x that policy-makers have at their disposal include the maximum level of concentration permitted in the financial system, the ease of entry of new firms, the extent of conglomeration permitted, and the terms and conditions for the provision of central bank liquidity or lender-of-last-resort loans. The policy action can work through a number of channels: it may affect the probability density function fx(i); it may also influence the loss function L(Pi) – for example, the government may provide some form of support to failed institutions. Policy may also influence the trade-offs implicit in the constraints.

4.2.1 An index of financial system instability

If we assume that the policy-maker is risk neutral with respect to the macroeconomic loss, then we can define an index of financial system instability (for a constant level of efficiency) to be:

The index S is the expected macroeconomic loss that results from financial system disturbances – low values of S indicate stability. This is consistent with our earlier notion that system stability describes both the probability and size of incidents of financial stress in terms of the impact on the real economy. Although risk neutrality on the part of the policy-maker is unlikely to be true in practice, this assumption is mostly one of convenience. Otherwise, the policy-maker may not be indifferent between two outcomes with the same levels of S if they are based on different variances of macroeconomic loss. Increasing the degree of convexity of the loss function L(Pi) would produce similar results to a model which incorporated risk aversion explicitly.

4.3 Consolidation and System Stability

The influence of consolidation on system stability remains an open question, and to date has been largely unanswered by existing empirical and theoretical studies (Boyd and Graham 1998). In this section we use our framework to examine a number of stylised examples relevant to the impact of consolidation on system stability.

For expositional purposes we present a simple version of our more general framework described in Section 4.2. We assume that the financial system initially consists of a number, n > 1, of equally sized banks. For reasons outside of the model, there is pressure for these banks to merge to form m banks of equal size (where 1 ≤ m < n). The policy-maker must determine whether or not to allow this consolidation to proceed – that is, their policy action can be described as x = n for no consolidation, or x = m for consolidation.

To focus our attention on system stability we assume initially that consolidation is neutral with respect to efficiency. Though efficiency is still a crucial consideration, we have already discussed the broad implications of consolidation for efficiency in Section 4.1. We also assume that the policy-maker is risk neutral with regard to the macroeconomic loss. These two assumptions simplify the problem to one of determining whether consolidation leads to an increase in system stability.

In the case where n = 2 and m =1, there are three states of the world: either no banks fail (P1 = 0); half the banks fail (P2 = 0.5); or all banks fail (P3 = 1). Of course in a system with only one bank, the probability that half the banks will fail is zero.

We assume the loss function is quadratic:

The probability density function for states of the world can be determined from the probability of individual bank failure, which is pj when there are j banks in the system (j = n,m). We simplify the analysis by assuming pn = pm = p, that is, consolidation does not, by itself, alter the unconditional probability of a single failure. This assumption is appropriate if the merger between two banks provides little scope for reduced risk through greater diversification.

The crucial question, however, is whether a bank failure is independent of other bank failures. This will depend on the nature of shocks that cause bank failures. We consider three cases: a common shock which implies complete dependence; an idiosyncratic shock which implies independence; and an intermediate case which arises because of contagion.

(i) Common (macroeconomic) shocks

Given the assumptions that we have made, if a bank failure is caused by a common shock (to the macroeconomy for example), then consolidation will have no impact on system stability. The instability indices, Sn and Sm, are both equal to the probability of individual bank failure p. In other words, if one bank fails, it indicates a bad macroeconomic shock, and all banks will fail.

(ii) Idiosyncratic (management) shocks

On the other hand, bank failure may be entirely due to idiosyncratic shocks. A common element of many cases of bank distress and failure is poor management and operational procedures.[22] Invariably during times of poor macroeconomic performance, some banks experience substantial losses while other banks – doing business in essentially the same market and under the same conditions – survive these periods relatively unscathed. Therefore, variation of management across banks can help to explain some of the variation in banking performance.

Almost by definition, consolidation will lead to a reduction in the degree of managerial diversification of the banking system. Such a reduction in diversification may be a good thing for both stability and efficiency if consolidation occurs through the acquisition of poorly managed banks by well-managed banks. However, if higher market concentration implies a reduction in competition, then managers may find it easier to reduce their efforts and the efficiency and the stability of the system may suffer.[23]

If all managers have equal ability and the probability of one bank failing is independent of the performance of other banks, then consolidation reduces the stability of the financial system. In a system with fewer banks we expect to see fewer bank failures, however, if these banks do fail there is a much larger macroeconomic loss because the banks are bigger.[24] This general result is demonstrated using our model for the case of n = 2 and m = 1 in Table 3. Consolidation increases our index of system instability from Sn = 0.5(p + p2) to Sm = p.

Table 3: Indices of Instability
  Proportion of
system failure
No consolidation Consolidation
State 1 0 0 (1 − p)2 1 − p
State 2 0.5 0.25 2p(1 − p) 0
State 3 1 1 p2 p
Index of instability, S Sn = 0.5(p + p2) Sm = p

In addition to consolidation leading to less management diversification, highly concentrated markets may make it more difficult for agents to monitor the performance of bank managers.[25] By reducing incentives for management to behave appropriately, a reduction in the ability to monitor management will reduce both the efficiency and stability of the banking sector. If banking performance depends on a common macroeconomic component and an idiosyncratic management component, which are only observed indirectly through their combined impact on banking performance, it will be impossible to perfectly observe management ability (or effort). However, the inference of management ability from a bank's performance will improve with the number of banks in the system. In a market with a sufficient number of banks, the law of large numbers implies that management ability will be reflected in the deviation of a bank's performance from the industry average. In contrast, in a market with only a few banks, bad luck and poor management may be more difficult to distinguish.

(iii) Contagion

Contagion can, in principle, lead to the failure of otherwise healthy financial institutions. This possibility can be incorporated into our framework as follows. Suppose that the state of the world is revealed in two stages. During the first stage, banks fail on their own account with probability pn and pm in the non-consolidated and consolidated cases respectively. In the second stage, conditional on the initial failure of at least one bank, other banks can fail as a result of contagion with probability qn and qm .[26]

Once again we assume that shocks that lead to failure in the first stage are independent (that is, due to idiosyncratic management behaviour), and that pn = pm = p and qn = qm = q. In the presence of contagion, it is no longer the case that consolidation always leads to greater system instability. This is easily demonstrated in the case of a system of only two banks which merge into a single institution. (Note that in the consolidated system there is no contagion because there is only one bank.) The system instability indices, Sn and Sm in Table 4 show that consolidation leads to greater instability only if the likelihood of contagion is not too great (that is, q < ⅓).

Table 4: Indices of Instability – with Contagion
  Proportion of
system failure
No consolidation Consolidation
State 1 0 0 (1 − p)2 1 − p
State 2 0.5 0.25 2p(1 − p)(1 − q) 0
State 3 1 1 p2+2p(1 − p)q p
Index of instability, S Sn = 0.5p + 0.5p(p + 3q(1 − p)) Sm = p

The intuition for this result is that for a given sized financial system, it is preferable to have many smaller banks if shocks are idiosyncratic. This ‘management diversity’ leads to greater system stability. However, with more banks, there is a greater chance of at least one bank failing, and therefore of contagion, which reduces system stability. The policy-maker can attempt to address the problem of contagion directly by providing some form of liquidity support to solvent banks, thereby helping to prevent runs on otherwise healthy banks. In our model this would be captured by a reduction in the probability of failure due to contagion, qn and qm. However, governments may face incentives to provide preferential support to larger banks.

The perception that banks might become too big to fail is closely related to the issue of contagion. There are three different effects to consider in this regard. First, there is the potential for greater contagion following the collapse of a larger bank. That is, qm > qn, which implies that consolidation lowers system stability. Second, governments recognising this effect may attempt to offset it by taking measures to avoid the failure of a large bank. The public perception of a large bank being too big to fail will tend to prevent runs on large banks following the collapse of some other bank – this will tend to increase system stability. Third, if the public perceive large banks as being too big to fail, then depositors and creditors face a reduced incentive to actively monitor and discipline very large banks so as to ensure both efficient and prudent behaviour. This can allow managers to take greater risks, which in our model represents an increase in pm – implying lower system stability. The net effect of these three opposing forces on system stability is unclear.

In summary we draw a number of broad conclusions from our analysis of the impact of consolidation on system stability.

  • There are circumstances where consolidation may reduce system stability. This is the case if a substantial proportion of shocks which lead to bank failure are idiosyncratic, if the probability of a bank failing of its own accord is not affected by size, and if there is little chance of contagion. In this case, consolidation has two opposing effects. A more consolidated system has fewer banks, therefore we expect to see fewer bank failures. However, in a more consolidated system the banks are bigger, so that a single bank failure has a much greater impact. If doubling the proportion of the system that fails more than doubles the macroeconomic loss, then this latter ‘size’ effect dominates and consolidation reduces system stability.
  • Contagion can introduce a third effect which may imply that consolidation increases system stability. Contagion describes the circumstance where the failure of any one bank (on its own accord) leads to the possibility of the failure of otherwise healthy banks. Consolidation reduces the number of banks, which reduces the possibility of at least one bank failing, and therefore, reduces the possibility of otherwise healthy banks failing from contagion. If contagion is sufficiently strong, system stability can increase with fewer banks in the system. On the other hand, if the failure of a larger bank is more likely to cause contagion (than the failure of a small bank), then the earlier result is restored – that is, consolidation may reduce system stability.
  • If a merged financial institution is perceived to be too big to fail, the incentives of managers and depositors can be distorted, leading to increased risk taking which reduces the stability of the system. An excessively large institution may impose large costs on any attempt to bail it out in the event of difficulties. On the other hand, the perception that larger banks may receive greater support during crises implies that consolidation can reduce the probability of contagion, thereby increasing system stability.


For an earlier discussion see Berger and Hannan (1989). [10]

In addition, inadequately supervised managers may choose to use market power to provide benefits for themselves and other employees while not necessarily increasing profits (Berger and Humphrey 1997). [11]

For a recent discussion of the ES hypothesis see Goldberg and Rai (1996). [12]

Walker (1998) is one of the few studies of economies of scale in the Australian banking industry. He concludes that significant economies of scale exist. However, his study is based on a very small sample of 12 banks – many of which are (or were) government-owned – over a period which straddles the significant deregulation of the mid 1980s. [13]

The smallest of the four largest Australian banks has domestic assets of about US$50 billion. The cut-off in terms of assets is estimated to range up to US$500 million. [14]

There is evidence that this has already occurred for some banks. For example, the proportion of branches of the State Bank of NSW (now the Colonial State Bank) outside of its home state rose from 3 per cent in 1990 to 16 per cent in 1997. [15]

For example, many back-office operations of Australian banks operating in New Zealand are being gradually shifted to Australia. [16]

There is some evidence that this is already occurring in Australia. For example, Westpac's loan-processing facilities currently provide capacity to at least one other institution. Also, the ANZ Banking Group has recently outsourced many of its electronic card operations (Australian Financial Review, 15th September 1998, p. 33). [17]

Risk-management services are an important part of wholesale banking. Intermediaries (especially banks) are the principal participants in markets for financial futures and options rather than individuals or firms (Allen and Santomero 1997). [18]

Developments in finance theory by Black, Scholes and Merton were instrumental in this regard. For a recent discussion of these developments see Schaefer (1998). [19]

One interesting avenue of exploration (which is beyond the scope of this paper) is the problem of dynamics – that is, the potential for entry and exit of firms, and the way in which the structure of the financial system evolves in response to significant crises. [20]

This assumption is not essential – the same qualitative results apply if the function L(Pi) is linear, so long as preferences are such that the policy-maker is risk averse in terms of economic losses. [21]

Dziobek and Pazarbasioglu (1997) found that management deficiencies caused problems in all cases of banking crises they studied, and that correcting these deficiencies was crucial for successful reform. [22]

Management discipline could be maintained in a more concentrated market through the threat of removal by the owners. However, given that distress is often used as a signal of poor management, this seems like an inefficient way of ensuring high management effort and management diversification across the system. [23]

The key to this result is the convexity of the loss function, L(Pi ). If this function is linear (for example, if the loss is proportional to the share of the system that fails), then consolidation will not alter system stability. [24]

Tevlin (1996) discusses issues related to management performance, monitoring and incentives. [25]

In a more complex model, the likelihood of failure due to contagion, q, could be made an increasing function of the proportion of the banking system that fails on its own account. [26]