Cost Efficiency of the Ghanaian Banking Sector: The Post-Liberalisation Experience

Using data from 1997 to 2008, this paper investigates the cost efficiency of the Ghanaian banking sector after financial liberalisation. The Ghanaian bank with the highest efficiency score is found to be operating at maximum possible efficiency. The average bank is however operating at high costs. Despite mixed evidence in the literature, there is the received wisdom that most cross-border mergers and acquisitions post-liberalisation result in failure due to factors that include poor credit quality, inadequate generation of fee income, and poor customer mix. In Ghana, the situation is different because the only foreign-acquired bank had prior knowledge of the local conditions and has managed to utilise this advantage, coupled with redundancy programmes and layoffs and without branch expansion following the acquisition, to operate at a relatively high level of cost efficiency.

be set by the market, and have permitted banks to lend on commercial considerations alone. It is expected that these practices would lead to greater competition amongst banks whose objective, in a liberalised economy, is to maximise profit. Proponents of liberalisation suggest that the banking system will become more efficient through competition, but the evidence is mixed (Note 1). There are different ways of measuring efficiency, and the measures are not necessarily correlated. For example, an examination of Korean data (Cho, 1988) suggests that, following liberalisation, allocative efficiency was improved. This implies that the rate of return from investment in different sectors to which the banks continued lending following liberalisation were equalised. This indicates a better allocation of loans to maximise profit. Amsden and Euh (1993) point out that Korea has achieved its objectives of modernising its financial sector, not because it relied exclusively on market forces to achieve desired goals, but rather by principally creating institutions or modelling old ones. The authors assert that the functional efficiency of Korean banks may be considered the most important performance-effect that the first round liberalisation achieved, and that, this type of efficiency has tended to be less heavily emphasised than allocative efficiency by advocates of financial liberalisation. Considering the high operating cost of Ghanaian banks, this argument is developed examining how functioning institutions have affected the efficiency of the Ghanaian banking system. More specifically, it is aimed to investigate how the market structure and regulatory framework under which the banks operate affect their efficiency.
Of particular concern is the received wisdom in the literature that most mergers and acquisitions post-liberalisation result in failure. Whereas some of the failures have been attributed to factors that include poor credit quality and inadequate generation of fee income, resulting in additional costs generated by the merger (Knapp et al., 2005), others have been attributed to poor customer mix post acquisition (Havrylchyk, 2006). The experience is however different in the Ghanaian context because the only foreign-acquired bank had prior knowledge of the local conditions and has managed to utilise this advantage, coupled with redundancy programmes and layoffs and without branch expansion following the acquisition, to operate at a relatively high level of cost efficiency.
This study contributes to the established literature on the impact of financial liberalisation in Africa.
Studies that have investigated banking system efficiency in both developed and emerging economies have yielded mixed results, due to different macroeconomic and sectoral conditions prior to the financial reforms. Many of these studies have focused on periods when some of these reforms had not been fully implemented. The present study examines a period following the implementation of all of the phases of the liberalisation process in the Ghanaian banking industry. This should help critically assess the benefits and non-benefits of the reforms. The results presented in this paper are among the earliest indications of the effects that financial sector reforms have had on the banking system efficiency in sub-Saharan Africa.
The remainder of the paper is organised as follows. Section 2 reviews the relevant literature surrounding the research topic. The empirical model employed is described in section 3, alongside www.scholink.org/ojs/index.php/asir Applied Science and Innovative Research Vol. 3, No. 3, 2019 192 Published by SCHOLINK INC. description of data used in the study. Section 4 presents the results and discussion of the study, followed by the conclusion in section 5.

Financial Liberalisation and Bank Efficiency: The Literature
The efficiency literature has grown rapidly over the past few decades. The vast majority of the studies covers the US and European countries (See, for example, Berger & Humphrey, 1997;Goddard et al., 2007;Asaftei, 2008;Beijnen & Bolt, 2009).
Efficiency studies on developing economies are less plentiful. The majority of the studies of emerging market banking systems after financial liberalisation refer to Asian countries including India (Das & Ghosh, 2006), Singapore (Rezvanian & Mehdian, 2002), Hong Kong (Drake et al., 2006), Malaysia (Okuda et al., 2002) and China (Berger et al., 2009). Cross-country studies of Asian banking system efficiency have also been undertaken (Williams & Nguyen, 2005).
Efficiency studies have also been reported for a number of countries in Central and Eastern Europe, as well as the Commonwealth of Independent States (Bonin et al., 2005;Koutsomanoli-Filippaki et al., 2009). In sub-Saharan Africa, bank efficiency studies after financial liberalisation cover economies including Uganda (Hauner & Peiris, 2005) and Botswana (Moffat & Valadkhani, 2008).
It is expected that financial sector reforms should increase competition, leading to improvements in bank efficiency. The empirical evidence has however been mixed. In a study of the Ugandan banking system between 1999 and 2004, Hauner and Peiris (2005) find that the level of competition has increased significantly and has been associated with an improvement in efficiency following liberalisation. Ataullah and Le (2006) report an improvement in the efficiency of Indian banks, especially foreign-owned banks, after financial reforms. Havrylchyk (2006) however finds that efficiency has not improved in the Polish banking industry between 1997 and 2001.
The comparison of efficiency between foreign-and domestic-owned banks suggests that foreign banks in emerging market economies have been able to utilise their advantages by achieving higher efficiency than their domestic peers (Hasan & Marton, 2003;Bonin et al., 2005). Hauner and Peiris (2005) show that on average foreign-owned banks are more efficient than domestic banks in Uganda. Moffat and Valadkhani (2008) find that foreign institutions exhibit higher levels of efficiency than public institutions in Botswana. In the case of the Polish banking industry, Havrylchyk (2006) finds that greenfield banks have been able to achieve higher levels of efficiency than their domestic peers, whilst foreign banks that acquired domestic firms had not succeeded in improving their efficiency.
Empirical studies on the relationship between bank size and efficiency have also produced mixed results of the impact of financial liberalisation on bank efficiency. Hauner and Peiris (2005) find larger banks to be more efficient than smaller banks in the Ugandan banking system. Moffat and Valadkhani (2008) show that small and large institutions have higher levels of efficiency than medium-sized banks in Botswana.

The Cost Efficiency Model
Under a liberalised financial system, banks are expected to realise efficiency gains via reorganisation with "prices reducing towards production costs under the pressure of competition" (Gardener et al., 2001). Operating cost (OC) is used as dependent variable in the efficiency model and this variable is proxied as the sum of personnel and other non-interest expenses.
The intermediation approach to modelling bank production is employed. This approach assumes banks to perform an intermediation role between depositors and borrowers. From this perspective, banks are considered to be intermediators of financial services that purchase inputs in order to generate earning assets as output (Sealey & Lindley, 1977).
More precisely, the value-added approach to the intermediation process is used, and estimate efficiency from the cost function by specifying two outputs (loans and deposits) and two inputs (price of labour and price of deposits). These categories of loans (LN) and deposits (DP) are considered as the key outputs, because they generate the great majority of value added. The price of labour is proxied as the ratio of personnel expenses to total assets, whereas the price of deposits is proxied as the ratio of interest expenses to the value of total deposits. These two variables enter the regression as the ratio of labour cost to the price of deposits (w). The dependent variable (OC) also enters the regression divided by the price of deposits.
A time trend variable (T) is included to account for the effects of technological change, together with other factors such as regulatory change. This captures the missing time dimension that is not explicitly modelled in the cost function.
Three bank-specific indicators that capture credit, capital and liquidity risk are also included to control for differences in the banks' risk profile, since measured efficiency may reflect variation in risk-taking strategies across banks.
Credit risk (CRD) is measured as the ratio of provisions for bad and doubtful debts to gross loans; capital risk (CAP) as the ratio of shareholders' funds to total assets; and liquidity risk (LIQ) as the ratio of liquid assets to total bank liabilities.
On the basis of the aforementioned variable definitions, the preferred cost efficiency model is specified using a two-output, two-input translog functional form. The efficiency estimate is derived from a cost where for the i-th bank in the t-th time period, and is assumed to be independently but not identically distributed according to a truncated-normal distribution.

Correlates of Bank Inefficiency
After estimating cost efficiency measures for Ghanaian banks, one may investigate further the factors that are correlated with bank inefficiency. To do this, the conditional mean model of Battese and Coelli (1995) is employed, which allows in a one-step procedure estimation of the cost function and identification of the correlates of bank inefficiency. Inefficiency scores are first derived from the model, and then, express these scores as an explicit function of a vector of predictor variables which in this case measure characteristics of functioning institutions.
The institutional variables employed to investigate the factors that are correlated with bank inefficiency include bank governance, competition, and bank size. These are discussed below.

Bank Governance
The issue of bank governance has been of concern to researchers since financial liberalisation was introduced in a number of reforming economies (Note 2). This has resulted in a proliferation of efficiency studies on bank ownership in recent decades. Causes of changes in ownership and other organisational restructuring include both domestic and foreign mergers and acquisitions, privatisation, restructuring of distressed banks, and bank closures.
The approach to measuring bank governance has varied with respect to the economies concerned, managerial structure and data availability. According to Berger et al. (2005), it is important to control for all of the major governance changes that affect the performance of an economy's banking sector in order to avoid misspecification, bias and misleading results. For this reason, several authors have developed a framework that captures the static, selection, and dynamic effects of changes in bank governance on bank performance. Governance changes are usually specified using dummy variables.
Due to data constraints, the method applied in this study only includes three static variables for foreign-owned, private domestic-owned and state-owned banks; one selection variable for cross-border merger and acquisition; and two dynamic variables for the selection dummy that include both shortand long-term performance effects for the bank concerned (see below). It is important to note that all the state-owned banks in Ghana underwent some form of restructuring during the reform period, and this included recapitalisation from public funds, the removal from balance sheets of non-performing assets, and the appointments of new boards of directors and senior executives. As mentioned earlier, the limited data sample for this study does not permit to investigate the selection and dynamic effects of the state-owned banks from the period these changes occurred.
Hence, these changes are treated as long-run static effects, because the government remained the majority equity stake holder in these banks during the study period.
Due to the aforementioned data constraints, I am unable to measure the selection and dynamic effects of foreign-and private domestic-owned banks from the period the government relinquished its part or entire minority shareholding in them. For this reason, I also treat these changes as long-run static effects in cases where foreign or private domestic owners retained majority shareholdings throughout the sample period. Many studies in the academic literature have applied these methods to investigate the performance effects of different types of bank ownership (see, for example, Fries & Taci, 2005;Lensink et al., 2008).
Following Nakane and Weintraub (2005) and Williams and Nguyen (2005), one exit variable is also included which identifies two state-owned banks that were liquidated. Unlike most ownership studies, this paper follows the approach of Fries and Taci (2005) by introducing two entry variables that capture newly established foreign and domestic banks that entered the market during and after the reforms.
Initially, I focus on the static effects of foreign-ownership (STATIC_FOR), private domestic-ownership (STATIC_DOM) and state-ownership (STATIC_STA) on the bank efficiency. Static dummy variables are created for these banks and assume that they have not undergone any major change in ownership composition over the sample period, and that they were still active at the end of the sample period.
These dummy variables take the value of one for the banks concerned for all time periods, and zero for other periods and for all other banks.
The literature suggests that the efficiency of foreign-owned banks in emerging economies differs from both private domestic-owned and state-owned banks. Foreign banks are assumed to possess superior management practices and technological advantage over local banks, and as such, are expected to capitalise on their advantages and exhibit higher efficiency levels than their local peers (Claessens et al., 2001). Following the literature on foreign banking, a bank is defined as foreign if more than 50% of its shares are owned by non-domestic residents. A static dummy variable is specified for foreign ownership (STATIC_FOR) that takes the value of one for all periods, and zero otherwise.
Contrary to the above discussion on foreign ownership, it has also been suggested that the opening of financial markets to foreign competition may increase the cost of domestic banks' operations. Stiglitz (1993)  State-owned banks are generally considered to be less efficient than privately-owned banks Nakane & Weintraub, 2005). In a case where government banks account for a substantial share of an entire banking market, non-commercial criteria may frequently be used to allocate credit, with resulting upward pressure on cost inefficiency. The argument behind the inefficiency of government banks is framed along the three alternative theories of state ownership: social, political, and agency (Sapienza, 2004). While the social view of state ownership assumes that state-owned enterprises (SOEs) are created to address market failure in financial and credit markets (Megginson, 2005), the political view assumes state-owned enterprises (SOEs) to be inefficient due to deliberate policies of politicians of diverting resources to their supporters (La Porta et al., 2002). The agency view on the other hand supports the idea of the social view that state-owned enterprises (SOEs) are created to maximise social welfare, but are subject to an inherent tendency to generate corruption and misallocation (Banerjee, 1997;Hart et al., 1997).
After major financial sector reforms, one would expect the inefficiencies in state-owned banks to decline through the strengthening of operational procedures and the existence of improved supervisory and regulatory systems. A bank is defined as state-owned if more than 50% of its shares are held by the government. A static dummy variable is specified for state ownership (STATE_STA) that takes the value of one for all periods, and zero otherwise.
During a period of liberalisation, it is common to observe banks withdrawing from the market by means of liquidation or through the change of activities from either a specialised operation into something else, or transfer of their assets and liabilities to other banks through merger and acquisition.
In Ghana, due to insolvency, a small number of state-owned banks were liquidated during the reform period. For this reason, exit dummy variables are defined for liquidated state-owned banks (EXIT), and specify these dummies to take the value of one for the closed banks during all the periods for which they are present in the sample, and zero otherwise.
As governments remove controls on entry, it is common for new foreign and private domestic banks to enter the industry. The Ghanaian banking system is seen to be expanding in numbers as new banks find their way into the reforming economy. Entry dummy variables are measured for both newly established foreign (ENTRY_FOR) and private domestic (ENTRY_DOM) banks, and specify separately these dummies to take the value of one for all the periods in which they are present in the sample, and zero otherwise.
Furthermore, mergers and acquisitions involving banks have been common in developing countries during and after periods of financial reform. Due to data constraints, I am unable to investigate the only domestic merger and acquisition deal that took place during the reform period. However, the same bank ownership. This is the only cross-border merger and acquisition recorded in the Ghanaian banking industry since the reforms took place, and the data sample permits investigation of its inefficiency effects. Selection and dynamic dummy variables are created for the merged and acquired bank.
The selection dummy variables are created for banks that have been involved in some form of ownership change over the sample period. For this reason, a dummy variable (SELECT_FOR) is specified to take the value of one for all periods of a bank that underwent cross-border merger and acquisition, and zero otherwise.
The dynamic dummy variables on the other hand are created for those banks for which the selection dummies take the value of one to date the exact moment when the ownership change took place, and zero otherwise. In other words, the dynamic dummy variables (DYNAM)_ST) take the value of one for the bank concerned for all the time periods following a given intervention, and zero for the periods prior to the intervention and for all periods for a bank that has not undergone any ownership change.
This treatment is assumed to identify the short-term performance effect of the intervention.
The dynamic dummy variable is assumed to capture the once-and-for-all changes associated to a certain intervention. In addition to this level effect, however, Nakane and Weintraub (2005)  However, all observations in the year when the intervention took place are excluded from the sample.
Thus, the dynamic time dummy variable, which reflects a long-term performance effect, starts with zero prior to the intervention and two for the second year following the ownership change. The intuition behind this treatment as noted by Nakane and Weintraub (2005) is to control for noise and some of the short-term transaction costs associated with the intervention. This may include discontinuities in previous policies, adoption of new strategies, as well as costs that are related to legal issues, consultancy services, due diligence, and any costs that may be associated with the corporate change. Each dummy equals zero for all periods for banks that did not experience any cross-border merger and acquisition.
Based on the aforementioned definitions for bank ownership and corporate changes, it is expected that the various measures will explain the variations in the cost-efficient frontier. This will allow assessing the effect of institutional differences on the efficiency estimate. Eventually, one is able to examine which of these ownership effects dominates the efficiency of the Ghanaian banking system after financial liberalisation.

Competition
Proponents of financial liberalisation suggest that as a consequence the banking system will become (UNIVERSAL) is set to take a value of one from the period when universal banking was introduced, and zero otherwise (Note 4).

Bank Size
Variation in the location of the cost-efficient frontier is allowed by bank size. Banks with larger asset holdings may operate more efficiently than their counterparts due to the use of different production technology. The evidence in the academic literature is mixed and inconclusive on this issue, as previous empirical studies have yielded different results.
It is also possible that banks of different sizes serve different groups of customers, and as a result, may face different levels of competition. From this perspective, the size indicator is expected to explain the efficiency of banks. Bank size (LOGASSETS) is measured as the natural logarithm of total assets.

Other Control Variables
As The following regression model is estimated for the determinants of cost efficiency in the Ghanaian banking system:

Data Sources and Classification
The data for this study were sourced from Bank of Ghana (BOG), which publishes balance sheet and income statement data for various Ghanaian banks. The dataset is an unbalanced panel that covers 28 banks over the period 1997-2008 with a sample size of 222 observations. The unbalanced panel dataset varies from 14 banks in 1999 to 25 banks in 2008. The study period has been chosen to reflect the post-liberalisation phase of the financial sector reforms period (Tables 1 and 2 report the summary statistics of the variables used in the empirical analysis).

The macroeconomic variables (discount rates and inflation) are obtained from the IMF International
Financial Statistics and World Economic Outlook databases.  Two state-owned banks were closed by means of liquidation.

Empirical Results
Cost efficiency frontier and bank inefficiency correlates were estimated on a panel of unbalanced sample. The results show a number of relevant implications of the cost function of Ghanaian banks and their correlates with inefficiency. Tables 3 to 5 report the results of the estimates. Robustness analysis are also conducted on the main findings, and the results are unaffected.

Average Cost Efficiencies
Tables 3 and 4 report the average measured bank efficiency scores by year at the sample means for banks classified by size, and by ownership status, respectively. The efficiency estimates take a maximum value of 1, indicating the best practice bank in the sample; and a minimum value of 0, corresponding to the most inefficient bank.
The results in Table 3  medium banks appear to be more cost efficient than small banks. Overall, the results suggest a mean efficiency score of 0.7553 for the full sample. This suggests that, when evaluated at the sample mean data point, the average bank is 75.53% efficient, or equivalently incurs 24.47% higher costs than the best practice bank facing the same conditions. On economic grounds, the mean cost inefficiency level suggests that, given their output level and mix, the average Ghanaian bank needs to reduce its production costs by 24.5% in order to use its inputs as efficiently as the most efficient Ghanaian bank.
This finding is interpreted as evidence of only limited gains in managerial efficiency having arisen from financial liberalisation.
The results in Table 4 are reported by governance indicators and year. Apart from the mean efficiency score for foreign banks, STATIC_FOR, which fell marginally from 91.85% in 1997 to 90.22% in 2008, all the other banks classified by governance indicators appear to have improved their mean efficiency scores over the same period. On average, the only bank selected for cross-border merger and acquisition, SELECT_FOR, seems to be more cost efficient than banks classified according to any of the governance indicators. This is followed by foreign banks, STATIC_FOR, and state-owned banks,

STATIC_STA. Both new and established private domestic banks, ENTRY_DOM and STATIC_DOM,
exhibit the lowest average cost efficiency measures. New foreign banks, ENTRY_FOR, also appear to be better at managing their costs than private domestic institutions.  Note. See Table 5.2 for variable definitions of governance indicators.

Bank Inefficiency Correlates
Employing the conditional mean model of Battese and Coelli (1995), factors that are correlated with bank inefficiency are investigated in the Ghanaian banking sector. In the model, (in)efficiency is derived from the cost function in equation (1) and subsequently expressed as a function of a vector of predictor variables. The results of this investigation, which include governance indicators, competition, and bank size, are presented in Table 5.

Bank Governance
The estimate in Table 5 shows that apart from the two dynamic variables (DYNAM_ST and DYNAM_LT), all the governance indicators have a statistically significant relationship with cost (in)efficiency. The omitted dummy variable among our governance indicators is bank exit, EXIT. Thus, the cost (in)efficiency with respect to bank governance is measured relative to this category. As expected, the results show that the two state-owned banks selected for closure were less cost efficient than other banks in the sample. The closure decision made by Bank of Ghana appears to be justified on cost efficiency grounds.  Note. The symbols *, ** and *** denote significance levels of 10%, 5% and 1%, respectively.
All the newly established private banks, ENTRY_DOM and ENTRY_FOR, that entered the market operate at relatively high levels of cost efficiency. The effect of each category also appears to be very similar in magnitude. The ability to pay high salaries arose from the fact that the newly established foreign banks had support from their parent companies. Concerns were raised as to how long this practice could be sustained. One would expect that the human capital investment made by the newly established foreign banks would eventually pay off, considering the fact that they had managed to bring into their teams a considerable number of staff and management with knowledge of local conditions in the banking industry. However, the unreasonable targets set by some newly established foreign banks forced a quite number of their workers to resign prematurely, as they were unable to meet their targets, and saw their salaries and incentives reduced accordingly. This added to the relatively high labour turnover in the banking industry during this period. Newly established foreign banks had to raise their game in order to compete with established banks. In spite of these efforts, the results of this study suggest they have not achieved the same levels of cost efficiency as their established foreign peers.
Among the group of static governance indicators, the result shows that established foreign banks, STATIC_FOR, are more cost efficient than state-owned, STATIC_STA, and established private domestic, STATIC_DOM, banks; with the latter showing the lowest average cost efficiency in the Ghanaian banking sector. This is partially consistent with theory which assumes that state-owned banks in developing countries are less cost efficient than their peers due to pervasive market inefficiencies and outmoded banking practices. In the Ghanaian context, our result shows that state-owned banks are more cost efficient than their established private domestic peers.
Despite the inefficiencies of government banks that are widely documented in developing economies, after the three phases of liberalisation in the Ghanaian banking sector, state-owned banks appear to have improved their efficiency levels according to the results reported in Table 5. There is also evidence that state-owned banks are more cost efficient than either foreign-owned, ENTRY_FOR, or private domestic, ENTRY_DOM, banks that have just entered the market. This suggests that a strengthening of operational procedures, and the existence of improved supervisory and regulatory systems, have helped to reduce the cost inefficiency of state-owned banks.
It is, however, noteworthy that the established foreign banks, STATIC_FOR, exhibit the highest average cost efficiency among the groups reported in Table 5. All of these banks were established during the colonial period; notably, Barclays and Standard Chartered. From this perspective, it appears that as a result of maintaining a long-term presence in the Ghanaian banking sector, these banks have managed to overcome the informational disadvantages they might have experienced with respect to local banks.
In geographical areas where they had consistently operated inefficiently, these banks have closed down their branches, despite adverse public opinion.
The higher cost efficiency of established foreign banks relative to domestic banks is also consistent with the theory that suggests foreign banks in developing economies have advantages over their domestic counterparts, such as superior management practices and technological processes (Claessens et al., 2001). As private domestic banks replicate and assimilate the modern banking technology and skills introduced by foreign banks, it is expected that this group of local banks would combine the modern practices with knowledge of local conditions they have to improve their efficiency levels.
The coefficient on banks selected for cross-border merger and acquisition, SELECT_FOR, is highly significant at the 1% level with a coefficient of 0.7865. This suggests that the only foreign-acquired bank has operated at a relatively high level of cost efficiency. This has been driven mainly by redundancy programmes and layoffs, and without branch networks expansion following the acquisition.
The experience in this case is therefore contrary to the received wisdom that most mergers and acquisitions result in failure (see, for example, Havrylchyk, 2006 In examining the dynamic effects of cross-border merger and acquisition, the results suggests that neither the short-term (DYNAM_ST) nor the long-term (DYNAM_LT) effects of governance changes has any statistical significant effect on measured cost (in)efficiency. This is a clear indication that the persistence of new policies and strategies adopted by the merged and acquired bank can only be fully realised after a passage of time.

Competition
As a measure of competition, the universal banking dummy variable, UNIVERSAL, though statistically insignificant is found to be positively related with measured cost (in)efficiency. As mentioned earlier, some advocates of universal banking argue that by despecialising development financial institutions (DFIs) and commercial banks, competition could be stimulated. The result of this study is found to be inconsistent with this argument, which may stem from the fact that before the Ghanaian authority replaced the 3-pillar banking model (development, merchant and commercial banking) with universal banking in 2003, banks were already engaged with universal banking activities to some extent before their licenses were officially changed. One can argue from this perspective that financial liberalisation has had only a limited effect on competition, because to a large extent banks continue to practice what they used to before the reforms.

Bank Size
The result shows that measured cost (in)efficiency is inversely related to bank size, LOGASSETS. The coefficient is significant at the 1% level. Though the established literature offers no consensus on this issue, there is evidence from this study that larger banks are significantly better at managing their costs than smaller institutions. This is consistent with the conventional wisdom which assumes that as banks grow larger, they are expected to control costs more effectively.

Conclusion
To examine the effects of financial liberalisation in Ghana, I empirically investigate the efficiency of its banking system from the period 1997 to 2008. Initially, I focus on the cost efficiency of banks to show the degree of progress that has been made in the banking sector after liberalisation. This is based on the expectation that greater cost efficiency should be associated with improved banking practices and better I therefore suggest that smaller banks, especially foreign-owned, could be encouraged to merge with established non-bank financial institutions that have adequate knowledge of local conditions. One such merger (UT Financial Services and BPI Bank) has been in progress at the time the writing of this paper was in progress. In the future, we expect the so-called "big giants", which have long-term experience in the Ghanaian banking industry, to face enhanced levels of competition. From a policy point of view, I conclude that even a liberalised financial system cannot be characterised by perfect competition, because of the need for risk pooling. At best, a liberalised system can achieve a form of oligopolistic competition that tolerates some degree of functional inefficiency. Thus, the cost of financial services might decline less than it would under perfect competition.