The efficiency of financial institutions

Published: Last Edited:

This essay has been submitted by a student. This is not an example of the work written by our professional essay writers.


The purpose of this essay is to reviews and summarizes the research to date on the efficiency of financial institutions. It analyses the contributions of the papers presented at the Atlanta conference, and suggest avenues of future research. Prior research on the topic of financial institution efficiency, surveys the contribution is this special issue, and suggests how future research on this important topic might proceed. The opinions expressed do not necessarily reflect those of the Board of Governors, the Reserve Banks, or their staffs. (U.S.A, for the 1990 and beyond).


Our contemporary world economy is structures of financial service industries that are changing rapidly, it is important to determine the cost and revenue efficiency of the evolving financial institutions. If the world institutions are becoming more efficient, then we might expect improved profitability, greater amounts of funds intermediated, better prices and service quality for consumers, and greater safety and soundness if some of the efficiency saving are applied towards improving capital buffers that absorb risk. Of course the converse applies if the evolution results in less efficient intermediaries, with the addition danger of taxpayer-financed industry bailouts if substantial losses are sustained.

It is clear that rapid changes in the economic and financial industry structure are occurring around the globe. In the US, the thrift industry has virtually collapsed, the insurance industry in under almost unprecedented financial pressure, and the banking industry is in the midst of a dramatic consolidation wave in which many of the nation's largest banking organizations are merging with one another. In Western Europe, there has been considerable consolidation of banks within countries in anticipation of EC integration and the accompanying consolidation across borders. In the Eastern Bloc, capitalist-style institutions that allocate financial resources on the basic of their prospects for financial success must almost be built from scratch out of the ruins of the Communist command economies. In Asia and South America, countries are restructuring regulations regarding the separation of commercial banking, underwriting, and insurance, in attempts to increase financial industry efficiency.

regrettably, the study of the efficiency of financial institutions has not kept pace with these changes. While scale and scope efficiencies have been extensively studied, primarily in the context of US financial institutions, relatively little attention has been extensively studied, primarily in the context of US financial institutions, relatively little attention has been paid to measuring what appears to be a much more important source of efficiency differences - X-inefficiencies, or deviations from the efficient frontier. That is, differences in managerial ability to control costs or maximize revenues appear to be greater than the cost effects of the choice of scale and scope of production. Research to date suggest that X-inefficiencies account for on the order of 20% or more of costs in banking, while scale and product mix inefficiencies, when they can be accurately estimated, are usually found to account for less than 5% of costs.

Even though a considerable amount of research has taken place on X-efficiency in general since its prologue in the introduction in the 1960s (see Leibenstein, 1966), published technical research on the X-efficiency of financial institutions has only appeared in the last few years. Moreover, prior to this special issue, nearly all such papers had measured X-efficiency for US commercial banks, with less than a handful of papers measuring the efficiency of nonblank financial institutions or banks outside of the US. Thus, in terms of both maturity and breadth, the efficiency research has not kept pace with the changes in the financial services industry.

The aim of the current research:

The purpose of this research and the opinion upon which it was based is to accelerate the growth rate of knowledge about the efficiency of financial institutions. The goals of this introductory article are to assess this progress and to suggest directions in which future research directions in which future research might be most fruitful.

The areas of research:

We subdivide the areas of this research into six categories:

  1. Scale and scope efficiencies in banking,
  2. efficiency in banking,
  3. The efficiency implications of bank mergers,
  4. The efficiency of thrifts and governmental financial institutions,
  5. The efficiency of the insurance industry, and
  6. The determinants of financial institution efficiency. For each category, we briefly discuss the prior state of knowledge, summarize the contributions of the articles in this special issue, and suggest directions for future research.

Scale and scope efficiencies in banking:

The prior literature on scale efficiency in banking suggests that the average cost curve has a relatively flat U-shape, with medium-sized firms being slightly more scale efficient than either very large or very small firms. (Humphrey 1990). The primary uncertainty expressed in this literature is the location of the bottom of the average cost U - the scale efficient point. Studies that used only banks with under $1 billion in assets, studies that used banks of all sizes, and one study that included all banks of over $100 million usually found average costs to be minimized between about $75 million and $300 million in assets. Studies that used only banks with over $1 billion in assets usually found the minimum average cost point to be between $2 billion and $10 billion in assets. These results suggest that the functional form employed in these studies may not be capable of incorporating the technologies of both large and small banks together in a single model, or that some important factor that varies with bank size may be excluded from the model.

Some show that the commonly used translog cost function specification gives a poor approximation when applied to banks of all sizes. (MC Allister and McManus 1993).The translog does not hold up as a reasonable gloal approximation because it forces large and small banks to lie on a symmetric U-shaped ray average cost curve and disallows other possibilities, such as an average cost curve that falls up to some output point and remains constant thereafter. Thus, it may be the case that the diseconomies found for larger banks are simply the imposed reflection of the economies found for small banks. In addition, the translog approximation may behave poorly away from the mean product mix, which can create problems in measuring scale efficiencies because large banks tend to have very different product mixes from the average. The solution to this problem is to replace the translog with one of several nonparametric estimation procedures.

There is a missing factor to the calculus of scale efficiency - risk, that as bank loan portfolios increase in size up to about $1 billion, the standard deviation of the rate of return fall precipitously, presumably because of diversification benefits. This reduction in risk lowers the amount of financial capital which must be held by the bank to keep the risk exposure of the bank's creditors (including the deposit insurer) at a given level.

Capital is the most expensive marginal source of funding; this creates a financial scale economy by which banks can lower their average costs of funds as scale increases by holding a smaller proportion of capital. This represents an improvement over tow previous attempts to incorporate risk into the cost function for financial institutions, one of which specified risk but did not include its cost and one of which measured risk by provisions for loan loss reserves, which reflect expected losses rather than the risk or variance of losses.

Another potential difficulty in the scale economy literature is that most studies do not use a frontier estimation method. Scales economics theoretically apply only to the efficient frontier, and the use of data from banks off the frontier could confound scale efficiencies with differences in X-efficiency. Fortunately, this potential problem does not appear to be of practical significance. Two previous studies that have compared scale efficiencies on and off the efficient frontier have found only small differences. The prior literature on scope efficiency for financial institutions is even more problematic that the scale literature. Major problems have been recognized. There is a problem in applying the translog specification to evaluate or test for scope economies. Computation of scope economies compares the predicted costs of producing a given bundle of outputs by two or more specializing firms versus joint production by a single firm.

The second recognized problem in estimating scope economies is that there are often little or no data on firms that specialize. In banking, virtually all firms produce the entire array of products specified in the cost function. In fact the dense part of the data set is usually well away form zero output, creating potentially significant problems of extrapolations. The effects of extrapolation, often combined with the problems of translog specification, can be quite dramatic - measured scope economies and diseconomies are often erratic and far exceed credible levels, at times over 1000 percent in absolute value. (Berger and Humphrey 1991, Pulley and Humphrey 1993, Mester 1993).

A way around this problem is to examine alternative measures that use product mixes that remain within the dense part of the data set, where more confident conclusions can be drawn. Expansion path subadditivity combines the scale and product mix effects of moving from each size class mean to the mean of the next largest size class, and provides what appears to be a more reasonable representation of the opportunity of existing banking firms to change their outputs than scope economies. These methods essentially sidestep the question of scope economies per se in favour of the potentially more interesting question of whether efficiency can be improved by changing scale and product mix simultaneously.

The third recognized problem in evaluating scope economies is that of using data that are not on the efficient frontier. As in the case of scale economies, scope economies are defined only on the efficient frontier, so that evaluation using data off of the frontier could confound scope economies with X-efficiencies. The empirical evidence suggest that, unlike the case of scale economies, this is a problem for estimating scope economies in banking.

More general 'Optimal scope economies' are based on the profit function instead of cost function instead of the cost function, and provide insights not available from the conventional scope economies concept. (Berger et al.1993). They redefine scope economies so that they are output-efficient as well as input-efficient. That is, the include all the revenue effects of output choices as well as the cost effects of output choices as well as the cost effects of input choices. Implementation of their new concept also provides at least partial solutions to all three of the recognized scope economy estimation problems.

This concept determines whether a firm facing a given set of prices and other exogenous factors should optimally produce the entire array of products of specialize in some of them. The optimizing choice is the output vector that maximizes profits over all possible output combinations for a firm that is fully X-efficient and scale efficient. They test whether the optimal quantity of every output is greater than zero for all the observed price vectors. If so, then optimal scope economies hold over the observed range of data. If not, then it may be profitable for some firms to specialize. They find that optimal scope economies prevail for most, but not all firms.

This concept contrasts with the contrasts with the conventional notion of scope economies, which simply addresses the question of how costs of producing a given output bundle may be minimized without determining whether than output bundle is optimal. Generally, the point of evaluation will incorporate output inefficiencies will be at a suboptimal scale and product mix, and therefore neither joint nor specialized production at that point is optimal.

Implementation of optimal scope economies address the three recognized problems in scope economy estimation: First, the problem of evaluating a translog or any other function at zero output is avoided - the profit function has as its arguments that output prices, which are always positive. Second, there should be little problem of extrapolation, since the price faced by banks are fairly close to each other. And third, the only points of evaluation are fully X-efficient from both the input and output standpoints, eliminating the potential problem of evaluating scope economies off of the efficient frontier.

The previous and current research on scale and scope efficiencies has several important implications for the course of future research. First, it strongly discourages the use of a translog cost function for research on banking or nay other industry where scale and product mix vary substantially. The translog is insufficiently flexible to describe an industry with increasing returns to scale up to some point and constant returns thereafter, and seems to have difficulties when firm tend to changes product mix significantly as they change scale. The translog and the Box-Cox approximation also typically perform poorly in estimating scope economies because they have trouble with evaluation at or near zero.

Second, future search should further investigate the concept of financial scale economies. The inclusion of the cost of risk has a modest effect on the scale efficiencies of some banks. What is not yet known is the robustness of this result, and how important it might be to other types of financial institutions, such as insurance companies, which may have very different risk-scale relationships.

Third, it is important in future applications to estimate scale and scope efficiencies only on the X-efficient frontier, where they are properly defined. Scope efficiencies measured off of the frontier appear to have been confounded with X-efficiency differences, although this problem does not appear to have occurred for scale efficiencies.

Fourth, additional research is needed on the optimal scope economies concept. This concept examines whether it is optimal from a profitability standpoint, including both costs and revenues, to produce all the products as opposed to specializing in one or more of them. Implementation of the new concept also provides at least partial solution to the known problems of scope economy estimation.

Finally, future research should also concentrate on estimating scale as well as scope efficiencies from the profit function, so that both revenue and cost effects on these efficiency measures can be taken into account. In the only prior study to estimate scale efficiencies from the profit function, found significant scale efficiencies using a conventional (nonfrontier) profit function. Future research should continue along these lines, but estimate scale efficiencies using the efficient frontier to avoid the possibility of confounding scale efficiencies with X-efficiencies. (Hancock 1992).

The efficiency of the banking industry:

A number of papers in recent years have considered the X-efficiency of US commercial banks. We use the term X-efficiency here to describe all technical and allocative efficiencies of individual firms, as distinguished from scale and scope efficiencies. The one result upon which there is virtual consensus is that X-efficiency differences across banks are relatively large and dominate scale and scope efficiencies. However, there is no consensus on the best method for estimating X-efficiency, or on the average level of X-efficiency of the banking industry.

The major econometric problem lies in distinguishing X-efficiency differences from random error that may temporarily give certain institutions relatively high or low costs. Four different approaches have been employed in evaluating bank data. Each of these approaches maintains a different set of assumptions about the probability distribution of the X-efficiency differences and random error for the purpose of distinguishing between these two explanations of cost dispersion. The econometric frontier approach (EFA) generally assumes that inefficiencies follow as asymmetric half-normal distribution, that random errors follow a symmetric normal distribution, and that both are orthogonal to the cost function exogenous variables. (Ferrier and Lovel 1990, Timme and Yang 1991).The thick frontier approach (TFA) assumes that deviations from predicted costs within the lowest average-cost quartile of banks in a size class represent random error, while deviations in predicted costs between the highest and lowest quartiles represent X-inefficiencies. The data envelopment analysis (DEA) approach generally assumes that there are no random fluctuations, so that all deviations from the estimated frontier represent inefficiency. Finally, the distribution-free approach (DFA) assumes the efficiency differences are stable over time, while random error averages out over time.

There is no simple rule for determining which of these methods best describes the true nature of the banking data. This would not be a problem if all of the methods arrived at essentially the same conclusion. Unfortunately, this is not the case - in fact, the choice of measurement method appears to strongly affect the level of measured inefficiency. Authors applying the EFA, TFA, and DFA methods to banking usually find average inefficiency to be about 20- 25% of costs, while authors using DEA find results ranging anywhere from less than 10% to over 5%. Perhaps a more important problem is that when these methods are compared with one another using the same data set, the rankings of individual banks often do not correspond well across methods, even when the methods find similar average efficiency levels. Moreover, the efficiency levels found are not invariant to the specification of which financial products are specified as inputs and outputs, which are fixed versus variable, or which output metric is employed, even when the measurement method is held constant. Thus, the results of using different methodologies and models are not mutually consistent, making it difficult to determine which institutions are most and least efficient, and which prospective industry entries, exits, and consolidations are most likely to improve overall banking performance.

In addition, there are very few studies of the efficiency of banks and other financial institutions outside of the US, and to our knowledge there are no prior frontier efficiency comparisons across international borders. For example, scale and scope efficiencies in Canadian, Israeli, and Finnish financial institutions without using frontier methods, applied the DEA frontier technique to Norwegian banks. Given the prior findings of relatively large X-efficiencies within one country and the increasing level of competition across countries, there is a clear need for measuring and comparing X-efficiencies across borders.

Several papers in the current special issues contribute to this literature. Applying the distribution-free approach using the profit function in place of the cost function, which brings output inefficiencies into the model as well as input inefficiencies. Banks can err by producing at the wrong level or mix of outputs as well as by employing the wrong level or mix of inputs, and only the profit function can incorporate all of these inefficiencies.

The profit function has never before been applied to estimating the X-efficiency of financial institutions and its few previous applications elsewhere did not separate out all of the input and output technical and allocative inefficiencies (defined below). The innovation of adding the output side appears to be quite important. The out inefficiencies are on average larger than input efficiencies. This is, most of inefficiencies are in the form of deficient revenues, rather than excessive costs, suggesting that use of the standard methods may substantially understate bank inefficiency.

On average, banks appear to lose about one-half of their potential profits to inefficiency also devise and implement a new method of decomposing total X-inefficiency into allocative and technical components that may be more useful than the standard definition. They define allocative inefficiency as the loss of profits from choosing a poor production plan, and model this as the effect of basing decisions on shadow prices instead of actual prices. They define technical inefficiencies as the loss of profits from failing to meet this production plan.

This decomposition of total efficiency into allocative and technical components differs substantially from the standard decomposition created. The standard method restricts technical inefficiencies to be in the form of radial, or equiproportionate overuse of all inputs, and force all deviations from input mix from the optimum into allocative inefficiencies. It is shown that the standard decomposition is a special case of Berger et al.'s decomposition. The new decomposition may also be more useful because it focuses on the source of inefficiency, i.e., poor plan versus poor implementation. Berger et al. find that most of the inefficiency is technical in nature - that banks err primarily in not meeting this production plans, as opposed to choosing unprofitable plans.

A surprising finding in this study is that larger firms are substantially more X-efficient on average, or closer to the frontier, than smaller firms. This finding may offset some of the scale diseconomies found for the larger banks in cost studies. Given that most of the measured inefficiencies are on the output side, this suggests that larger firms may have an advantage in terms of achieving high-value output bundles.

Several other papers, (Grabowski et al. 1993) in the special issues also estimate X-efficiency for commercial banks. Using the econometric frontier approach, find average input efficiency to be about 88% for a sample of large US bank holding companies, and using DEA, find average input inefficiency to be about 68% for a sample of US multibank holding companies and branching banks. These papers will be discussed in detail below in the section on the determinants of financial institution efficiency. The following section on bank mergers also describes three papers that estimate the change in bank efficiency associated with mergers.

The past and present papers on bank X-efficiency suggest several important topics for future research; First, the lack of correspondence among the efficiency levels and rankings for the different measurement approaches suggests that more researching comparing these techniques in needed. One step in this direction would be to determine more information about the standard errors of the efficiency estimates to facilitate comparisons across approaches. It may also be the case that the maintained assumptions of each of the four main approaches about the probability distributions of the efficiencies and random error may best fit different data sets. If so, and there is no unambiguously best method for all data sets, then research leading to the most appropriate method for a given data set may be useful.

Second, the evidence provided by Berger et al. and English et al. suggesting that output inefficiencies are as large or larger than input inefficiencies strongly implies that more research is need using the profit function, output distance function, and other methods that include output inefficiencies. To our knowledge, no case has ever been made for excluding output inefficiencies - banks and other firms seem just as likely to make mistakes in maximizing revenues as in minimizing costs.

Third, more research is needed comparing these output-inclusive approaches with other approaches. For example, a profit function model and a cost function model estimated using the same data set could be compared. If the profit function model were to find much larger inefficiencies, this would support the case that cost function models seriously understate efficiencies. If not, it might suggest something about the levels and relationships between input and output inefficiencies.

Fourth, additional research on Berger et al.'s alternative technical and allocative efficiency decomposition, using the cost, revenue, or profit functions, is in order. Given that the standard decomposition is a special case of their definition, it would be of inters tot determine how far apart the standard and new methods are empirically by estimating both on the same data set using the same efficiency criterion (i.e., cost minimization, or revenue or profit maximization).

Finally, much more research is needed measuring and comparing the efficiency of banks and other financial institutions across international borders, as in Berg et al. the integration of EC markets, as well as the general globalization of financial markets, means that the most efficient institutions may eventually dominate world markets. The cross-country comparisons may also shed some light on the efficiency effects of various regulatory policies. Substantial differences in efficiency across nations would tend to suggest that regulatory polices be coordinated and made roughly equal.

The efficiency implications of bank mergers:

The recent wave of large bank mergers in the US, combined with the prospects for sweeping international mergers subsequent to EC integration, has put the spotlight on the efficiency implications of bank mergers, particularly mergers between large institutions. If these mergers are successful in improving banking industry efficiency, substantial benefits may accrue to the customers and claimholders of these banks, and the level of competition within the banking industry may be considerable increased. Moreover, the efficiency effects of mergers constitutes an important policy question on its own, since merger applicant often cite prospective efficiency benefits as a justification for merger approval.

Most prior studies of bank merger efficiencies compared simple pre-merger and post-merger financial rations, such as operating costs divided by total assets, or the return one equity or assets. (Rhoades 1986, 1990, Linder and Crane 1992, Cornett and Tehranian 1992, cornett and Tehranian 1992, Spindt and Tarhan 1992, Srinivasan 1992, Srinivasan and Walls 1992).

In most cases, the authors of these studies corrected for the change in these financial rations for other firms in the industry to avoid confounding secular change within the industry with the efficiency effects of mergers. One prior study also simulated the pre- and post merger costs of hypothetical merger partners using a cost function to determine the potential effects of mergers on costs.

Most of these studies found no benefits of average from mergers. The exceptions are Cornett and Tehranian (1992) and Spindt and Tarhan (1992). Interestingly, these two exceptions found most of the merger benefits on the revenue or output side, while most of the studies that did not find merger benefits examined only the cost or input side. We will return to this issue below.

There are several problems with these prior studies that examine simple financial ratios. First, financial rations may be misleading indicators of efficiency because they do not control for product mix or input prices. Implicitly, studies using a cost-to-asset ration assume that all assets are equally costly to produce and all locations have equal costs of doing business. In addition, the use of a simple ratio cannot distinguish between X-efficiency gains and scale and scope efficiency gains. This inability greatly reduces the predictive power of the ratios in determining which types of mergers are likely to be successful in improving efficiency, since scale and scope efficiencies automatically changes when a merger is consummated, but X-efficiencies may or may not change.

Only one prior study (Berger and Humphrey 1992a) which we are aware used a frontier method to determine the efficiency effects of bank merger. Using the distribution-free approach, that study found very small, statistically insignificant average X-efficiency benefits from mergers among banks with over $1 billion in assets each. These benefits were more than offset by the scale diseconomies created by merging banks that were generally larger than efficient scale, resulting in a small total efficiency loss that was sometimes statistically significant. They also found that there were no efficiency gains associated with mergers in which the acquirer was more efficient than the acquired bank or in which both banks were represented in the same local market, two conditions often thought to be conducive to merger efficiency gains.

There are several implications of these findings for future research into the merger efficiency question. First, the literature should move away from the use of simple financial ratios and into the use of frontier efficiency techniques. As noted above, frontier techniques take account of product mix and input prices and need not make extreme assumptions about all assets in all locations having the same minimum efficient costs. In addition, frontier techniques allow for the separate estimation of X-efficiency from scale and scope efficiencies, which may have very different merger implications.

Second, further research is needed to determine the factors which predict merger efficiency gains or losses. Although most of the literature found no efficiency benefits on average, all of the studies found that some of the mergers were successful, while others were not. For policy purposes, it is important to know the factors that predict efficiency benefits, given that merger participants often claim such benefits when applying for regulatory approval. Unfortunately, the research to date has only found conditions that do not seem to work in predicting merger success - market overlap and greater relative efficiency of the acquiring firm.

Third, the profit function should be applied to merger efficiency analysis. The past and present research that found average efficiency benefits from mergers all included revenue effects, while the research that did not find such benefits generally used only cost data. This 'coincidence' suggests that mergers might improve efficiency, but only on the output side. This could occur if the merger helps the consolidated bank better achieve a higher revenue output bundle through improved marketing, product innovation, repricing, risk management, or other revenue-enhancing effects. The results of current and prior research consistently point to the possibility of output efficiency gains from mergers that could be captured and analyzed using a profit function frontier analysis. Finally, future analysis will be needed on the question of whether merger efficiency effects are time-dependent.

The efficiency of thrifts and governmental financing institutions:

The efficiency of thrift institutions, such as savings and loan (S & Ls) and credit unions (CUs), has not been studied extensively in the literature to date. Even less is known about governmental financial institutions, such as the Federal Reserve, that complete with the private sector in providing measurable financial services as well as performing traditional governmental functions. Given the substantial inefficiency found in banking, the study of efficiency questions for these competitors to banks takes on added importance.

The previous literature on thrift efficiency has concentrated on measurement of scale and scope efficiencies Hermalin and Wallace (1992) applied a DEA approach. Controlling for other factors, Cebenoyan et al. found no significant difference in efficiency between mutual and stock S & Ls, while Hermalin and Wallace found efficiency to be directly correlated with an S & L's product mix.

A study (by Mester 1993) uses the econometric frontier approach to provide further evidence on efficiency in mutual and stock S & Ls. Mester makes an important extension of the standard EFA model to permit both the cost frontier and error structures to differ between the two organization forms. A likelihood ration test indicates that the data supports this unrestricted model. This finding implies that efficient mutual and stocks use different production technologies. Mester also finds mutual to be more efficient on average than stocks, contrary to conventional wisdom, her own previous results. Finding is discussed further in the section below on the determinants of efficiency.

The past and current research on the efficiencies of thrifts and the FRS suggest that much research remains to be done on these topics. First, the recommendations for future banking research given in the previous three sections of this paper apply with equal or greater vigor to thrifts. Despite this industry's importance, its efficiency remains understudied.

Second, the finding of Mester that the cost frontier and error structures differ between mutuals and thrifts implies that future studies that compare the efficiency of two or more groups should consider checking whether the groups differ in all the available dimensions. Misleading comparison could be made if, for example, cost of profit function coefficients were incorrectly forced to be equal across groups.

Third, the model or Fried et al. suggests that care should be taken in matching the output or service characteristics with the organizational forms. If the form is cooperative or mutual, then price and service variety should be specified as beneficial outputs, while the standard outputs are likely sufficient for the standard corporate form.

The efficiency of the insurance industry:

The insurance industry currently faces many challenges, including increased competition, solvency risks, and a changing regulatory environment. The question of the efficiency of the firms in this industry is clearly important to determining how the industry will respond to these challenges and which firms are likely to survive. Despite the importance of such analysis, there has been very little prior research examining the efficiency of insurers. (Weiss 1991). Moreover, the previous insurance cost literature typically focused exclusively on scale and scope efficiency. However, it is also possible that the reported efficiency-size relation ship is indicative of heteroskedasticity problems, as discussed. (Yuengert 1993).

First, there is no consensus on the best measurement of outputs and costs for insurers, or on the average level of efficiency in this industry. Similar problems have been encountered in research on other financial services industries. Further research on the robustness of results to the choice of outputs and costs is needed.

Second, these papers indicate that the insurance industry displays substantial dispersion both across groups and across firms in measured X-efficiency. Further research is needed in identifying the determinants of efficiency in this industry in order to predict which groups and individual firms are likely to survive the increased competition of the future.

Third, more research along the methodological lines started by Yuengert (1993) should be conducted using both insurance and noninsurance data. His research strongly suggests that heteroskedasticity, as well as the specification of the distributions of efficiency and random error, may substantially affect the results reported by other investigators using standard techniques. If future research finds that these results are robust, a wholesale change in the methodology for measuring efficiency may be in order.

Fourth, all of the insurance efficiency studies to date use a cost frontier approach to examine input efficiency. Additional research using other methods such as DEA, a profit function, or output distance function, should also be applied to examine the robustness of the results, and could provide a more complete understanding of the health of this industry.

Finally, to our knowledge, no research to date has examined the X-efficiency of insurers outside the US nor have transnational insurers been specifically examined. Research into X-efficiency in the international arena should provide insights into the likely results of future cross-border competition as well as the need for internationally coordinated restructuring of insurance regulation.

The determinants of financial institution efficiency:

Since X-efficiency was introduced in the 1960s, great strides have been made in developing techniques to measure it. However, relatively little empirical research has been devoted to developing and understanding of those factors which influence a firm's efficiency. Some prior studies focused on the impact of regulation and organizational form on costs and scale and scope efficiencies, but these earlier studies did not relate this factor directly to X-efficiency.

A number of papers reverse this trend and make inroads in examining the determinants of financial institutions' X-efficiency. Factors that are likely to influence a firm's X-efficiency may be grouped into (1) agency problems between owners and managers, (2) regulation and organizational and legal structures, and (3) scale and scope of operations.

Financial institutions' X-efficiency may also be related to the scale and scope of operations. The efficiency-scale relationship is important for making proper managerial decisions and has recently been the focus of attention by regulators and other government officials. Berger et al. (1993) find that X-efficiency is strongly positively related to a bank's scale of operations. Given that most of their X-efficiency differences are on the output side, this implies that larger firms may be better able to reach their optimal mix and scale of outputs, increasing output efficiency.

The studies make important inroads into finding the determinants of financial institution efficiency. They also highlight many of the challenges and potentially fruitful avenues for future research. First, the determinants of efficiency should be given more theoretical underpinning. For example, in recent years there have been many significant advances in our understanding of principal-agent relationships, but prior to the current papers, little or no research had linked these agency problems to efficiency. More research is warranted in the area of linking these theoretical advances to X-efficiency.

Second, more attention should be given to the potential for sample selection problem when comparing groups of firms. When firms shift among groups, such as the mutual-to-stock conversions cited by Mester (1993), control should be used to avoid sample selection biases. For example, firm that shift could be deleted from the samples entirely or studied separately.

Third, some studies (Fare and Primont 1993) establish the maximum theoretical efficiency gains owing to a single-unit organization's better ability to allocate resources over a multi-unit organization. However, various factors such as agency costs, managerial skills, and local market limitations may significantly inhibit single-unit organizations form achieving these maximum gains. Ideally, establish the maximum gain for the phenomenon under investigation and this value would be compared to the observed actual efficiency gain. Such studies would be further enhanced by ascertaining the factors which impeded the achievement of the maximum gain.

Fourth, further research is needed into the relationships among scale, scope, and X-efficiencies. For example, if larger firms are more X-efficient (i.e., closer to the frontier), this may offset scale diseconomies (i.e., a higher cost frontier for larger banks).The research citied here strongly suggests that such research should be performed using methods that incorporate output inefficiencies, since the output side seems to be where the relationship between X-efficiency and scale in strongest. Moreover, prior research has indicated that the relationship between input measures of efficiency and profitability is relatively weal (Berger 1991, Timme and Yang 1991), suggesting that differences in input efficiencies may be somewhat offset by differences in output efficiencies or other factors.

Finally, I believe that more econometric procedures should be employed to establish the casual relationship between X-efficiency and its hypothesized determinants. A limitation in some of the studies is that measured X-efficiency is regressed on a set of these hypothesized determinants, such as size, ownership structure, capital-asset ration, etc. This may create problems of interpretation in efficiency is also a causal factor for some of the regressors. For example, X-efficiency may be positively related to size because larger firms become efficient (e.g., by virtue of their ability to achieve optimal output), or because more efficient firms compete more effectively and become large. To our knowledge, no study to date has utilized a methodology which distinguishes the causality of efficiency from the effects of efficiency on the other variables.


Each session of this assay contains the discussions comments and panel discussions of the experts in order to provide comprehensive view of world economic. I also include the comments of three panels comprised of academic government, and industry experts. The essay discusses the myths and realities of the benefits form consolidation in the US banking system. These panelists' comments immediately follow the papers and discussants' comments on bank mergers. Also it concerns the efficiency of financial service delivery around the globe, particularly Eastern and Western Europe and Japan. In addition it views about the future of financial service firms, and includes separate discussions of the future of the banking thrift, securities, and insurance industries.


· Berger, A.N., 1993, 'Distribution-free' estimates of efficiency in the U.S. banking industry and tests of the standard distributional assumptions, Journal of Productivity Analysis 4 (forthcoming).

· Berger, A.N., and D.B. Humphrey, 1991, The dominance of inefficiencies over scale and product mix economies in banking, Journal of Monetary Economics 28, 117-148.

· Berger, A.N and D.B. Humphrey, 1992a, Megamergers in banking and the use of cost efficiency as an antirust defence, Antirust Bulletin 37 (Fall), 541-600.

· Cebenoyan, A.S., E.S. Cooperman, C. Register and S. Hudgins, 1993, The relative efficiency of stock vs. mutual S & Ls: A stochastic cost frontier approach, Journal of Financial Services Research 7 (forthcoming).

· Cornett, M.M. and H. Tehranian, 1992, Changes in corporate performance associated with bank acquisitions, Journal of Financial Economies 31, 211-234.

· Ferrier, G.D. and C.A.K. Lovell, 1990, Measuring cost efficiency in banking: Econometric and linear programming evidence, Journal of Econometrics 46, 229-245.

· Grobowski, R., N. Rangan and R. Rezvanian, 1993, Organizational forms in banking: An empirical investigation of cost efficiency, Journal of Banking and Finance 17, 531-538 (this issue).

· Handcock, D., 1992, Testing for subadditivity and economies of scope in banking using the profit function, Working paper, Board of Governors of the Federal Reserve System.

· Humphrey, D.B., 1990, Why do estimates of bank scale economies differ?, Federal Reserve Bank of Richmond, Economies Review 76 (Sept./Oct.), 38-50.

· Linder, J.C. and D.B. Crane, 1992, Bank Merger: Integration and profitability, Working paper (Hayvard Business School).

· McAllister, P.H. and D.A. McManus, 1993, Resolving the scale efficiency puzzle in banking, Journal of Banking and Finance 17, 389-405 (this issue).

· Mester, L.J., 1993, Efficiency in the savings and loan industry, Journal of Banking and Finance 17, 267-286.

· Pulley, L. and D.B. Humphrey, 1993, The role of fixed costs and cost complementarities in determining scope economies and the cost of narrow banking proposals, Journal of Business 66.

· Rhoades, S.A., 1986, The operating performance of acquired firms in banking before and after acquisitions, Staff Economic Studies 149 (Board of Governors of the Federal Reserve System, Washington, DC).

· Rhoades, S.A., 1990, Billion dollar bank acquisitions: A note on the performance effects, working paper (Board of Governors of the Federal Reserve System).

· Spindt, P.A. and V. Tarhan, 1992, Are there synergies in bank mergers? (Tulane University, New Orleans, LA).

· Timme, S.G. and W.K. Yang, 1991, On the use of direct measure of efficiency in testing structure-performance relationships, Working paper (Georgia State University).

· Weiss, M., 1991, Efficiency in the property-liability insurance industry, Journal of Risk and Insurance 58, 452-479.

· Yuengert, A., 1993, The measurement of efficiency in life insurance: Estimates of mixed normal-gamma error model, Journal of Banking and Finance 17, 483-496