الگوبرداری و بهره وری کل عوامل مبتنی بر فن آوری: پیشنهادهای جدید و برنامه کاربردی

The present study fills a gap between the benchmarking literature and multi-output based efficiency and productivity studies by proposing a benchmarking framework to analyze total factor productivity (TFP). Different specifications of the Hicks–Moorsteen TFP index are tailored for specific benchmarking perspectives: (1) static, (2) fixed base and unit, and (3) dynamic TFP change. These approaches assume fixed units and/or base technologies as benchmarks. In contrast to most technology-based productivity indices, the standard Hicks–Moorsteen index always leads to feasible results. Through these specifications, managers can assess different facets of the firm's strategic choices in comparison with firm-specific relevant benchmarks and thus have a broad background for decision making. An empirical application for the Spanish banking industry between 1998 and 2006 illustrates the managerial implications of the proposed framework.

Literature on benchmarking focuses on the selection of a unit of strategic value against which performance is compared [1]. Another series of academic studies analyze the efficiency and productivity of firms with multiple inputs and outputs. So far there seems little or no link between these two streams of research. In this paper we propose to bridge this gap by defining novel total factor productivity (TFP) benchmarking methods. These are devised to include cross-sectional and inter-temporal perspectives not only concerning unit to unit benchmarking, but also efficiency frontier benchmarking. These various perspectives are introduced stepwise starting with static indices, continuing with fixed base and unit, and ending with dynamic benchmarking. This provides managers of any industry with a new set of TFP benchmarking indices for decision making. Both benchmarking and TFP analysis represent key tools in business economics. For instance, Balk [2] points to two main actions a manager constantly carries out: the monitoring activity (i.e., assessing how the firm is doing over time) and the benchmarking activity (i.e., comparing firm performance with respect to its main competitors). Although both activities aim at enhancing performance, monitoring is internally oriented while benchmarking has an external focus. Benchmarking is defined as the search and emulation of the industry's best practices and it thus is an objective setting procedure [1]. Through benchmarking, a firm can deduce whether it has a best or worst practice. Thus, it can aim at maintaining superiority or at closing the gap to its competitors [1]. Therefore, benchmarking appeals most to firms with similar strategic orientations or facing comparable problems and opportunities [3] and [4]. Empirical applications suggest different methods for monitoring or benchmarking activities. In managerial studies of performance, the simplest method is the use of output-input ratios or any other kind of ratios for that matter (see [5] and [6]). Managers care about profitability and implicitly about productivity: “the most encompassing measure of productivity change, TFP change, is nothing but the “real” component of profitability change. Put otherwise, if there is no effect of prices then productivity change would coincide with profitability change” ([2]: 6). The above TFP measures are easily adaptable to benchmarking purposes. One can simply divide the firm's TFP change (or performance) ratio to the one of a chosen competitor. However, in multiple inputs and outputs technologies various problems emerge related to the use of ratios for benchmarking. When comparing two firms, different partial productivity ratios (built by dividing different outputs by some inputs) can point to different results. The management literature suggests a way to remedy this problem. Specifically, in the presence of prices, multiple outputs and inputs productivity indices are proposed by the American Productivity Center (APC) method [7]. Turning attention to efficiency and productivity analysis, this literature uses frontier methods with economic underpinning in production theory to handle multiple inputs yielding multiple outputs. These non-parametric techniques have known an important upsurge and are probably best known under the label Data Envelopment Analysis (DEA) (see [8] and [9]). DEA methods compute the degree of inefficiency separating a certain Decision Making Unit (DMU) from the efficiency frontier. In this case, the comparison is done against the whole analyzed sample, not against some specific strategic competitor as in benchmarking. Thus, in DEA benchmarks are the efficient units on the frontier against which the other DMUs are projected using some efficiency measure (see [8] and [9]). Therefore, it is highly unlikely that a single benchmark is found for all units evaluated in the sample. In inter-temporal analyses, the efficiency and productivity literature captures the potentially shifting efficiency frontier usually through index numbers. The Malmquist productivity index is probably the best known measure that has been extensively used in past research.1 However, there are some pitfalls to the use of Malmquist indices. First, it is not always a TFP index: while the TFP properties are maintained under constant returns to scale, shortcomings appear in the presence of variable returns to scale (VRS) which mostly represents the true technology [13]. Second, there is the possibility of having infeasible results.2 For example, Glass and McKillop [15] find infeasibilities for up to 7% of the analyzed UK building societies.3 This issue could have an important impact on benchmarking analysis, since managers wish to obtain firm level results that may not always be available.4 As a result, there are two main issues with the Malmquist index that need to be resolved: TFP interpretation and infeasibilities. To address these problems, one can turn to Bjurek's [19] proposal for a Hicks–Moorsteen TFP (HMTFP) index (see also [20]: footnote 18). The HMTFP index is defined as a ratio of an aggregate output quantity over an aggregate input quantity index. More precisely it measures the change in output quantities in the output direction and the change in input quantities in the input direction, instead of exclusively adopting an input- or output-orientation as Malmquist indices usually do. The TFP characteristics of the HMTFP index solve the limitations of the traditional Malmquist productivity index in the presence of VRS. Furthermore, this HMTFP index is well-defined under general assumptions of variable returns to scale and strong disposability.5 However, in spite of its attractive properties, the HMTFP has been scarcely empirically applied.6 Various benchmarking applications have been developed in the non-parametric efficiency and productivity analysis framework by isolating reference frontiers or DMUs. In the non-TFP context, Berg et al. [25] adapt the Malmquist productivity index to have a base year frontier as a benchmark frontier, and measure productivity growth or regress relative to this fixed basis. Similarly Berg et al. [26] adapt the Malmquist productivity index to make comparisons across countries with respect to a fixed basis (i.e., a single country) for a given year. Also, single benchmark TFP analyses have been undertaken by Zaim et al. [27], Färe et al. [28] and Zaim [29]. Manipulating a Hicks–Moorsteen index, their proposals include both cross-sectional and inter-temporal analyses by mixing a single DMU and TFP benchmarking. Zaim et al. [27] use a five years sample of OECD countries to analyze the well-being of individuals in each country as compared to a benchmark country. Similarly environmental performance is measured against a benchmark DMU in Färe et al. [28] and Zaim [29]. While the former study looks upon OECD countries at cross-sectional level, the latter analyzes US states from both cross-sectional and inter-temporal perspectives. A small existing literature thus proposes efficiency frontier comparisons using productivity indices combined with some form of unit to unit benchmarking. But, while consensus is reached regarding the usefulness of benchmarking, less agreement exists with respect to the choice of benchmarks. In a strategic analysis setting, the interest of a firm may be to know its relative performance to a certain specific competitor, instead of comparing itself to a frontier potentially composed of all firms in the sector. The benchmark could differ for each firm, even though it could remain the same over a certain time period. In addition, awareness of TFP positioning is useful in both static and dynamic environments. Efficiency coefficients (static) and TFP indices (dynamic) relative to a given benchmark are equally relevant and could represent the basis of strategic decision making. For instance, in the case of similar strategic configurations, firms constitute strategic groups and may choose their benchmark within their relevant cluster. In this case, the benchmark unit can be the leader of the strategic group or any other unit, say the local competitor, regardless of its performance. To develop a systematic framework to analyze these issues, this study proposes a TFP benchmarking framework by adapting Bjurek's [19]HMTFP index for benchmarking purposes. The introduced HMTFP indices for benchmarking include the features of the traditional HMTFP together with some of the properties of the indices in Berg et al. [25] and [26], Zaim et al. [27], Färe et al. [28], and Zaim [29]. Various specifications of the HMTFP index measure distances (and catching-up effects) between analyzed DMUs and their selected benchmarks: these indices offer TFP interpretations with respect to static, fixed base or changing efficiency frontiers. The empirical application considers the Spanish banking sector over the period 1998–2006, a post-deregulation growth phase. The sector experienced consistent growth following the disappearance of regulatory constrains and due to the competition between private and savings banks. In productivity and efficiency terms, the sector has been looked at from a multitude of perspectives.7 In addition, there is a wide range of DEA studies that focus on some alternative aspects of benchmarking. We mention some recent examples. Bougnol et al. [37] show how DEA can be used by practitioners to enhance standard performance evaluations such as benchmarking or constructing rankings based on scorecard assessments. Moreover, the versatility of DEA models for benchmarking allows to evaluate multiple-stakeholder perspectives using common sets of variables [38]. In a similar vein, DEA-based benchmarking can also be used for analyzing bank branch efficiency suitable for both line managers and senior executives [39]. However, our study is unique in focusing on integrating a benchmarking perspective into frontier-based TFP measures. This paper is structured as follows. Section 2 develops the HMTFP index adapted to benchmarking purposes. Section 3 presents sample information together with the variables and methods of analysis. The empirical application is found in Section 4, while the final section is dedicated to some concluding remarks.

This research is founded in the traditional view of benchmarking as the search and emulation of best practices. By applying the HMTFP index [19], this study aims at closing the gap between benchmarking and multi inputs and outputs TFP frontier analysis. In this way, TFP benchmarking can be a new way to set strategic objectives for managers and to analyze firm performance for regulators and researchers. The advantages of the proposed tool for benchmarking are various. First, this Hicks–Moorsteen type index, which is currently rather scarcely used, solves known problems of TFP measurement in the presence of variable returns to scale. Indeed, under weak assumptions of strong disposability and VRS, this index is always feasible. This property is crucial for benchmarking analysis as firm-specific results have to be provided. Thus, one implication is that the HMTFP index deserves greater attention. Second, through straightforward manipulations of the HMTFP index, versatile tools for benchmarking analysis are obtained. Pursuing a global image of TFP benchmarking, three measures result from diverse assumptions: (1) static benchmark analysis, (2) fixed base and unit benchmark analysis, and (3) dynamic benchmarking analysis using a new decomposition. These benchmarking viewpoints assume fixing a particular DMU as a benchmark (very little used in previous analyses) and/or base technologies (a classical benchmark approach) together with the pros of the standard HMTFP index. Each of these settings enables managers to see a certain facet of the firm's activity. These benchmarking indices are stand alone tools, but can also be potentially combined to offer a broad perspective for decision making. For the empirical analysis, this paper used benchmarking criteria based on current interests of banking institutions: technical efficiency and good stress tests results. While technical efficiency has always been considered an important aspect, stress tests attracted a lot of interest in the aftermath of the recent financial crisis. Results confirm the growth phase of the Spanish banking industry and illustrate how TFP scores evolve in the sector. This consistent growth phase originated at the end of the deregulation of the savings banks sector. For instance, for these banks Grifell-Tatjé and Lovell [30] found that productivity declined at the end of the 1980s, beginning of the 1990s. For the following period, recent studies find significant productivity increases due to either deregulations [33], or technological change (e.g., [45] and [30]). Furthermore, the fluctuations encountered in banks' TFP may be due to expansion or consolidation strategies, such as mergers. For example, Cuesta and Orea [41] find immediate efficiency decreases for savings banks involved in mergers, followed by significant increases. In our empirical application these fluctuations are revealed through significant catching-up effects. Moreover, this sudden decrease which follows expansion strategies can explain the TFP evolution of Bancaja, the benchmark. Also, Bancaja's significant TFP growth at the end of the analyzed period is in line with the results of Illueca et al. [36], who state that savings banks that expand outside their original markets achieve greater productivity gains. Throughout the paper, findings are first scrutinized by comparing the sample level results with the established benchmark under the various benchmarking scenarios. While key results are stable between the different approaches, differences may appear mostly due to the chosen treatment of the fixed or changing technology. Next, the same scenarios are analyzed in a unit to unit analysis revealing dissimilar behaviors of banking units. This study makes headway for future research since the proposed methodological tools can employ benchmarking criteria adapted to any scenario or industry. For instance, in the standard benchmarking approach comparisons against efficient units reveal the firm's position in the market and the distance separating it from the efficient units. This is a method to discover, understand and implement new organizational practices. In this line it could be interesting to define analyses by benchmarking against strategic groups' leaders. However, in some cases managers may want to compare performance against their local competitor, even if this may be an inefficient firm. For savings banks, this local competitor may be a unit that is developing its branch network in the same region. All these benchmarking options contribute to organizational learning and strategic planning and reveal how decision making can contribute to the performance of firms over time. Finally a limitation of this study – an avenue of future research – is the absence of risk variables in TFP indices. The importance of including risk measures has become acute following the recent financial crisis. One option could be to obtain risk-adjusted estimations of TFP (e.g., the work done by Hughes and Mester [50] in a cost function approach could be adapted to TFP indices). For instance, future studies could introduce the risk variables through outputs such as the credit-risk expressed as the amount of off-balance-sheet items. Alternatively one could use banking ratios defining the risk environment (e.g., percentage of insolvency provisions or simply the risk of assets).