This study is devoted to decompose the gMMPI into a set of components related to productivity change using equations (18) and (20). Recall that a value of these components greater than one indicates a productivity gain, while a value less than one indicates a productivity loss. As far as the metafrontier is concerned, the left part of Table 2 presents the rates of change in gMMPI and its individual components per annum. The average rate of gMMPI is found to be equal to 0.21% per year, which is consistent with the finding of Casu et al. (2004) and Casu and Girardone (2005) over the period 1994-2000, Murillo-Melchor et al. (2009) over the period 1995-2001, as well as Barros et al. (2010) over the period 1996-2003. It is noticeable that this measure of productivity growth consists of technical efficiency gains (0.06%), technical progress (0.05%) and the effect of scale improvement (0.10%) toward constant returns to scale. Figure 1 draws the foregoing measures of productivity
change over the sample period. Measure gMMPI varies closely with TEC*
particularly before 2001. SEC* also fluctuates over time, while TC* exhibit a gradual upward trend initially, accompanied by a slight downward trend.
[Insert Table 2 and Figure 1 here]
We divide the entire sample period into three sub-periods. Evidence is found that West European banking industries experienced a faster positive productivity growth during 1993-1998, immediately after the establishment of the single market for European financial services leading to international economic integration in this area, than the latter two sub-periods. This productivity gain arose probably from the regulatory reform in banking, such as financial services restructuring and consolidation, aiming at increasing competition among financial institutions and across borders. It is worth mentioning that there exists an increasing positive scale effect contributing to the productivity growth during the sample period. In fact, this effect constitutes the main source of the productivity gain, resulted from, e.g., bank consolidation and the enlargement of banking markets.
The gMMPI is further decomposed into technical efficiency change, technical change, and scale efficiency change based on the group frontiers. The results are listed in the middle of Table 4, while various catch-up effects are shown in the right part of the table. Recall that MMPIt t,+1=TECt tk,+1×TCt tk,+1×CUTt t,k+1×PTCt tk,+1, which differs
from the gMMPI due to the omission of scale effect, i.e., term SEC*. Since the average rate of productivity growth for MMPI is 0.11% per year, this omission leads to an underestimation, as compared with the 0.21% of the gMMPI. Among the four sources, technical efficiency change (SEC) of the group frontiers plays the most important role, followed by the PTC, while the remaining two sources curb productivity.
18
The scale effect on the ground of group frontiers is essential, not only for capturing scale efficiency change, but also for helping shed light on the effect of catch-up in scale. The results show that an average bank sustains a slow improvement of its production scale, 0.02% per annum, toward constant returns to scale during the sample periods. Most importantly, the ratio of the scale efficiency change of the metafrontier to that of group frontiers exceeds one, suggesting that there exists catch-up in scale. That is, the production scale measured on the group frontiers is adjusting toward CRS faster than that measured on the metafrontier. Finally, before the advent of European Economic and Monetary Union (EMU) in 1999 Rao’s catch-up effect (MTRC) is slightly greater than unity and sustains over the following three years, then falls below unity thereafter, mainly entailed by the regress of CUT.
This reveals that the stimulation effect of the EMU on the banking productivity appears to last in a short period. Figure 1 depicts the three measures over time. It is seen that MTRC and CUT are almost fluctuating synchronously, but the curve of PTC hardly varies with a smooth downward trend.
[Insert Table 3 here]
Table 3 reports the mean values of these measures in a descending order according to the gMMPI. Banks in nine out of the fifteen countries have positive measures of the gMMPI, implying that their productivity grows with time evaluated against the metafrontier, while banks in the remaining countries undergo a productivity decline. The highest productivity growth is found by Belgium (1.05%), followed by Sweden (0.99%), and United Kingdom (0.99%), while the lowest rate of growth occurs in Spain and Portugal (both are found to be equal to -0.56%), followed by Switzerland (-0.19%), Norway (-0.18%), Italy (-0.16%), and Netherlands (-0.13%).
Our findings are similar to Pastor et al. (1997), Kondeas et al. (2008), and Murillo-Melchor et al. (2009) without relying on the metafrontier framework. Taking
a closer look at the three elements, all but one (Denmark) and eleven out of fifteen countries show scale efficiency gain and technical progress, while merely seven out of fifteen countries enjoy technical efficiency gain. In particular, banks in Spain and Portugal are found to experience the slowest productivity growth due primarily to worse technical efficiency change.
It is crucial to note measures TEC, TC, and SEC, shown in the middle columns of Table 5 are not comparable among countries, because they are computed against distinct country frontier. The evidence found in the previous paragraph may be attributed to the trend toward consolidation in response to the intensified competition among banks within the EMU. According to Goddard et al. (2001), all West European countries, apart from Portugal, have undergone a decline in the number of banks since 1989. Faced with this new atmosphere, a bank must manage to enhance its competitiveness by lowering production costs by means of, e.g., merger and acquisition activities to expand the production scale and adopting innovations in a prompt manner to stimulate technical progress instantly.
Merger and acquisitions may incur extra costs of adapting to the new environment and reconsidering business strategies, disappearance of traditional revenues, and any other unfinished banking reforms. Another explanation related to the economic environment stresses that the lower the level of economic conditions, the harder is to engage in banking activities, as noted by Lozano-Vivas et al. (2001, 2002). Our finding confirms this argument in that, on the basis of Table 1, Spain and Portugal have the lowest levels of per capita income among the sample states, leading to technical efficiency losses.
We respectively classify the entire sample into eight groups based on total assets and the ratio of equity to total assets. Table 4 presents the outcomes. According to panel A of Table 4, smaller and larger banks appear to grow faster than medium-sized
20
banks do. Evidence is found to support the establishment of large banks as this form of banks exhibit swift rates of growth in TC* and SEC*, accompanied by a gradual regress in TEC* as a consequence of merger and acquisitions. Conversely, the major contribution to productivity growth of smaller banks stems from TEC*. Figure 2 shows a U-shape for the average gMMPI curve against various size classes. Curve TEC* decreases as the bank size grows, while the curves of TC* and SEC* rise with bank sizes. It is also evident that small banks exhibit higher catch-up in technology, while larger banks exhibit higher catch-up in scale.
[Insert Table 4 and Figure 2 here]
Panel B of Table 4 shows the productivity differences for banks with diverse attitudes toward risk. Since equity capital provides a buffer against portfolio losses, it is expected that the higher the capital to assets ratio, the less likely is a bank involving in insolvency, and vice versa. It is interesting to note that the highest three classes of banks corresponds to the quickest productivity growth, mainly because they have the highest average values of TEC*. More conservative banks tend to grow in a rapid rate.
Similarly, TC* improves with ETA for the last four classes, while SEC* varies irregularly. The outcomes appear to be quite reasonable, since there is incentive for a risk-averse bank manager to engage in monitoring and supervising activities and making circumspect decisions for the purpose of avoiding the exposure of excess risks (Kumbhakar and Wang, 2007). Figure 3 illustrates that MTRC and its two components have gradual increasing trends, though fluctuate.
[Insert Figure 3 here]