Abstract Productivity change and shareholder value have been analysed in the banking sector in the last few years, although it should be noted that these two important aspects have been studied separately. In this regard, the main contribution of our study is to link these two lines of research by verifying whether those banks characterized by higher levels of efficiency and productivity change have a higher shareholder value. To measure changes in efficiency and productivity we use the Malmquist nonparametric technique, which is calculated from Data Envelopment Analysis (DEA) linear programming approach. The Malmquist total factor productivity index enables separation of the 'catching up' effect, i.e. changes over time in technical efficiency, from 'technological change', i.e. the shift of best practice frontier over time due to technological progress. Our results for a sample of listed Spanish banks in the period 2000 to 2004 confirm that those banks with higher efficiency and productivity changes have a higher shareholder value, even after controlling for the impact of traditional measures of performance, such as return on assets. Notes 1 Other studies measure productivity change using a parametric approach (Berger and Mester, Citation1999, Citation2001). However, Casu et al . (Citation2004), in their analysis of productivity change in European banking during the period 1994 to 2000, conclude that both parametric and nonparametric methodologies do not yield markedly different results in terms of identifying the main components of productivity change. 2 A further decomposition of the 'technical efficiency change' component to take into account variable returns to scale (VRS) technology has been proposed, which distinguishes between 'scale efficiency' and 'pure technical efficiency change' (Färe et al ., Citation1994). However, this further decomposition has been subjected to a number of criticisms (e.g. Ray and Desli, Citation1997). In this regard, there seems to be a consensus that the Malmquist index is correctly measured by the ratio of the CRS distance function even when the technology exhibits VRS (Casu et al ., Citation2004). 3 DEA is a technique originally introduced by Charnes et al . (Citation1978) as a reformulation of Farrell's (Citation1957) efficiency measure to the multiple-output, multiple-input case. This technique has been usually applied to evaluate efficiency in different economic sectors, specially for financial institutions (see Berger and Humprey, Citation1997 for a survey). 4 The hypothesis of CRS subsequently modified to allow for VRS (Banker et al ., Citation1984), which is the most commonly used specification in the 1990s, because the CRS assumption is only appropriate when all DMUs are operating at an optimal scale. The VRS approach provides technical efficiency scores that are greater than or equal to those obtained using the CRS model. 5 The DEA models can take two different orientations. The first one, called input orientation, seeks to identify technical inefficiency as a proportional reduction in inputs usage. The second one, referred to as output orientation, seeks to identify technical inefficiency as a proportional increase in output production. To date, the theoretical literature is inconclusive as to the best choice among these two alternatives. These two orientations yield equal values under CRS, but not when VRS is assumed (Thanassoulis, Citation2001). 6 The most-ebated issue regards the role of deposits. Under the production model, which views banks as service-producing organiztions, deposits are considered as an output. Under the intermediation approach, banks are viewed as financial intermediaries whose primary business is to borrow funds from savers and lend those funds to customers to obtain profits. Hence, in this case, deposits are regarded as inputs. 7 To run the DEA model, personnel expenses and administration expenses have been summarizd into a single variable in order to avoid an excessive number of inputs so that the proposed model can be accepted regarding the total number of variables (El-Mahgary and Ladhelma, Citation1995).