ABSTRACT
At this point, we may draw up a future research agenda. Our first recommendation concerns the use of data. Although we are aware of the difficulty of finding good (proxies for) output prices, this book has once again emphasized the need for output prices in an analysis of bank competition and profitability. Without output prices, we are unable to calculate banks’ mark up on costs or to derive the reaction functions to their competitors. As a result we know very little about the differences between banks in a single market. The increase in concentration in all markets reviewed here makes this an important concern. A change in the definition of the production process of banks would also be welcome. We need to rethink the traditional intermediation approach and focus more on other types of income. Our choice of variables in all models described here is mostly determined by banks’ balance sheets. An increasing part of the action in today’s banking markets, however, takes place off the balance sheet. Including off-balance sheet items in the intermediation approach therefore is a first step towards a more balanced view of bank production. Our second recommendation regards the theoretical foundation of the models employed to measure market power and efficiency.As we have shown, models focusing on a single variable may suffer from identification problems. In particular, we emphasize the distinction between market power and efficiency, for example using the efficiency hypothesis test in Chapter 6 or through a comparison of cost and profit efficiency. In addition, the fact that we observe such strong trends in banking calls for time-dependent models. In particular, we stress the need for making both the price-elasticity of demand ηD and conjectural variation λ time-dependent. Our third recommendation concerns the market under examination. For reduced-form market structure models such as the SCP model, we advocate their application to a wider range of specific submarkets. (Sub)markets that are not very contestable and have experienced less internationalization (e.g. deposits or mortgages) lend themselves particularly well to this type of analysis. For non-structural models such as Iwata, Panzar-Rosse and Bresnahan we suggest estimating a different H or λ for different size classes and submarkets.
