The purpose of this paper is to provide an econometric model which can be used to explain and determine the demand for cash in Norway, and to use this model to forecast the demand for cash in the period 2004.3-2007.4.
From a central banks perspective, there is particularly one reason to be interested in the demand for cash. The central bank is the issuer of notes and coins. Since one of its objectives is to satisfy the demand for cash, it is of the central banks interest to ensure that they supply the society with notes and coins in the most efficient way. Obviously this is also of great importance for the society. For the central bank to do this, an efficient amount of notes and coins must be produced. In order to be able to plan the production in the best way, it is therefore of interest to investigate closer what determines the demand for cash. Further, one may also argue that it is of interest to investigate the demand for cash for monetary policy purposes. The reason for this is that the amount of cash held for transaction purposes is closely related to domestic spending, and hence to domestic price developments.
The demand for cash increased steadily year by year until 1999, and has decreased somewhat since then. In the same period the use of alternative payment instruments, such as payment cards, increased rapidly. One would therefore expect that the use of cash should have been reduced dramatically, something that has not yet happened. This paper studies the determinants of cash demand in Norway during the period 1980-2004. A thorough discussion of the different determinants for cash demand is carried out. The starting point for the discussion is the concept of money and its role, the evolution of the payment system and theoretical models of money demand. Since most of the theoretical models of money demand focus on a quite narrow concept of money, it is believed that some of these models are also relevant when investigating the demand for cash. This seems to especially be the case for models considering the private households demand for money. Based on this discussion, empirical models of demand for real cash are developed. Rather than estimating the theoretical models themselves, they are used to determine which explanatory variables that should be considered when an empirical model is specified.
In the empirical analysis, quarterly data and a general-to-specific approach are used. For this analysis I have used the software programs PcGive 10.1 and TSP 4.5. Due to problems with autocorrelation in the residuals and lack of data, a VAR model seems not appropriate to use for estimating the demand for cash. Instead a single equation equilibrium correction model is used. Considering different initial models, eliminating insignificant variables and applying different statistical tests, suggests that there are two competitive, but also quite similar, models of relevance. Both models have only private point-of-sale consumption and real deposit interest rate as significant explanatory variables in addition to a linear trend, seasonal variables and impulse dummies. The rapid evolution in the payment system seems to have a negative effect on the demand for cash, which may dampen the increase in the demand for real cash. This effect is represented in the models by a negative linear trend.
The estimation results show that the two models short-run effects differ slightly, while the long-run effects are quite similar. The long-run elasticity for consumption was found to be approximately 0.63 and the semi-elasticity for the interest rate was found to be approximately 0.02 in both models. These results differ compared to the results obtained by Fischer, Köhler and Seitz (2004) for the Euro area. The explanatory variables were found to be weakly exogenous with respect to all parameters in the structural equation for real cash, validating a single equation approach. In addition to this, the highly significant equilibrium correction term suggests that an equilibrium correction model is appropriate.
In order to choose between one of the two competitive models, parameter stability and ex post forecasting properties are investigated. However, this does not lead to a clear conclusion of which model to choose. Both models seem to have parameters that are reasonable stable. On the other side, both models have some problems when it comes to forecasting in the sample period 2002.1-2004.2. An explanation for this may be that this forecasting period was a rather turbulent period for the Norwegian economy. The main reason for this was probably that the economy needed some time to adjust to the change in the monetary policy regime in March 2001. Since the model selection analysis gives no clear suggestion of which model to choose, both models are treated on an equal basis. For given evolution paths of the exogenous variables, ex ante forecasts for the period 2004.3-2007.4 are obtained by the two models. In addition to this forecasts produced by a simple AR(8) model are also considered. The forecasts produced by the two competitive models that are developed, suggest that the demand for real cash will rise for the next couple of years, and then slowly decrease from the end of 2006. In contrast to this the simple AR(8) model suggests that the demand for cash will increase throughout the whole forecasts period. In order to make the forecasts more robust for potential breaks, forecasts with intercept correction are conducted. Finally, I produce forecasts under alternative assumptions of the evolution paths of the exogenous variables.