Demand for money in India: 1953–2003

Abstract

The demand for money, especially in the developing countries, is an important relationship for formulating appropriate monetary policy and targeting monetary variables. In this paper the demand for narrow money in India is estimated and its robustness evaluated. It is found that there is a stable demand for money for almost half a century from 1953 to 2003. There is no evidence for any significant effects of the 1991 financial reforms.

Acknowledgements

The authors are grateful to Professor Ashok Parikh for help with some computations and conceptual clarifications. Also they thank a referee and the editor of this journal, Professor Mark Taylor, for many useful suggestions. However, errors in this paper are the authors responsibility.

Notes

¹ Equation 1 is derived from an equilibrium relationship:

augmenting with a partial adjustment equation:

where λ is the speed of adjustment. See Cuthbertson (1988) and Taylor (1994).

² Attention is drawn to recent publications with the real rate of interest. An early study by the IMF and a recent one by Jayaraman and Ward (2000) have used the real rate of interest in the demand for money for Fiji. Ahmed (2001) has also used the real rate for Bangladesh. Perhaps there are several other empirical works in the developing countries with similar weaknesses in the demand for money.

³ It is well known by now that Hendry and Mizon have popularized the London School of Economics approach to modelling dynamic equations with the general to specific approach (GETS). GETS is a pragmatic and flexible solution (compared to PAM) to reconcile a methodological conflict between the equilibrium nature of theoretical relationships and the data from the real world that is seldom in a state of equilibrium. Although the LSE-GETS approach had a mixed reception, there is a renewed interest due to the development of an automated model selection software, PcGets, by Hendry and Krolzig (2001) and the seminal contributions by Hoover and Perez (1999, 2004). See also Rao and Singh (2005) on PcGets.

⁴ Some early studies based on the unit roots and cointegration approach to demand for money in India are: Nag and Upadhyay (1993), Ghatak and Ghatak (1994), Parikh (1994) and Rao and Shalabh (1995). However, Vasudevan (1977) and Bhoi (1992) have used equations based on the PAM.

⁵ This test is essentially testing the restriction that the coefficients of the lagged values of ln(M/P) in the block of equations explaining the variables ln Y and i are zero.

⁶ The maximal eigenvalue and trace test statistics for the null that there is no cointegration are 25.2268 and 34.5347 respectively. The 95% critical values, respectively, are 22.16 and 30.77. For the null that there is one cointegrating vector, the corresponding computed values, with the critical values in the parentheses, are: 9.3079 (15.44) and 9.3079 (15.44) respectively.

⁷Estimation of the cointegrating equations for the pre- and post-1979 periods was also tried, but found that it is not possible to obtain a satisfactory cointegrating relationship for the first period.