Demand for money in the selected OECD countries: a time series panel data approach and structural breaks

Abstract

Time series panel data estimation methods are used to estimate the cointegrating equations for the demand for money (M1) for a panel of 11 Organization for Economic Cooperation and Development (OECD) countries for which consistent quarterly data are available. The effects of financial reforms are analysed with structural break tests and estimates for alternative sub-samples. Our results for the post-reform sub-samples show that the income elasticity of the demand for money has decreased and response to interest rate changes has increased.

Keywords:

• demand for money
• Pedroni method
• financial reforms
• Westerlund method

JEL Classification::

• C50
• E41

Notes

¹ Some studies support the Taylor rule-based interest rate targeting. See Orden and Fisher (1993) for Australia, McPhail (1991) and Haug (1999) for Canada, Nagayasu (2003) and Maki and Kitasaka (2006) for Japan, Papadopoulos and Zis (1997) for Greece, Vega (1995, 1998) for Spain, Oxley (1983) and Caporale and Gil-Alana (2005) for UK and Breuer and Lippert (1996) for USA.

² Following the advanced countries, central banks in many developing countries have also switched to targeting the rate of interest without any credible evidence that their money demand functions have become unstable. This shift has been questioned by Bahmani-Oskooee and Rehman (2005), Bahmani-Oskooee and Gelan (2009), Rao and Kumar (2009), Sumner (2009) and Yu and Gan (2009). These authors have found no instability in the demand for money functions using alternative estimation methods and data mainly from the Asian countries.

³ Nielsen (2004) examined the UK money demand (M2, M3 and M4) for the period 1873 to 2001 and stability results suggest that the long-run demand for money is stable. According to Friedman and Kuttner (1992), the US demand for M1 is cointegrated for the period 1960 to 1979, but becomes unstable if the sample is extended from 1980. Ball's (2001) study for the US, notes that stability tests did not show breaks in the demand for M1 up to 1987, but becomes unstable when the samples are extended up to 1996.

⁴ The cross-section and time series studies are well-surveyed earlier by Sriram (1999, 2001). Recently, Duca and VanHoose (2004) provided a useful survey of the theoretical and empirical literature on the demand for money.

⁵ These EU countries are Cyprus, Czech Republic, Estonia, Hungary, Latvia, Lithuania, Malta, Poland, Slovak Republic and Slovenia.

⁶ These countries are Austria, Belgium, Finland, France, Germany, Ireland, Italy, Luxembourg, Netherlands, Portugal and Spain.

⁷ The International Financial Statistics (IFS) and the World Bank database did not publish data on M1 after 1998 for some major OECD countries (e.g. France, Germany, Greece, Ireland, Netherlands and the UK) and this has constrained our selection of countries. Furthermore, for compatibility we have used the IFS and World Bank sources for data.

⁸ Estimates of the individual country cointegrating parameters are not reported to conserve space but are available from authors upon request. FMOLS estimates are without common trends. DOLS estimates are similar and not reported to conserve space.

⁹ Deleting the inflation rate did not make any difference to the coefficients of other variables. However, deleting this variable made the cointegration tests weaker. Therefore, we retained this variable and proceed to test and estimate the cointegrating equations for sub-samples for Equations 1 and 3 only.

¹⁰ Although it is also possible to test for multiple breaks, we decided to test for one dominant break because our data covers only a shorter period of about 30 years and after some immediate and initial effects of reforms might have taken place, i.e. from 1975 due to gaps in data. Multiple breaks may also give conflicting break dates and increase the number of sub-samples. Therefore, testing for a single dominant break is a pragmatic option.

¹¹ We thank a referee of this journal for asking us to make this point clear.

¹² Hadri tests provide support that inflation rate is I(1); the sub-sample unit root tests are not reported to conserve space.