The monetary model of the exchange rate and equities: an ARDL bounds testing approach

Abstract

This study examines a version of the monetary model of the exchange rate, which incorporates a stock price measure. Using the ARDL Bounds testing approach, we produce evidence of cointegration, well-specified ECMs and forecasts that outperform a random walk.

Acknowledgements

I would like to thank Simon Broome and an anonymous referee for providing helpful comments and suggestions. The standard disclaimer applies.

Notes

¹ An advantage of the ARDL bounds testing approach is that the critical values produced by Pesaran et al. (2001) allow for the inclusion of a mix of I(0) and I(1) variables in the cointegrating relationship. Although the Johansen ML approach can also be used with a mixture of I(0) and I(1) variables, Rahbek and Mosconi (1999) suggest that including I(0) series in a VECM can produce nuisance parameters in the asymptotic distribution of the trace for the cointegration rank. See Wickens (1996) for further concerns over the Johansen ML procedure and Tawadros (2001) for a discussion of the difficulties in interpreting the results from this technique for the monetary model.

² As noted we assume that causality runs from stock prices to the exchange rate. For a justification of this assumption and a discussion on the causality between these variables see Bahmani-Oskooee and Sohrabian (1992), also see Smyth and Nandha (2003), for an alternative approach when using developing countries markets.

³ Regressive expectations are commonly used in exchange rate models, such as Dornbusch (1976), Frankel and Froot (1987) provide evidence to support their existence in the financial markets.

⁴ In other studies monthly data have been used, however this required the use of industrial production as a proxy for income, which in general is not as good as GDP, for which only quarterly data were available. Other measures of stock prices, such as dividend yield have also been linked with money demand (Friedman, 1956), however when incorporated into the model they did not perform as well as the main market indexes.

⁵ The UK and the USA were used as both countries have financial systems based around financial markets, rather than the banking sector as in Germany or France. As shown by Morley (2002), this has an important effect on this relationship. Stock market indexes are as follows: US: Standard and poor Composite index; UK: FTSE All Share Index. The following data was taken from the IFS: GDP–line 99b, treasury bill rate–line 60c and prices–line 64, money supply data was taken from the national accounts. Microfit 4.1 was used to carry out the regressions.

⁶ The tests were repeated including a trend in the tests; these results supported the findings of the non-trended tests.

⁷ The two models were also tested for the existence of a stable long-run relationship using other standard cointegration methods. These all supported the findings of a stable long-run relationship between the variables.

⁸ The conventional monetary model was also tested for the presence of a long-run relationship, but the unrestricted model indicated no evidence of cointegration, so further tests on the restrictions and forecasting was not possible.