Monetary policy and dividend growth in Germany: long-run structural modelling versus bounds testing approach

Abstract

This study examines the long-run relationship between monetary policy and dividend growth in Germany. For this purpose, cointegration is tested for between both variables in the period 1974 to 2003. However, problems related to spurious regression arise from the mixed order of integration of the series used, from mutual causation between the variables and from the lack of a long-run relationship among the variables of the model. These problems are addressed by applying the bounds testing approach to cointegration in addition to a more standard long-run structural modelling approach. In principle, both procedures are capable of dealing with the controversial issue of the exogeneity of monetary policy vis-à-vis dividend growth. However, the structural modelling approach still leaves a certain degree of uncertainty about the integration properties of the interest rate and the dividend growth. Hence, one feels legitimized to refer to the bounds testing procedure and to conclude that in the longer term short-term rates drive stock returns but not vice versa.

Acknowledgements

We would like to thank Mark Taylor and an anonymous referee for helpful comments and suggestions on an earlier version of this paper. We gratefully acknowledge the hospitality of the Oesterreichische Nationalbank (OeNB) where the first author was a visiting researcher while parts of this paper were written.

Notes

¹ For this kind of reasoning see, for instance, Bernanke and Gertler (2001), Bohl et al. (2003), Durham (2003), European Central Bank (2002), and Rigobon and Sack (2004).

² The regressions for dividend and profit growth are potentially subject to the omitted variables problem because, in this case, expected stock returns introduce noise. To circumvent this problem, the difference between h and Δd, h − Δd, were also calculated and used in the bounds testing procedure.

³ See Kaul (1996), p. 284.

⁴ Detailed proofs can be found in Pesaran and Shin (1999) and Pesaran et al. (1996, 2001).

⁵ The following estimations – like all other computations in this study – have been carried out using the 2001 version of the Microfit 4.11 package (see Pesaran and Pesaran, 1997).

⁶ For instance, monetary policy could have systematically and preemptively reacted to the emergence of asset price bubbles. More generally, asset prices as predictors of the future course of the economy might have triggered some monetary policy action. See, for instance, Bean (2004), Dupor and Conley (2004), European Central Bank (2002) and Robinson and Stone (2005) for good summaries of this discussion in the literature.

⁷ If a sample size is small, e.g. if observations take single-digit or low double-digit values, the relevant critical values potentially deviate substantially from the critical values reported in Pesaran et al. (2001). Therefore, exact critical value bounds have to be tailored to these sample sizes and are calculated for instance by Narayan and Smyth (2004a, b). Several previous studies have applied the ARDL approach to sample sizes smaller than the present study. Pattichis (1999) uses the ARDL approach to estimate a disaggregated import demand function for Cyprus based on annual data for 1975–1994 (20 observations). Tang (2001) employs the ARDL approach to model inflation in Malaysia using annual data for 1973–1997 (25 observations). Tang (2002) uses it to estimate a money demand function for Malaysia and annual data for 1973–1998 (26 observations), whereas Tang and Nair (2002) apply the approach to estimate an import demand function for Malaysia using annual data for 1970–1998 (29 observations).

⁸ For instance, Narayan and Smyth (2003), p. 1651, use a sample size comparable to the present study's and, exactly like it, refer to the critical values tabulated in Pesaran and Pesaran (1997), p. 478, Case II.

⁹ For a recent application of a Sen-type unit root test that allows for a simultaneous structural break in the intercept and slope see Narayan (2005). Since the ADF-tests on the first differences in the variables throughout lead to a rejection of the null hypothesis of non-stationarity, at least the variable i1m (but also Δd at even higher MA-orders) cannot a priori be excluded to be integrated of order one.

¹⁰ We are grateful for this important argument to an anonymous referee.

¹¹ However, these uncertainties that surround the integration properties of our stock return measure indicate that the Pesaran et al. (1996, 2001) approach – which does not rely on the exact identification of the order of integration of the underlying variables and has been applied in this study to test for cointegration between German monetary policy and stock returns – is more robust in the present context.

¹² This appears to be a quite important robustness check since Pesaran et al. (1996) admit that a shortcoming of their approach is that is not appropriate in situations in which there are more than one cointegrating vectors. See also Pattichis (1999).

¹³ See among others Bahmani-Oskooee and Ng (2002), Faria and Ledesma (2000), Halicioglu (2004), Morley (2003), and Payne (2003).

¹⁴ See Narayan and Smyth (2004), p. 5.

¹⁵ In contrast to the present study, Islam (2004), p. 997, finds diverging results of the bounds testing and the Johansen procedure in a study on the long-run relationship between openness and government size. Hence, it is interesting to cross-check the results based on the bounds testing procedure with the Johansen approach even if there is uncertainty on the orders of integration of the variables involved. See explicitly for this line of reasoning Islam (2004).

¹⁶ See Bean (2004), Dupor and Conley (2004), Domanski and Kremer (1998) and European Central Bank (2002).