Business cycle comovement in the G-7: common shocks or common transmission mechanisms?

Abstract

What are the sources of macroeconomic comovement among G-7 countries? Two main candidate explanations may be singled out: common shocks and common transmission mechanisms. In the article it is shown that they are complementary, rather than alternative, explanations. By means of a large-scale Factor Vector Autoregressive (FVAR) model, allowing for full economic and statistical identification of all global and idiosyncratic shocks, it is found that both common disturbances and common transmission mechanisms of global and country-specific shocks account for business cycle comovement in the G-7 countries. Moreover, spillover effects of foreign idiosyncratic disturbances seem to be a less important factor than the common transmission of global or domestic shocks in the determination of international macro-economic comovements.

Acknowledgements

The authors are grateful to two anonymous referees for constructive comments and to MIUR (PRIN project 2005) for financial support.

Notes

¹ See also Doyle and Faust (2002), Heathcote and Perri (2002), Helbling and Bayoumi (2003) and Monfort et al. (2003), for evidence of a reduction in G-7 business cycle syncronization over the most recent period.

² See for instance Stock and Watson (2003). See also, Justiniano and Primiceri (2006) and Fogli and Perri (2006) for recent contributions.

³ Kose et al. (2005) have found, for instance, evidence of a regional factor for the US and Canada. Also, Helbling and Bayoumi (2003) have found evidence of geographical clusters, pointing to two groups of countries, namely, the US, the UK, Canada and France, Italy, Germany, respectively. Moreover, Stock and Watson (2005b) point to the existence of a common Eurozone factor for the 1984 to 2003 period. Also, Bagliano and Morana (2006) have found that regional similarities seem to characterize more the real side of the economy than the nominal side. Finally, interesting regional similarities have been pointed out by Andreano and Savio (2007) concerning asymmetries in business cycle fluctuations and by Furceri and Karras (2007) concerning comovements within the Eurozone.

⁴ See Stock and Watson (2005a) for details on alternative identification strategies.

⁵ In particular, Bai (2003) considers the generalization of the principal components analysis to the case in which the series are weakly dependent processes, establishing consistency and asymptotic normality when both the unobserved factors and the idiosyncratic components show limited serial correlation and the latter also display heteroscedasticity in both their time-series and cross-sectional dimensions. In Bai (2002) consistency and asymptotic normality is derived in the case of I(1) unobserved factors and I(0) idiosyncratic components, also allowing for heteroscedasticity in both the time-series and cross-sectional dimensions of the latter component. Moreover, Bai and Ng (2004) have established consistency also for the case of I(1) idiosyncratic components. As pointed out by Bai and Ng (2004), consistent estimation should also be achieved by principal components techniques in the intermediate case of long-memory processes and Monte Carlo results reported in Morana (2007) support this conclusion.

⁶ In fact, what is denoted as global factor in Pesaran et al. (2004) is just a summary feature for all the variables which may have an impact on a given country, but for parsimony reasons are not modelled in detail. This is because when the unobserved component is estimated, the own country variables are neglected. However, it is hard, for instance, to justify the exclusion of US data when the global factors for the US are computed.

⁷ Nominal money balances are given by M2 for the US, M2+CD for Japan, M3 for the Euro area and Canada and M4 for the UK. The aggregates employed are the ones usually employed to measure broad money in each of the countries investigated.

⁸ The source of the Euro-area aggregate data is the European Central Bank. All other data are taken from Datastream.

⁹ For instance, the setting of the policy interest rate by the central bank renders the latter a step-wise deterministic process, inducing a nonlinear deterministic trend both in short- and long-term interest rates series.

¹⁰ Hence, the deterministic component included in the i-th equation of Equation 1 is specified as . Detailed results are not reported for reasons of space, but are available upon request from the authors. See also Bagliano and Morana (2006).

¹¹ See also Ehrmann et al. (2005) and Hamori (2000) for additional evidence in favour of the interpretation of US macroeconomic shocks in terms of global shocks. See also Harvey and Mills (2005) for additional evidence of comovements in G-7 macroeconomic variables.

¹² The real exchange rate changes (e) display little evidence of comovements: the fraction of the overall variance attributable to the first principal component amounts to 0.37 and is widely dispersed across regions (being heavily influenced by the US series). On this basis we conclude that there is no compelling evidence of a global factor driving real exchange rates.

¹³ The more idiosyncratic behaviour of the Japanese economy over the time span investigated is consistent with the very different macroeconomic conditions (economic stagnation) which have characterized this country, relative to the other economies, over the 1990s.

¹⁴ More detailed results of the first step of the analysis are reported in Bagliano and Morana (2006).

¹⁵ Evidence of serial correlation at the 1% level is detected only for the UK and US output growth rate equations. Significant ARCH effects are found for the UK output growth and short-term rate equations and for the Euro area long-term rate equation.

¹⁶ The median forecast error variance decomposition, as the median impulse response functions, have been obtained using Monte Carlo simulation, as suggested in Granger and Jeon (2004). For reasons of space, only the results for the within period and the 5-year period horizons have been reported in the tables. A full set of results is available from the authors upon request.

¹⁷ See Bagliano and Morana (2006) for additional details.

¹⁸ For reasons of space, plots are not reported. They are, however, available upon request from the authors.

¹⁹ SEs have been computed by simulation. The statistical significance has been evaluated at the 5% level.

²⁰ The finding of price and liquidity puzzles, given the large information set employed in the modelling, is quite surprising. The above puzzles are in fact usually related to misspecification of the information set and, for instance, the inclusion of commodity prices tends to lessen the problem. Yet, the evidence is coherent with previous results of Dees et al. (2007), where an even larger information set is used.

²¹ The robustness analysis, carried out by comparing the orthogonal impulse responses with the generalized impulse responses (Pesaran and Shin, 1998), in general, supports the above findings, particularly for the real output shock and the real effective exchange rate shock. In fact, the comparison allows to strengthen the interpretation of the former shock in terms of a productivity disturbance, since a negative correlation between real output and prices is found in all regions, apart from the Euro area. Moreover, the negative correlation between the exchange rate and output developments is also a robust finding, as well as the transmission of interest rate shocks along the term structure and the effects of short-term rate shocks on real output (with the only exception of the UK). Detailed results are available upon request from the authors.

²² For reasons of space, plots are not reported. They are, however, available upon request from the authors.

²³ In general, the analysis of the generalized impulse responses support the results obtained from the orthogonalized shocks, particularly as far as the foreign output shocks (apart from the effects on the exchange rate at the within quarter horizon), the foreign inflation shock (except for the effects on the domestic stock market in the long term), the foreign stock market shock (apart from the effects on the stock market). On the other hand, less robust results are found for the nominal money balance and interest rate shocks. Finally, the findings are in general robust also across countries, apart from Japan, for which, when the generalized shocks are employed, no reaction to foreign shocks is found in the short term.