Impact of global uncertainty on the global economy and large developed and developing economies

ABSTRACT

Global uncertainty shocks are associated with a sharp decline in global inflation, growth and interest rate. Global uncertainty shocks have more protracted, statistically significant and substantial effects on global growth, inflation and interest rate than U.S. uncertainty shocks. When controlling for domestic uncertainty, the decline in output following a rise in global uncertainty is statistically significant in each large country, except for the decline for China. For most economies, a positive shock to global uncertainty has a depressing effect on prices and official interest rates – exceptions are Brazil, Mexico and Russia, which are economies with large capital outflows during financial crises.

KEYWORDS:

• Global
• uncertainty shocks
• monetary policy
• FAVAR

JEL CLASSIFICATION:

• D80
• E44
• E66
• F62
• G10

Acknowledgments

We thank Baker Scott, Nicholas Bloom, Efrem Castelnuovo, Mardi Dungey, Enrique Martinez-Garcia and Francesco Ravazzolo as well as our discussant James Hansen for useful comments of earlier versions. We also thank members of Melbourne Institute, Centre for Applied Macro and Petroleum economics, Norges Bank and BI Norwegian Business School and seminar participants at CAMP-Melbourne Institute Applied Macroeconometrics Workshop (2015).

Disclosure statement

No potential conflict of interest was reported by the authors. The views in this paper are those of the authors and do not necessarily reflect the views of the Federal Reserve Bank of Dallas or the Federal Reserve System. Any errors or omissions are the responsibility of the authors.

Notes

¹ Note that Bloom, Bond, and Van Reenen (2007) show that share-return volatility is significantly correlated with alternative measures of uncertainty proxies.

² The methodology underlying the Global Economic Indicators (DGEI) database is provided in Grossman, Mack, and Martinez-Garcia (2014).

³ An important thread in the literature is that uncertainty faced by the individual firm is embodied in its own stock price volatility, as discussed in Leahy and Whited (1996), Bloom (2009), Bloom, Bond, and Van Reenen (2007) and Baum, Caglaynan, and Talavera (2010), among others.

⁴ This first principal component accounts for around 40% of the data variation.

⁵ We attempt to estimate this index for G20 economies. However, data for Indonesia, Iran, Thailand Nigeria and Poland were not available for the full sample period. An alternative measure of global uncertainty including these countries for a shorter span is discussed in section 8.6.

⁶ Data from the stock market are not available for all countries from 1981. The index is constructed with data on the countries for which data are available. A shortcoming of this approach is that for the earlier period, missing data are more apparent for developing countries. Nevertheless, we argue that this is not necessarily a problem, given that in the first part of the sample (1980–1995), the relative weight of developed economies in the global economy is more important than in the more recent period (following China’s unprecedented growth starting in mid-1990s). The availability of stock market data for each country is reported in Table A1 in online Appendix A.

⁷ Note that in Figures 1 and 2, the 12-month moving average indexes are shown to optically capture important global events. However, we follow Bloom (2009) in using the stock market volatility index in the VAR system without moving average transformation.

⁸ We follow Bloom (2009) and Jurado, Ludvigson, and Ng (2015) in setting p = 12, which allows for a potentially long-delay of effects of uncertainty shocks on the economy and for a sufficient number of lags to remove serial correlation.

⁹ We omitted the variables stock prices, wages, working hours and employment because these variables are not available at the global level.

¹⁰ We deal with missing data in early observations for some series by building the factors with series available at this time to maximize the number of time series observations.

¹¹ Note that the optimal lag selection for the FAVAR model according to the Bayesian information criterion (BIC) is 3 lags, while the Akaike information criterion (AIC) selects 6 lags. We also estimate the model using 12 lags following Bloom (2009).

¹² When the models are specified with 12 lags, the greatest response occurs after 6 months, with a quick return to positive values after 12 months.

¹³ Note that results are very similar when the coefficient $a_{44}$ is estimated (rather than set to zero), as well as when the order of U.S and global uncertainty are interchanged in separate estimations.

¹⁴ The starting period for these estimations starts later than 1981 for some countries due to data availability. In particular, the starting period for Brazil is October 1996, January 1994 for China, January 1994 for India, January 1997 for Russia and January 1990 for South Africa. For all other countries, the full period sample is available from January 1981 to October 2014.

¹⁵ The EMH predicts that prices on traded assets (and/or future prices) already reflect all past publicly available information. Consequently, the residual of this Equation can be interpreted as uncertainty using Jurado, Ludvigson, and Ng (2015)’s rationale.