Macro News And Commodity Returns
Brunel University – Centre for Empirical Finance
Brunel University London – Economics and Finance
October 13, 2015
This paper adopts a VAR-GARCH approach to model the dynamic linkages between both the mean and the variance of macro news and commodity returns (Gold, Corn, Wheat, Soybeans, Silver, Platinum, Palladium, Copper, Aluminium and Crude Oil) over the period 01/01/2001-26/09/2014. The chosen specification also controls for the effect of the exchange rate. The results can be summarised as follows. Mean spillovers running from news to commodity returns are positive with the exception of Gold and Silver. Volatility spillovers are bigger in size and affect most commodity returns. Both firstand second moment linkages are stronger in the post-September 2008 period. Overall, our findings confirm that commodities, despite not being financial assets, are sensitive to macro news (especially their volatility), and also suggest that the global financial crisis has strengthened such linkages.
Macro News And Commodity Returns – Introduction
The existing literature on the effects of macro news mainly focuses on the stock and bond markets, and typically considers two sources of news effects: scheduled macroeconomic announcements that do not correspond to agents’ expectations (the announcement effect) and unscheduled announcements (the surprise effect). Most studies analyse the former, calculating the difference between news releases and their expected value, and then defining positive and negative news accordingly (see Kocenda and Hanousek, 2011, and Hanousek, et al., 2009). Stock prices have been shown to be affected by news about monetary variables such as money growth and interest rates (see, e.g., Chen, 1991; Cornell, 1983; Pearce and Roley, 1983, 1985), and in some cases also by real sector news (see, e.g., McQueen and Roley, 1993, and Boyd et al., 2005). Birz and Lott (2013) use newspaper headlines, and also find that news on GDP and unemployment affect stock returns. Caporale et al. (2014a) consider both mean and volatility spillovers in the case of the euro area.
Various studies have also been carried out for bond markets. For instance, Gurkaynak et al. (2005) show that long-term interest rates respond to the unexpected component of macro and monetary news releases, Balduzzi et al. (2001) and Andersen et al. (2005) find effects on US Treasury bond futures contracts, and Brenner et al. (2009) on bond return volatility. Beetsma et al. (2013) examine the impact of news on interest rate spreads vis-à-vis Germany in various countries in the euro area, and Caporale et al. (2014b) provide evidence of dynamic linkages in both the first and second moments.
Fewer studies have examined the effects of macro news on commodity prices. Despite not being financial assets, the latter have been shown to be affected by variables such as interest rates (Frankel, 2008) and the US dollar exchange rate, both of which are known to respond to news announcements. Frankel and Hardouvelis (1985) provide evidence of a statistically significant response to US money supply announcements; effects of macro news on various commodity prices are also found by Cai et al. (2001), Hess et al. (2008), Kilian and Vega (2008); commodities futures prices have been reported to be affected as well (Barnhart, 1989; Ghura, 1990). Roache and Rossi (2010) in particular show that they are influenced by the surprise element in macro news, with evidence of a pro-cyclical bias after controlling for the effects of the US dollar, the only exception being gold, which reacts counter-cyclically given its role as a safe heaven and store of value, and is more sensitive to bad news and higher uncertainty. Unlike most other authors, typically using OLS, they estimates a GARCH(1,1) model given the evidence of time variation and clustering of volatilities (Cai et al., 2001, is
another of the few papers using a GARCH framework, specifically to examine the impact of news on gold futures prices).
Some recent literature focuses on investor psychology to explain the relationship between news and financial markets. For instance, De Long et al. (1990) distinguish between two categories of traders: rational arbitrageurs updating their Bayesian beliefs on the basis of economic fundamentals, and noise traders with random beliefs. In their model, because of risk aversion and other constraints for investors, low sentiment has a (temporary) negative effect on prices but increases volume, as noise traders react to negative belief shocks by selling shares to rational arbitrageurs (see also Campbell et al., 1993). Coval and Shumway (2001) and Antweiler and Frank (2004) instead relate investor sentiment to trading costs, with the perception of a more negative outlook resulting in lower trading volumes.
Tetlock (2007) examines the links between media “pessimism” (generated by “bad news”) and low investor sentiment in the US by estimating a VAR model. The former could be interpreted as a proxy for either investor sentiment or risk aversion, in which case pessimism should increase volume, or for trading costs, implying that pessimism should decrease volume. Also, pessimism could either forecast future or reflect past sentiment, and be due to negative information about asset prices not already incorporated in them or about dividends already reflected in them, with different implications for price behavior. The empirical evidence suggests that models of noise and liquidity traders can account for the effects of low investor sentiment on financial markets (see also Tetlock et al., 2008, for further results). 1 Fang and Peress (2009) use a wider dataset including more US daily newspapers and a cross-section of countries and find that media coverage can increase the degree of recognition and therefore the corresponding returns on stocks only recognized by a few agents and consequently not sufficiently diversified; therefore it affects asset prices by disseminating information broadly, even if it does not represent news (see Merton, 1987).
The present paper adopts a VAR-GARCH approach to model the dynamic linkages between both the mean and the variance of macro news and commodity returns. This is in contrast to the vast majority of earlier contributions, which only examined level effects. Analysing simultaneously the interactions between the first and second moments sheds new light on the issues of interest. The layout of the paper is the following. Section 2 outlines the econometric modelling approach. Section 3 describes the data and presents the empirical findings. Section 4 summarizes the main findings and offers some concluding remarks.
See full PDF below.