Financial markets aren't ignoring geopolitics
Co-author: Lize Nauta
- Geopolitical risks have risen since the 2016 US elections, but markets appear unperturbed
- We construct a geopolitical risk index using Bloomberg news volume data, following the method of Caldara and Iacoviello (2018)
- Government bond yields, equity markets and some exchange rates show a statistically significant response to changes in our index
- With the important exception of US Treasuries, there is little evidence that the response to geopolitical risk has declined, and in the case of global equity markets it has increased
- However, the effect is apparently not large enough to dominate other factors
- Markets don’t appear complacent about geopolitical risks, but any effect should only be noticeable in the (uncomfortably realistic scenario) that geopolitical tensions rise further
Geopolitical tensions have increased in recent years resulting in more attention to international relations. Regardless, equity markets have climbed to record highs, equity volatility remains low and interest rates are on the rise. Are markets ignoring geopolitical risks?
In order to study the effects of geopolitical risk, Federal Reserve economists Caldara and Iacoviello construct an index by counting the number of news articles in major English language newspapers that mention keywords on topics like international conflict, war and terrorism. Using this index they find that heightened geopolitical risk weighs on economic growth, equity prices and interest rates in advanced economies. We used the same index to conclude that geopolitical risks have a significant negative effect on the Dutch economy.
In this Economic Report we take a closer look at the impact of geopolitical risks on various financial markets. Rather than use the Caldara and Iacoviello index, which is based on general interest newspapers, we construct our own index using Bloomberg news data. We run regressions of daily changes in different equity market indices, interest rates and exchange rates controlling for economic data surprises. The coefficient on the daily changes in the geopolitical risk index are generally statistically significant and stable enough over time to suggest that markets do care about geopolitical risk and continue to do so. The effects, however, are not strong enough to dominate other factors moving markets.
A geopolitical risk index using Bloomberg news
Caldara and Iacoviello (C&I) construct a geopolitical risk index based on the incidence of words related to geopolitical risk in US, Canadian and British newspaper articles. They include different combinations of relevant terms aimed at capturing the risk and incidence of conflicts of global importance. We refer the reader to their Working Paper and Iacoviello’s geopolitical risk index website for more information on their methodology and data. They do different types of robustness checks to validate their search terms and show the usefulness of their index in analysing the economic impact of geopolitical risk. They also do some financial market analysis, particularly for equity markets, at a monthly frequency.
We use C&I’s validated search terms to construct a similar index using Bloomberg news (see the appendix for list of keywords). Bloomberg news data has important advantages over newspaper data when it comes to analysing financial markets, especially at a daily frequency. First, Bloomberg is a real-time news service, so news is more likely to be reported on the day that it occurs whereas many newspapers may report the news the morning after. Second, Bloomberg is aimed at financial market professionals and more likely to focus on news relevant for financial markets and this news also more likely to reach actual market participants. Third, although C&I do share daily data, it is only available to the public with a delay. Statistical results shown in the appendix show that Bloomberg news data does indeed provide a better basis for the analysis of financial markets, particularly bond and equity markets.
The index we construct is based on the relative frequency of geopolitical news. That is, we use the news volume function in Bloomberg to count the number of articles with the search terms defined by C&I. Then we divide by the total volume of Bloomberg news on a given day.
As figure 1 shows, our Bloomberg-based geopolitical risk index shows similar patterns to C&I’s index, but they are not completely identical either. Both indices also capture important geopolitical events or periods of tension.
Geopolitical news continues to move markets
We examine the impact of daily shifts in geopolitical risk perception on various financial markets. To do this we regress daily movements in different financial markets on shocks in our Bloomberg-based geopolitical risk index (GPR-B), controlling for economic surprises. Our technical appendix contains details of the regression, alternative specifications and stability checks.
The history of economic surprises available means that our regressions samples run from 2002, at the earliest, to June 2018. The samples are shorter for the regressions using MSCI equity indices, namely since 2008.
We report the main results in the table below. Because the coefficient on the geopolitical risk index does not have an intuitive explanation, we show instead the coefficient multiplied by the largest daily change in the index (24 September 2002, in the run up to the Iraq War). We only show results that are statistically significant (p-value of lower than 10%). We also do stability tests to check if the coefficients of the regression are statistically different over time. When they are instable, we show the results of a regression for the period from January 2016 to June 2018.
Overall the results show a statistically significant relationship between geopolitical risk and financial markets, particularly for government bond yields and equity markets. The signs of these responses are as you would expect, higher risks leads investors to sell equities and buy sovereign bonds with low default risk, leading to lower equity prices, higher bonds prices and lower interest rates. The impact on exchange rates is less pronounced. Although, the results here support the view that the Japanese yen acts as a flight-to-safety currency, which appreciates when risk increases.
In most cases the relationship is stable over the sample. Markets do not appear to be responding differently to political risk than in the past. An important exception is the US Treasury yield, which over the last few years has not responded to moves in geopolitical risk, unlike other developed country government rates. It is not clear why this relationship has broken down. It could be that market participants fear an armed conflict would be very expensive and become a risk to US creditworthiness, now that US government debt has increased considerably. Another theory is that market participants might expect China top stop buying large amounts of Treasuries if geopolitical tensions with the US rise.
Another important exception is the MSCI world equity index, where we see that the global equity index has become more sensitive to political risk. We also examine if Dutch equities respond differently to geopolitical risks than global equities, and we find no evidence that they do.
Overall the impact of geopolitical risk is small in these estimates. We would be cautious in concluding that this means markets care very little about geopolitical risk. The daily swings in the news volume about geopolitical risk are large, “watering down” the effect identified and resulting in a smaller coefficient. Lower frequency data may better reflect news that is persistent, explaining why C&I find stronger effects. Nevertheless, even C&I’s effects are not so large that we would expect recent increases in geopolitical risk to be the dominant factor for financial markets. While lower frequency data could potentially offer better estimates of the scale of the impact of geopolitical risk, to answer the question whether markets are responding differently than in the past daily data are better suited. This is because we need to be able to split the sample and still have a sufficient number of observations for statistical inference.
Our analysis suggests that most financial markets aren’t responding differently to geopolitical risk than in the past. The lack of apparent response thus suggests that the impact of the current levels of geopolitical risk are simply not strong enough to overcome other factors driving markets. It does not suggest markets are complacent and could suddenly ‘come to their senses’ resulting in geopolitical risk rapidly being priced in. It also suggests, however, that if there is an even larger increase in geopolitical risks (which is all too plausible) there will be a corresponding response in markets with lower equity prices and interest rates as a result.
 We follow C&I in using the residuals of a ARIMA process to measure shocks in the financial markets variables. The intuition is that any change in the news index that can be predicted by its own history cannot be a surprise to the markets and shouldn’t move them. As we report in the appendix, we also examine several alternative specifications and generally find similar results.
 Here using daily data shows its value, as with weekly or monthly data it would be difficult to do regressions on such short periods.
 Interestingly, the C&I daily index does show a statistically significant response for major exchange rates, reflecting an appreciation of the dollar against other currencies. That while for interest rates and equity markets, the C&I index does not show statistically significant results while our Bloomberg variant does. See the appendix for details.