©2005 All content on site protected by copyright
War and Peace
The paper War and the World Economy: Stock Market Reactions to International Conflicts, by Gerald Schneider and Vera Troeger, is an honest and yeomanlike study of the influence of hot and cold wars on stock markets, and is well worth study.
The paper uses a daily index of the increase in tension caused by events in the Iraq wars, the Israel/Palestine conflict and the Yugoslavia wars as a foundation for differentiating between two hypotheses:
The authors conclude that markets react negatively to increases in the likelihood and intensity of war. They contend that "international traders only welcome conflictive events whose anticipated costs lift the uncertainty over the future course of action and promise a less costly resolution of the conflict than originally anticipated." They also conclude that conflicts increase the volatility of stock markets in an asymmetric way, with increases in tension increasing the volatility more than positive developments decrease it.
There are many defects in the paper. Foremost is the use of parametric rather than non-parametric statistical tests to carry their points. Also, the authors don't discuss the magnitude of the reduction in uncertainty or prediction error of the stock market that their model provides. Nor do they take account of the multiplicity of their retrospectively-concocted hypotheses that seem to fit the data. A more direct and convincing test could have been performed with a Siegel-Tukey test. Furthermore, the lumpiness of their data relative to the exact timing of the events undercuts the utility of their analysis.
This is one of those papers where the road is much better than the inn, i.e. the path taken is better than the conclusions. Many readers will be introduced to a class of models developed by Jean-Michel Zakoian in Threshold Heteroskedastic Models, Journal of Economic Dynamics and Control, 1994, which model volatility as a function of the the surprise and the over/underestimates in the previous period, with a dummy variable to differentiate between the overestimates, which cause increased volatility, and the underestimates, i.e. the positive surprises, that have no impact. Such a model could be much more meaningfully tested by enumerating the 10 biggest surprise events that seem to account for the only meaningful reactions that the market had during the period, and seeing whether the negative events affected variance more than the positive ones. The use of headlines from Dow Jones News or the New York Times would have been a useful adjunct here. Nevertheless, the methods used would find a much more tractable and useful set of data involving the reaction of individual company news to cardinal events like earnings announcements or the reaction of markets to economic reports such as the month employment reports.
I learned many things from this paper:
In short, although none of the conclusions of the Schneider and Troeger paper are carried by their data and techniques, there are many interesting facts and approaches outlined that have applicability and utility in many fields of specinvestment warfare and research.