اظهار نظر در مورد جانسون و سوئنون (2004): بازار سهام ایالات متحده و ارزش های بین المللی دلار آمریکا

Is there a relationship between the performance of US equity markets and the value of the US dollar? The question is of practical and regulatory significance. Previous attempts to answer the question relied upon on the lagged-causality approach of Clive Granger and his coauthors. Given that financial markets are efficient, most of the correlation would be undetectable by such methods. In groundbreaking work, Johnson and Soenen (2004) used an estimator by Geweke (1982) that allows for contemporaneous or instantaneous effects, and found that there was always and everywhere an instantaneous link between the US equity and currency markets. Given the importance of Johnson and Soenen's results, we attempted to replicate their study. We argue that Johnson and Soenen's results hinge on a simple substitution error. After recalculation, we find little evidence of consistent instantaneous correlation between returns in the US equity markets and the value of the US dollar.

What is the relationship between the US stock market and the value of the dollar? This question is important and relevant to investors and policy-makers alike. Answering this simple question, however, is complicated by the fact that financial markets are efficient so that much of the relationship between these two variables may be contemporaneous, and thereby rendered untestable using traditional methods (such as VARs) which require lagged causality. Financial markets, however, are efficient so the correlation between equity markets and exchange rates is likely to escape detection by models using the traditional approach. To circumvent this problem, Johnson and Soenen (2004) use an estimation procedure developed by Geweke (1982) that allows for instantaneous and lagged correlation between variables. Geweke's procedure provides a measure of the size (but not the sign) of the instantaneous correlation between two variables. Johnson and Soenen (henceforth, JS) attempt to implement Geweke's procedure and conclude that “using Geweke measures of feedback, we find a high percentage (92% at the 1% level) of contemporaneous association between daily movements in the S&P 500 index and changes in the value of the dollar against seven foreign currencies” (p. 480). We argue that these seemingly powerful results stem from a simple substitution error. Below, we identify the error and recalculate the instantaneous correlation measures indirectly from Johnson and Soenen's tables, as well as directly from US financial data. In contrast to Johnson and Soenen, we find little evidence of instantaneous correlation between returns in the S&P 500 and the relative value of the US dollar. This paper is outlined as follows. First, we provide a mapping of notations between JS and Geweke. The models used by JS and Geweke are compared in the second section. Statistics for the contemporaneous feedback measure and details of the error are provided in the third section. Revised results are provided in section four.

The relationship between equity market performance and the value of the currency is of obvious relevance to investors, policy makers, and academics alike. Disentangling the relationship has proven difficult in the past because of the reverse and instantaneous causality between these two variables. Johnson and Soenen's original findings seemed to provide powerful and incontrovertible evidence that US equity markets are always and everywhere related to the value of the dollar. Johnson and Soenen (2004, pp. 473–474) report that of the 224 Geweke contemporaneous feedback measures in [their] Table 1 (7 exchange rates * 32 years)… 92% are significant at the 1% level… providing strong empirical evidence that daily movements in the S&P 500 coincide with changes in the value of the US dollar against the seven foreign currencies [on the] same day. Johnson and Soenen are correct to apply Geweke's novel approach to this interesting problem; however, their numerical calculations do not withstand closer inspection. When estimated correctly, we find that only 52 of 224 (23%) instantaneous feedback measures are significant at the 1% level, with most of these clustering in several years. The conclusion is that there is only weak evidence for instantaneous, or any other, causality between the S&P and the value of the US dollar. Whenever there does appear to be a statistically significant relationship with one country's currency (say, the Swiss Franc/US dollar) and the S&P, there is almost aways also one in the same year for another large country's exchange rate (the British Pound or the Japanese Yen, for example). It seems likely that the correlation between equities and exchange rates is due to shocks at the global level. Thus, we recommend that future research focus on discovering confounding variables that influence these seven countries simultaneously. We leave this for future research.