This paper re-examines the relationship between U.S. defense spending
and the dollar-mark exchange rate during the 1960:1 to 1990:4 period.
Previous work, under the guise of the ''safe haven'' argument, finds a
cointegrating relationship between real military expenditures, real o
utput and the real dollar-mark exchange rate. Here, we examine the dyn
amic time series properties of a cointegrated vector autoregression un
der a more traditional real exchange rate specification. Through vario
us restrictions placed upon the estimated cointegrating vector, we tes
t specific hypotheses about the long-run relationships between the var
iables. Our results reveal that previous work suggesting that military
expenditures may significantly affect the real dollar exchange rate v
alue may be misleading.