Tests of convergence in a time series framework have historically been formulated as univariate tests of stationarity of the output differential. However, convergence is a multivariate hypothesis, so a multivariate framework is more natural for testing the hypothesis of convergence. One such framework is the cointegrated vector autoregression, with the Johansen procedure providing the statistical basis for tests of convergence. This approach yields novel results relative to those obtained from the univariate approach. For instance, the multivariate approach detects output convergence among France, Germany, and Italy over the last century, whereas the univariate approach fails to detect any such convergence.