We investigate the phase difference of the sunspot cycles in the two hemispheres and compare it with the latitudinal sunspot distribution. If the north-south phase difference exhibits a long-term tendency, it should not be regarded as a stochastic phenomenon. We use datasets of historical sunspot records and drawings made by Staudacher, Hamilton, Gimingham, Carrington, Spouml;rer, and Greenwich observers, as well as the sunspot activity during the Maunder minimum reconstructed by Ribes and Nesme-Ribes. We employ cross-recurrence plots to analyse north-south phase differences. We show that during the last 300 years, the persistence of phase-leading in one of the hemispheres exhibits a secular variation. Changes from one hemisphere to the other leading in phase were registered near 1928 and 1968 as well as two historical ones near 1783 and 1875. A long-term anticorrelation between the hemispheric phase differences in the sunspot cycles and the latitudinal distribution of sunspots was traced since 1750.