This paper analyzes employment fluctuations in the Brazilian states vis-à-vis the aggregate employment for the country as a whole. The objective is to verify whether it is possible to establish a long run relationship between state employment and national employment (Blanchard & Katz, 1992; Martin, 1997). The paper applies the traditional cointegration analysis methodology (Engle & Granger, 1987) and also the unrestrict error correction model, as proposed by Pesaran et alii (1996). The results lend support to the general hypothesis which states that both state and national employment levels fluctuate along a common trend, but with permanent differentials in the long run.