In this paper, we investigate whether government investment crowds out or crowds in private investment for Fiji over the period 1950–2001. We begin by searching endogenously for break points in the data series using the Zivot and Andrews [J. Business Economic Stat. 10 (1992) 251–270] test. Upon finding that 1975 is the statistically significant break date, we divide the sample into two. Using the error correction mechanism test, we find that government and private investments are cointegrated over the period 1950–1975, but not for the period 1976–2001. We also find that in the former period government investment has crowded in private investment, while in the latter period the relationship between government and private investments has been statistically weak.