This paper investigates the determinants of long-term bank relationships using a new data set for the time period between 1998 and 2008, and comprised of 9,476 firms in Norway. We document that firms are more likely to end a relationship as the relationship matures. This result casts doubt on theories suggesting that firms become locked in by their banks. Further, looking at firm specific variables we find that old, highly profitable firms with high liquidity and high creditor concentration maintain longer relationships. We extend our study to bank- and relationship specific effects and report that relationships with savings banks are longer, as are relationships where the firm holds both deposits and loans in the bank. Our study is robust to left-censoring, alternate specification for the distribution of relationship duration, and control variables such as the bank market concentration in Norway and primary banks. Overall the main contribution of our study to the existing literature is that the unique data set enables us to study new explanatory variables on a relatively large sample.