Too-connected versus too-big-to-fail: banks’ network centrality and overnight interest rates

What influences banks’ borrowing costs in the unsecured money market? The objective of this paper is to test whether measures of centrality, quantifying network effects due to interactions among banks in the market, can help explain heterogeneous patterns in the interest rates paid to borrow unsecured funds once bank size and other bank and market factors that affect the overnight segment are controlled for. Preliminary evidence shows that large banks borrow on average at better rates compared to smaller institutions, both before and after the start of the financial crisis. Nonetheless, controlling for size, centrality measures can capture part of the cross-sectional variation in overnight rates. More specifically: (1) Before the start of the crisis all the banks, independently of their size, profit from different forms of interconnectedness, but the economic size of the effect is small. Bank reputation and perceived credit riskiness are the most relevant factors to reduce average daily interest rates. Foreign banks borrow at a discount over Italian ones. (2) After August 2007 the impact of banks’ interconnectedness becomes larger but changes sign: the “reward” stemming from a higher centrality becomes a “punishment”, which possibly reflects market discipline. Bank reputation becomes even more important. (3) After Lehman’s bankruptcy the effect of centrality on the spread maintains the same sign as after August 2007, but the magnitude increases remarkably. Foreign banks borrow at a relevant premium over Italian ones; reputation becomes outstandingly more important than in normal times.