Banks, Government Bonds, and Default

Recent events in Europe have illustrated how government defaults can jeopardize domestic bank stability. Growing concerns of public insolvency since 2010 caused great stress in the European banking sector, which was loaded with Euro-area debt (Andritzky (2012)). Problems were particularly severe for banks in troubled countries, which entered the crisis holding a sizable share of their assets in their governments’ bonds: roughly 5% in Portugal and Spain, 7% in Italy and 16% in Greece (2010 EU Stress Test). As sovereign spreads rose, moreover, these banks greatly increased their exposure to the bonds of their financially distressed governments (2011 EU Stress Test), leading to even greater fragility. As The Economist put it, “Europe’s troubled banks and broke governments are in a dangerous embrace.” These events are not unique to Europe: a similar relationship between sovereign defaults and the banking system has been at play also in earlier sovereign crises (IMF (2002)).

Despite the relevance of these phenomena, there is little systematic evidence on them. In our recent ECGI working paper, Banks, Government Bonds, and Default: What Do the Data Say?, we try to fill this gap by documenting the link between public default, bank bond holdings, and bank loans. We use the BANKSCOPE dataset, which provides us with information on the bond holdings and characteristics of over 20,000 banks in 191 countries and 20 sovereign default episodes between 1998 and 2012. We address two broad questions:

• 1. Does banks’ exposure to sovereign risk affect lending? In particular, do the banks that hold more public bonds exhibit a larger fall in loans when their government defaults?
• 2. Why do banks buy public bonds, becoming exposed to default risk in the first place? Is it because they buy bonds in normal times and are then surprised by crises, or is it because they actively chase risk during crises?

The goal of our analysis is to document robust stylized facts and not to identify causal patterns, which our data does not allow us to do. Our main findings are as follows:

• Holdings of public bonds are large in normal times, particularly for banks that make fewer loans and are located in financially less developed countries. In non-defaulting countries, banks hold on average 9% of their assets in public bonds. Among countries that default at least once (which are financially less developed), average bank bond holdings in non-default years are 13.5%. In both groups of countries, bondholdings in non-default years are decreasing in bonds’ expected return.
• During default years, average bond holdings increase from 13.5% to 14.5% of bank assets. Critically, this increase is concentrated in large banks. Moreover, during default years, bond holdings are increasing in bonds’ expected return.
• During sovereign defaults, there is a large, negative and statistically significant correlation between banks’ bond holdings and subsequent lending activity. A one dollar increase in bonds is associated with a 0.60 dollar decrease in bank loans during defaults. Strikingly, about 90% of this decline is accounted for by the average bonds held by banks before the default takes place; only 10% of this decline is explained by the additional bonds bought in the run-up to and during default.

Our results support the notion that banks’ holdings of public bonds are an important transmission mechanism of sovereign defaults to bank lending. These findings are broadly consistent with the following narrative. Public bonds are very liquid assets (e.g., Holmstrom and Tirole (1998)) that play a crucial role in banks’ everyday activities, like storing funds, posting collateral, or maintaining a cushion of safe assets (Bolton and Jeanne (2012), Gennaioli, Martin, and Rossi (2014)). Because of this, banks hold a sizable amount of government bonds in the course of their regular business activity, especially in less financially developed countries where there are fewer alternatives. When default strikes, banks experience losses on their public bonds and subsequently decrease their lending. During default episodes, moreover, some banks deliberately hold on to their risky public bonds while others accumulate even more bonds. This behavior could reflect banks’ reaching for yield (Acharya and Steffen (2013)), or it could be their response to government moral suasion or bailout guarantees (Livshits and Schoors (2009), Broner et al. (2014)). Whatever its origin, this behavior is largely concentrated in a set of large banks and is associated with a further decrease in bank lending.

One important feature of our dataset is that it covers a wide sample of default and non-default years. Because of this, it allows us to evaluate the relative contribution of bond holdings accumulated before and during sovereign defaults to the transmission of such events to private lending. The data provide a rather clear result: in the countries and periods that we consider, average bond holdings in non-default years, which reflect banks’ normal business activity, play a significantly larger role than bonds accumulated in the run-up to and during default years. The main driver for this result is that, in our sample of defaulting countries, banks hold many bonds in normal times (13.0% of assets) and the average increase in bond holdings during crises is rather small by comparison (less than 2% of bank assets).

These results provide a new perspective on the mechanisms whereby the sovereign default-banking crisis nexus comes into existence and operates. Much of the recent work on this nexus has focused on the role of risk-taking by banks during crises. Although this may well be the right strategy for the European context, our panoramic view of sovereign debt crises calls for paying close attention also to the bonds held by banks in normal times. This insight has both positive and, potentially, normative implications.

From a positive standpoint, our analysis suggests that the unfolding of sovereign crises is fundamentally different in emerging and advanced economies. In the latter, the accumulation of bonds during crises is larger relative to total bond holdings, and is therefore likely to be responsible for a larger portion of the adverse costs of defaults. From a normative standpoint, our results suggest that caution may be warranted in modifying bank regulation to apply higher risk weights to government bonds. If banks demand a sizable amount of government bonds to carry out their normal business activities, as seems to be particularly the case in emerging economies, sovereign defaults will undermine the functioning of the banking sector and bank lending over and above its risk taking during the crisis itself. In this context, proposed regulations to increase the risk weight of government bonds during sovereign crises may backfire, because they might exacerbate the pro-cyclicality of bank balance sheets without having much of an effect on the link between sovereign risk and the banking sector.

The full paper is available for download here.