Weak instruments in estimating business cycle effects on banks' interest income

Abstract

This article explores the link between the real business cycle and core bank earnings. Using bank-level data and an estimation technique which corrects for weak instruments, evidence confirms that pre-provision Net Interest Income (NII) is determined by the term structure of interest rates rather than output fluctuations. Output growth is only found to be significant when Loan-Loss Provisions (LLP) are taken into account.

Keywords:

• net interest income
• system-GMM
• weak instruments
• business cycle

JEL Classification:

• C33
• E32
• G21

Acknowledgements

We thank Leonardo Gambacorta, Bank for International Settlements, for his helpful comments.

The views expressed in this article are those of the authors and do not necessarily reflect those of the European Central Bank or its members.

All errors remain our own.

Notes

¹NII is typically defined as interest income minus interest expenses.

²An alternative specification based on NII per interest earning asset was considered to take account of the potential impact of mergers and acquisitions activity on the sample. The results, available from the authors on request, were remarkably similar to those displayed in Table 1.

Table 1. Determinants of NII

³In their model, Albertazzi and Gambacorta (2009) also control for lending growth and stock market volatility.

⁴OECD Financial Indicators (Main Economic Indicators) taken from OECD StatExtracts: http://stats.oecd.org/index.aspx.

⁵Those countries were Austria, Belgium, Luxembourg, France, Germany, Ireland, Italy, the Netherlands, Portugal and Spain.

⁶More specifically, only banks with (1) a Bankscope C*, C1 or C2 consolidated report and (2) which did not have a single ultimate owner of more than 50% according to the BvDEP Independence Indicator were used. Banks with a BvDEP Independence Indicator of C or D were, therefore, excluded.

⁷The change to international accounting standards was not deemed problematic, as accounting practices for NII do not vary significantly. Furthermore, the use of a dummy variable in Equation 1 controls for different standards.

⁸Two-step estimation controls for panel-specific autocorrelation and heteroskedasticity. SEs were Windmeijer corrected to control for downward bias in this context.

⁹On the basis of the Difference-GMM results shown in Table 1, short-term interest rates were also found to be statistically insignificant.

¹⁰The point estimate of 0.18 on the AR coefficient, obtained from the one-step System-GMM estimator, implies that a unit increase in long-term interest rates increases NII in the same year by 18%. The long-run effect calculated from is even higher at 34.1%. For the short-term interest rate, a unit decrease increases NII by 8.1%, with a long-run impact of 15.3%.

¹¹A better measure of this impact could be obtained by using NII adjusted for net charge-offs. Data limitations rendered this impractical, however, as it would reduce the data sample by approximately 75%.