Discussion by Alistair Milne, Cass
Business School, 22 nd May 2008
• Panel data estimates: effect of monetary policy (Δ i t
( Δln (L) t
)
) on loan growth
– Euro-zone data (Bankscope), 1999-2005,
• Innovations/ features
1.
Inclusion of “expected default frequency” EDF
– as a level term
– interacted with interest rates
2. time varying bank specific variables (liquidity LIQ, log assets SIZE , capital asset ratio CAP)
– Measured relative to sample mean (so interaction terms sum to zero)
– Both as level terms and interacted with interest rates
– In line with (but also different from) literature on ‘bank lending channel’
• Findings
– All these levels and interaction terms highly statistically significant
– Coefficients match conventional bank lending view
• constrained banks (low LIQ, SIZE, CAP) respond more to monetary policy
– Lower EDF associated with rapid lending growth
• Interpretation less risky banks respond more to monetary policy
• This paper is highly topical and has a great motivation
• Since July 2007
– sharp drop in bank equity prices (higher EDF)
– evident constraints on bank funding/ loan growth
• So potential to reveal important lessons about the credit crunch
• Better relationship to existing literature
• Distinguishing bank loan demand from bank loan supply
• Improving econometrics
• Confusing literature
– Many papers, few relate clearly to each other
• Bank balance sheet characteristics may affect supply of intermediated credit
– Because of constrained access to wholesale funding markets
• Different aspects of this story
– Cross-sectional: some banks constrained
– Time series: access to wholesale funding varies over time.
• In my view original cross-sectional theory (Bernanke and
Blinder (1988), Stein (1988)) flawed
– Stein (personal correspondance) admits the theory is ambiguous,
• Too simple to say “results in line with bank lending channel”
– Literature says almost anything goes!
• The key empirical issue – must be able to distinguish loan demand from loan supply
• eg work of Kashyap and Stein (2000)
– Go to great lengths to argue that their results reflect differences in loan supply
• EDF is correlated with loan demand
– High loan growth associated with higher equity price and hence lower EDF
– You have to find instruments for EDF that are correlated with loan supply and not loan demand
• Other variables (SIZE, LIQ, CAP) may also be correlated with loan demand
• Essential to use time-country dummies
– In Euro area different response of loan demand to interest rates in different countries
– I think time country dummies may be only way to correct for this
• Model includes level variables
– Therefore must include bank fixed effects
• Unclear from draft if you do this
• Why are interaction results (SIZE, CAP, LIQ) so different from Ehrman et. al. ?
– Is this choice of data period? Please investigate