Comments on

1
Comments on Bank Risk Taking and Competition
Revisited New Theory and New Evidence, by Boyd,
De Nicolo, and Al Jalal

• Discussant
• Bilal Zia
• World Bank

2
Quick Summary

• This paper
• Presents contrasting models that relate banking
  competition to stability
• Tests the relationship empirically
• Theory
• CVH (Charter value hypothesis) Competition
  reduces bank profits and lowers franchise value
  ? lower incentives for good loans (higher MH) ?
  correlation between bank concentration and risk
  of failure is negative
• BDN (Optimal contracting) Int. rates directly
  influence firm project choice (MH) ? correlation
  between bank concentration and risk of failure is
  positive
• Empirics
• BDN trumps CVH

3

• Theoretical part of paper is strong and well
  presented
• Focus here on empirical section
• Finding High concentration is positively
  related with risk
• Measure of concentration Hirschman-Hirfendahl
  Index
• Measure of risk Z-score (ROA EQA) / sd(ROA)
• A host of controls

4
Comments

• Cross-sectional data from US rural banks and
  Bankscope panel from developing countries two
  very different datasets
• US data
• Banks with branches in multiple counties
  omitted, but is bank activity concentrated to
  within-county clients only? Dont banks compete
  across neighboring counties? (HHI is at county
  level)
• Are there different bank types (such as banks
  that serve specific rural-finance functions)? If
  so, incentive structure to lend may be different
  (state mandated? Perhaps not in US)
• Does calculating market size using mean rather
  than median change results? (Can winsorize and
  then take mean if worried about outliers)

5
Comments

• Using state FEs is useful but regression is at
  county level, so should do county FEs. (There is
  footnote in paper, but should run these as well).
  
• Can you try putting in size controls
  non-parametrically (size deciles)? LHS is loans /
  assets and size control is log(assets), so a
  mechanical relationship exists.
• Surprisingly large (and significant) coefficient
  on Farm coefficient. Seasonality?
• IV need to show first stage. Not immediately
  obvious why state dummy is a good instrument for
  bank size?

6
Comments

• International Sample
• 2700 banks from 134 developing countries
• Major issue No mention of differences between
  government owned, private, and foreign banks.
  Large literature on politicization of loans
  through government banks. Could large govt. banks
  be driving results?
• Does sample include ALL banks, or only banks
  above a certain size? HHI measurement will be
  affected.
• Need a better control for demand conditions.
  GDP per capita is very imprecise. Do you have any
  measures of business development?
• Need to show results using relative size of banks
  rather than absolute size across countries.