Transcript Stephen Millard

Discussion of
Long-run effects of idiosyncratic
uncertainty in a model with credit
market frictions
by Morozumi and Ormaechea
Stephen Millard
Bank of England
12 November 2010
1
Motivation
• Many commentators have looked at the role
of uncertainty in business cycles
• But less investigation of ‘long-run’ impact
on output and credit supply of uncertainty
• But important what you mean by ‘long run’
– More on this later
• Additionally, recent crisis saw a rise in
uncertainty, fall in credit and a fall in output
2
Key findings
• Increase in uncertainty leads to a long-run
fall in output and a long-run fall in the
credit-to-output ratio
• Can explain long-run correlation between
the two
– But what do we mean by long-run correlation?
• More later
3
Modelling Approach
• Entrepreneurs need to borrow to finance
working capital
– cf My paper with Michael McMahon and others
to be presented later
• Uncertainty about output (sales?) at time
loan is made leads to possibility of default
• This implies a risk premium on these
otherwise interest-free loans
4
Modelling Approach
• I’m not sure about interest-free loans
• Loans are ‘within period’ within the model
but surely this still implies some time?
• Banks typically have to fund their loans
from somewhere so I’d expect to see some
sort of funding cost
– Both aspects present in Fernandez-Corugedo et
al.
– But may not matter for the results of this paper?5
Modelling Approach
• Distortion arises from the ‘costly state
verification’ problem
– Key parameter is m, the monitoring cost
– As ever, this might be better thought of as
‘losses arising in bankruptcy’ costs
– Technical comment (1): Can banks seize a
defaulting firms capital? They should do.
• Is this distortion really relevant for such
short-term loans?
6
Modelling Approach
• Intuitively, it is the uncertainty associated
with investment projects that you might
think gives rise to a costly state verification
problem
• Yet, all investment in this model is done out
of retained earnings
• Is there really much (any) uncertainty about
production within a period?
– There may be about sales …
7
Modelling Approach
• Technical comment (2):
– I think the entrepreneurs budget constaint might
be wrong (though it could just be me!)
– Should it not be:
f t At F H t , K t   Cte  K te1  1  d K te
– This error carries through until you make the
assumption that d equals 1
– But I’m not sure it makes and qualitative
difference to your results anyway
8
Results and comments
• Compare steady states of the model with
different degrees of uncertainty
• An increase in uncertainty – measured as a
mean preserving spread (MPS) in
idiosyncratic productivity shocks – leads to
a fall in steady-state output and the creditoutput ratio
– Provided distribution of shocks is symmetric
and bell-shaped and the default rate is small
enough.
9
Results and comments
• As long as a MPS leads to an increase in the
deadweight loss associated with the CSV
problem, then output and the credit-output
ratio will fall
– I must admit to being confused over the
difference between the deadweight loss and the
‘entrepreneurs share’ distortion; I’d appreciate
some intuition as to whether or not this is a
separate distortion (I think not) and why it adds
to the existing deadweight loss
10
Results and comments
• For their particular calibration, the result
follows
• Does the result depend on using a beta
distribution?
– Would increasing the standard deviation of
other distributions – normal, t – achieve the
same effect?
– Is a beta distribution reasonable for the sort of
uncertainty they’re thinking about?
• The authors flag further work that’s relevant here.
11
Results and comments
• Section 3.5 gives the sign of some
derivatives given their calculations
– What does it add over Section 4?
• The effect on output looks very small
– Or is it that the movements in uncertainty they
look at are small? I’m afraid I don’t know how
to think about this.
– Back to my original motivation: the recent
credit crisis seems to have lowered output by
roughly 8%.
12
General comment
• What do we mean by ‘long-run’ effects?
• Authors compare their results to the
literature that says that output and the
credit-output ratio are positively correlated
in the long run
• To me, that means that as output has grown,
so has the credit-output rate
– Surely idiosyncratic uncertainty cannot have
trended downwards?
13
General comment
Potential
output
Credit-output
ratio
Actual output
Increase in
uncertainty
Increase in
uncertainty
• Authors’ model seeks to explain the break in the trend rather
than the trend itself
14
• The literature on financial deepening is all about the trend