Transcript Discussion by A. Galindo

Illiquidity, Financial Development and the
Growth-Volatility Relationship
By Enisse Kharroubi
Comments by: Arturo Galindo
Universidad de los Andes
The Growth and Welfare Effects of Macroeconomic Volatility
Barcelona - March 17-18, 2006
General Comment



This is a nice paper that analyzes both from a
theoretical and empirical perspective the link
between volatility, growth and financial development
The paper concludes that the negative relationship
between growth and volatility is stronger in
countries with lower financial development.
…and more likely to be positive in countries with
deeper credit markets.
Two Sets of Comments


Clarifying questions on the theoretical part of
the paper
Comments on the empirical tests
Summary of the Model



The model assumes that lenders choose the allocation of short
term and long term loans, and that there is interim moral
hazard (the possibility of the borrower deviating long term
fund from their original use)
To reduce this type of moral hazard lenders discipline
borrowers by issuing short term debts
This solves a micro problem but generates a coordination
problem and can lead to multiple equilibria


One in which s.t. loans are rolled over
Inefficient run eq.
Summary of the Model


There is a level of short term payments that increases the
desire of the entrepreneur from not deviating form his L.T.
technology. This allows for an incentive compatible solution
in which the lender supplies incentives to the entrepreneur to
continue in the L.T. technology, and can be reached as long as
the L.T. technology is not too illiquid (safe financing
strategy).
There is also a risky lending strategy when the production
technology is sufficiently illiquid. Entrepreneurs make
decisions based on expected rollover probability. If this
probability is low the entrepreneur finances his investment
with less S.T: debt. This is risky in the sense that it depends
on the rollover risk.
Clarifying questions about the model





What is the economic interpretation of h? There is
none in the paper.
Should it be interpreted as a liquidity shock?
Is it known in t=0?
Is there any uncertainty on h?
Is uncertainty about h relevant in the moral hazard
story?
Clarifying questions about the model

Is there uncertainty about R or r? Is it always the
case that R>r2?


Intuitively, why is there a difference between t and
t’? In particular why is t > t’?


Why would the entrepreneur who at t=0 chooses the L.T:
technology want to move to the S.T. technology?
In a two period world (that is in a non repeated game set
up), why is it more costly to default on a l.t. debt than on a
s.t. debt if defaulting on the s.t. debt implies reducing the
roll over probability?
If t’ is sufficiently large would that rule out interim
moral hazard?
Clarifying questions about the model

How is the interest rate structure determined?



In f.n.8: interest rates are exogenous and such that
investors are indifferent between lending s.t. and l.t.
Is the fact that s.t. loans are perfectly enforceable and l.t.
loans are not, the possibility that there is interim MH, and
h, incorporated in the interest rate structure?
Why is rs independent of a and b? Does the competitive
structure of lenders matter? If lenders are competitive one
would expect that they break even, in such a case
wouldn’t a and b and other parameters affect r?
Summary of the Model



In the mixed strategy eq, expected
growth decreases with d and growth
volatility increases with d
In the pure strategy eq. expected growth
increases with d if and only d<m2 +z1
In the pure strategy eq. growth volatility
increases with d if and only d<m2 +z2
Summary of the Model


When d is low the mixed strategy
equilibrium prevails and there is a
negative correlation between growth and
growth volatility
When d is high the pure strategy
equilibrium prevails and there is a
positive correlation between growth and
volatility
Summary of the Model




However, despite the fact that there is a negative correlation,
the impact of d – financial development - on growth and
growth volatility, seem to be counter intuitive for countries
with low d!
Moreover according to the model there is always a negative
relationship between d and growth!
Several research pieces suggest that there is a positive
correlation between d and growth for developing countries
(Levine 2004). The model suggests that this correlation
should be negative!
Moreover research also has suggested that there is a negative
correlation between d and the volatility of growth for
developing countries. (Bekaert, Harvey, Lundblad 2004,
Easterly, Islam, Stiglitz, 2000)
Comments on Empirics

The paper estimates the following regression:
gvoli ,t  ai + bt +  1d i ,t+ 2 g i ,t + 3d i ,t gi ,t + xi ,t +  i ,t


Where d is a measure of financial
development, g is the growth rate of GDP per
capita and x is the log of GDP.
The crucial empirical findings are that
 2 < 0 and  3  0

The author claims that this finding supports
the theory, but one should look at the results
carefully.
According to the data:



In very little cases there is a positive correlation
between g and volatility
According to the estimates of table 2, there is a
positive correlation when ll>0.95
This occurs in only 3% of the observations
County
CHE
JPN
1961
1.0232
1971
1.0889
1.0451
1981
1.3268
1.3924
1991
1.4008
1.8114
According to the data:


According to the estimates of table 3, there is a
positive correlation when fia>1.7
This occurs in only 3.6% of the observations
Country
CHE
JPN
NLD
USA
1961
2.129751
1.884389
1.962291
1.740548
1971
2.586255
Comments on Empirics

The author does not claim any causal effects in the
estimations. In fact endogeneity appears to be a source
of concern.


Not only in g, but also in d
But even if you are looking only at the correlations
you should be cautious with:


High collinearity between d and g, induces high variance in
your estimates so it is difficult to make any inferences
Omitted variables that affect the precision of the estimates:
previous studies have also included developing country
dummies, (M+X)/GDP, volatility of money growth,
volatility of real wages, level and volatility of capital flows,
among others.
The model provides an explanation of
how s.t. and l.t loans are chosen





From the perspective of development countries I have doubts
that the setting is completely relevant.
“Original sin” view, seems to be more important (at least in
LAC).
IDB 2005 explores determinants of loan composition in LAC
and find that regulatory barriers and lack of matching funding
are the main determinants of the composition of lenders
portfolio.
In the case of L.T. vs S.T. loans the driving force for low
share of L.T. loans is lack of L.T. liabilities and restrictions on
maturity mismatches.
Macro instability, prob of S.S., etc. seems to be a more
plausible story for the determination of the debt structure of
firms in LAC