Transcript Dr.David Tennant

MODELLING THE RELATIVE EFFECTS OF
FINANCIAL SECTOR FUNCTIONS ON
ECONOMIC GROWTH IN A DEVELOPING
COUNTRY CONTEXT USING COINTEGRATION
AND ERROR CORRECTION METHODS
PAPER PRESENTED AT THE FRANCIS
CONFERENCE
September 28 – 30, 2007
DAVID TENNANT, CLAREMONT KIRTON &
ABDULLAHI ADBULKADRI
DEPARTMENT OF ECONOMICS, UWI, MONA
OVERVIEW OF PRESENTATION:
• Background and Objectives
• Proxies of Financial Sector Functions
• Control Variables
• Methods
• Results and Findings
• Conclusions
BACKGROUND and OBJECTIVES
• The Monterrey Consensus (2002) notes that many
developing countries increasingly depend on local funds to
finance their development needs.
• Financial institutions can facilitate economic growth by
mobilizing savings, allocating these savings to the most
productive investments, and by facilitating the smooth flow
of trade needed in a market-driven economy.
• Theoretical models supporting this view have been
developed by many economists.
BACKGROUND and OBJECTIVES Cont’d
• Many of the empirical studies of the finance-growth
relationship use broad proxies of financial sector
development.
• It would however be very useful to determine which of the
distinct functions of the financial sector have the greatest
impact on economic growth.
• Holden and Prokopenko (2001), Levine and Zervos (1998),
and De Gregorio and Guidotti (1995) all cite difficulties
involved in developing proxies to accurately and
comprehensively capture the many different functions
performed by the financial sector.
BACKGROUND and OBJECTIVES Cont’d
• Levine’s (1997) five basic functions of the financial sector
are used as the basis of this study.
• The Jamaican case study is used.
• In the post-liberalization phase of the evolution of the
Jamaican financial sector, the country experienced a
financial crisis, and a government-led restructuring and
subsequent divestment of the sector.
• Important conclusions are presented regarding the relative
importance of financial sector functions to the creation of
economic growth, and the impact of financial crises on the
economy.
PROXIES of FINANCIAL SECTOR FUNCTIONS:
Savings Mobilization
• SMOB = Deposits / (Total Assets – Loans)
• A positive relationship between SMOB and economic
growth is therefore expected.
Risk Diversification
• The measure of risk diversification used (DRISK) focuses
on the degree of diversification by sector.
PROXIES of FINANCIAL SECTOR FUNCTIONS:
• DRISK for each type of financial institution is calculated by
first finding the percentage of total loans allocated by sector.
• The standard deviation of the percentage of total loans
allocated to each sector is then used to measure the spread
for each institution from the state of uniform distribution.
• A measure of the degree of diversification for the entire
financial sector is calculated using a weighted average.
• A negative relationship between DRISK and economic
growth is expected.
PROXIES of FINANCIAL SECTOR FUNCTIONS:
Resource Allocation
• RESAL = credit to private sector production / total loans.
• It is expected that RESAL will have a positive relationship
with economic growth.
Corporate Control
• The proxy used for corporate control (CORPC) assumes
that as connected party loans and/or connected party
financial investments increase, then the ability of financial
institutions to exert corporate control decreases.
PROXIES of FINANCIAL SECTOR FUNCTIONS:
• CORPC = connected party loans and financial investments
/ total loans and financial investments.
• CORPC is expected to have a negative relationship with
economic growth.
Ease of Trading
• Using the approach of Levine and Zervos (1998), ETRAD
= value of shares traded / current GDP.
• Levine (1991) initially argued that a positive relationship
with economic growth should be expected, he later notes
the argument that very liquid markets encourage investor
myopia, thus actually hurting economic growth (Levine
1996).
CONTROL VARIABLES
• A measure of trade openness (TRADE) was included,
and calculated as the sum of the country’s imports and
exports divided by current GDP.
• Exchange rate volatility (XRATEVOL) for each
quarter is calculated as the standard deviation of the
percentage change in the US$ real exchange rate for
the four preceding quarters.
• Two dummy variables were used to account for
structural changes affecting the Jamaican financial
sector – liberalization and crisis.
METHODS
• Data
• Quarterly time series from 1986-2005
• Due to missing data most estimations limited to
1989-2005
• Augmented Dickey-Fuller test used to determine
stationarity
• Cointegration Analysis
• Used to determine long-run relationships among
data series
• Two specifications were considered: (1) using
individual proxies (2) with interaction terms
METHODS
Model (1):
ln GDPt   0   1 ln TRADE t   2 ln XRATEVOLt   3 ln ETRADt   4 ln CORPC t
  5 ln DRISK t   6 ln RESALt   7 ln SMOB t   8 STRUCTURE t  et
Model (2):
ln GDPt   0   1 ln TRADE t   2 ln XRATEVOLt   3 ln CORPC t   4 ln RESALt
  5 ln SMOB t * ln RESALt   6 ln SMOB * ln ETRADt   7 ln SMOB t * ln DRISK t
  8 STRUCTURE t  et
METHODS
• Error Correction Model
• Used to examine short-run dynamics
• Enables us to determine short-run adjustments to
long-run equilibrium
L
L
L
l 0
l 0
l 0
 ln GDPt    ECTt 1   1l  ln TRADE t l    2 l  ln XRATEVOLt l    3l  ln CORPC t l
L
L
L
l 0
l 0
l 0
   4 l  ln RESALt l    5l  ln SMOB t l * ln RESALt l    6l  ln SMOB t l * ln ETRADt l
L
   7 l  ln SMOB t l * ln DRISK t l   8 STRUCTUREt   t
l 0
RESULTS
• Stationarity
• LnGDP I(2), LnETRAD and LnTRADE I(0), all
others I(1)
• Johansen Cointegration test indicated nine possible
cointegrating relationships for Model (1) and six for
Model (2)
• Selection of reported equation was based on conformity
with theory
Table 1 – Estimates of Initial Model Cointegrating
Equation with lnGDP as Dependent Variable
Variable Name
Coefficient
t-Statistic
CONSTANT
-6.972*
-10.329
lnSMOB
1.206*
4.691
lnRESAL
-1.870*
-3.503
lnCORPC
0.224
1.432
lnDRISK
-1.507*
-9.442
lnETRAD
-0.145*
-5.513
CRISIS
0.270
0.791
lnTRADE
0.271
0.736
lnXRATEVOL
0.286*
8.085
* indicates significance at 1% level
Table 2 – Estimates of Alternative Model Cointegrating
Equation with lnGDP as Dependent Variable
Variable Name
Coefficient
t-Statistic
CONSTANT
-6.441*
-48.186
lnCORPC
-0.222*
-7.686
lnRESAL
1.342*
15.490
lnSMOB*lnRESAL
1.232*
9.985
lnSMOB*lnDRISK
-0.785*
-7.704
lnSMOB*lnETRAD
-0.361*
-12.317
CRISIS
0.797*
15.474
lnTRADE
1.110*
13.285
lnXRATEVOL
-0.117*
-15.842
* indicates significance at 1% level
FINDINGS
•Model (2) outperforms Model (1)
•Has more significant variables and conforms better with
theoretical expectations
•Establishes long-run relationship between GDP and
dependent variables.
•ECM indicates
•No short-run relationship between GDP growth and
dependent variables
•The speed of adjustment is ~35%
•CRISIS has a negative, though, insignificant effect on
GDP growth, compared with positive and significant
effect in the long-run
CONCLUSIONS
• Sophisticated proxies of financial sector function
have been developed
• Based on Favara (2003)
• ensure accuracy and conformity to theory
• Savings mobilization is an essential factor through
which other proxies impact economic growth
• Resource allocation
• Ease of trading
• Risk diversification
CONCLUSIONS
• Financial sector reform should:
• Focus on savings mobilization and the efficient
allocation of mobilized savings
• Emphasize removal of government distortions in
financial markets
• Not be expected to yield immediate results