External debt and growth: An econometric study

1
External debt and growth An econometric study

• Marianna KoliIDPM, University of Manchester
• ESDS International Conference
• London, 3 December, 2007

2
Research objectives

• To assess the significance of debt as a
  determinant of growth
• To test the debt overhang hypothesis
• To find the critical debt ratios where the impact
  of debt on growth becomes negative

3
Context

• Growth
• Mainstream variable
• Often proxy for development
• Debt
• Instrument of public dev. discourse
• Need hard evidence of real situation
• Little research into exact critical ratios
• Debt overhang hypothesis (Krugman)

4
Debt overhang hypothesis
GDP per capita growth
Debt/GDP
Growth-maximising debt level
5
Previous literature

• A wealth of studies on determinants of growth
• Large country samples (100)
• Some studies on debt/growth and debt overhang
  hypothesis
• Pre-2000 papers 10-30 countries
• Recent papers 20-100 countries

6
Original contributions of this study

• Middle-income countries only
• Smaller subgroups of countries
• Lat Am/non-Lat Am
• Severely/moderately/less indebted
• Determination of critical debt level
• New data compared to previous literature on same
  topic

7
Sample

• 25 middle-income countries as classified by UN
  based on GDP pc
• Middle-income versus low-income
• Lower levels of absolute poverty, HIV and other
  urgent wide-scale problems
• Better possibilities for strategic planning
• Better institutions
• Better government?

8
Sample

• Exclusions
• Population less than 1 000 000 (islands)
• Independence after 1960 (EECA, SSA)
• China excluded for poor data availability
• 85 data points of 7700 (1.1)had to be estimated
  using combination method two moving averages and
  two econometric techniques, with equal weights

9
Subgroups

• Indebtedness
• Severely/moderately/less indebted
• Effect of debt level on debt/growth relationship
• Latin America
• Endemic debt crises in 1980s
• To find out whether debt/growth relationship is
  different in Latin America

10
Methodology

• Regression with fixed effects
• chosen through logical method and OLS/FE/RE
  diagnostic statistics comparison
• Robustness tested through country groups

11
Results

• Significance in several subgroups debt/GDP, debt
  service/X, short-term debt/total debt, budget
  balance/GDP, terms of trade, investment/GDP and
  economic openness
• Results reflect the current theories about
  financial crises in middle-income countries

12
Results debt variables

• External debt/GDP is highly significant in all
  groups except non-Latin America
• External debt without lag gives U-shaped Laffer
  curve
• External debt with lag gives inverted U-shaped
  curve
• All sub-samples give same shape for curve

13
Results Cubic Laffer curve
Lagged variables
Non-lagged variables
GDP per capita growth
Debt/GDP
Growth-maximising debt level
14
Results insignificant variables

• Education, population growth, tax revenue/GDP,
  economic freedom
• Solow model variables found insignificant,
  possibly because of heavy focus of model on debt
• Taxes and economic freedom possibly neutralised
  by other variables (budget balance, investment)

15
Results Being in Latin America

• Small number of significant variables
• Significant variables external debt, budget
  balance and short-term debt
• Debt service is insignificant
• Different causes of growth slowdowns
• Lat Am debt overhang
• Non-Lat Am debt service crowds out investment,
  reducing growth

16
Results Indebtedness level

• SICs debt, stability, openness (negative!)
• LICs debt, investment and stability
• MICs debt, ST debt, terms of trade, education,
  economic openness, investment rate
• Conclusion debt is significant in all groups,
  but reasons may be different

17
Results for debt overhang

• Mixed evidence for DOH (total of 24 squared
  coefficients, of which 12 significant, of which 6
  negative)
• Some evidence for DOH found in all subgroups
  except Less Indebted Countries

18
Critical debt ratios(90 confidence interval)
19
Explanation of critical ratios debt/GDP

• High error margins in some groups, but all
  critical debt/GDP ratios over 100
• Higher thresholds than earlier research
  improved debt tolerance?
• Previous studies included 1970s and included both
  low- and middle-income countries
• Latin America shows lower tolerance than whole
  sample of 25 countries

20
Explanation of critical ratios DS/X

• We find lower critical ratios than Pattillo et al
  (2002)
• Possibly explained partly by sample composition
  high exporters give low DS/X ratios (e.g.
  Thailand, Philippines)
• No clear statistical explanation for lower ratios
  does this mean debt service tolerance has
  fallen?

21
Laffer curve findings debt/GDP

• Four countries found on wrong side of the Laffer
  curve during 1980-2001
• Bolivia (1986-87)
• Jordan (1987, 1989-1995)
• Panama (1988-1989)
• Syria (1983-2001)
• All but Bolivia are still severely indebted

22
Laffer curve findings DS/X

• Seven countries exceed the 30 mark for longer
  periods of time, mostly Lat Am
• Argentina (range 25-80)
• Bolivia (30-60)
• Brazil (20-80)
• Chile (15-70)
• Hungary (20-50)
• Mexico (20-50)
• Peru (10-50)

23
Conclusions

• Crisis avoidance and growth objectives are
  equally important
• Infant institutions argument is valid,
  particularly in SICs
• Policies should be tailored to country
  circumstances
• Model may be specific to conditions where debt is
  important