Forgive or Buy Back: An Experimental Study of Debt Relief

1
Forgive or Buy Back An Experimental Study of
Debt Relief

• Vivian Lei, Steven Tucker, and Filip Vesely

2
Motivation

• Bono, Brad Pitt, the Dalai Lama, the late Pope
  John Paul II, , and the Jubilee Debt Campaign
• Call for 100 cancellation of the massive
  external debt owed by the worlds poorest
  countries.
• Demand an end to the scandal of poor countries
  paying money to the rich world.
• I encourage you in your advocacy for total debt
  cancellation for poor countries because, frankly,
  it is a scandal that we are forced to choose
  between basic health and education for our people
  and repaying historical debt. (President Mkapa
  of Tanzania, 2005)

3
Motivation
4
Motivation
5
Motivation

• HIPCs has received significantly more capital
  inflow in the form of new lending and foreign aid
  than their debt service.
• Capital outflow in the form of debt service 3
  of the GDP
• Capital inflow in the forms of new lending and
  aid 15 of the GDP
• Reducing poor countries heavy debt burden has
  always been on developed countries agenda since
  1970s.
• The Paris Club
• Rescheduled payment deadlines for 81 countries
  between 1976 and 1988
• The Brady Plan
• Reduced US60 billion of debt for 16
  middle-income countries during the early 1990s.
• The HIPC (Highly Indebted Poor Countries)
  Initiative
• Reduced US37 billion of debt for 30 HIPC
  countries by the end of 2005.

6
Motivation

• Question 1 Will debt relief really help poor
  countries and also benefit their creditors?
• Krugman (1988) Yes, as long as a debt overhang
  is present.
• Debt overhang
• The expected present value of a countrys future
  resource transfers is less than its debt.
• Impede investment and growth and thus increase
  the probability of a default in the future.
• Decrease the expected value of repayments.

7
Motivation

• Question 2 Which debt relief scheme is best to
  relieve debt burden?
• Krugman (1989)
• Compare debt forgiveness vs. more market-based
  schemes such as debt buybacks
• Forgiveness a once-and-for-all reduction in the
  future obligations of a debtor country
• Buyback allows a debtor country to buy back its
  own debt at a discount
• As long as the debtor country is initially on the
  downward-sloping side of the debt Laffer curve,
  both creditors (acting collectively) and debtors
  should be indifferent, in expected terms, between
  the two schemes.

8
Motivation

• Question 3 What does the empirical literature
  say about the efficiency or effectiveness of
  different relief schemes to solve for the problem
  of debt overhang?
• Not much.
• Most empirical studies aim to investigate if debt
  overhang really exists.
• Regress growth rate of GDP/investment on debt
  stock/flow, using linear/nonlinear specifications
  and various techniques to control for
  endogeneity.
• Results are far from conclusive.
• No study has compared the relative effectiveness
  of different relief schemes because developed
  countries use the case-by-case approach to deal
  with poor countries debt problems.

9
Objective

• To investigate the effectiveness of debt
  forgiveness and debt buybacks in the presence of
  debt overhang in the lab.
• Study the impact of different relief schemes on
• creditors behavior
• How much debt are creditors willing to reduce?
• debtors behavior
• How much effort are debtors willing to exert to
  improve their economic conditions?
• expected payoffs of both sides

10
Design

• 2x2 design treatment variables are
• Relief scheme
• Debt forgiveness
• Debt buybacks
• Number of creditors
• One creditor
• Two creditors

11
Design

• 4 Treatments
• Forgiveness/1 creditor
• Buyback/1 creditor
• Forgiveness/2 creditors
• Buyback/2 creditors

Due to project overhang!
12
Numerical Example

• Consider a risk-neutral debtor country
• Inherits a nominal debt of 120, which is greater
  than its current resources, 40.
• Has a chance to invest and, with some
  uncertainty, generate more income in the future.
• With probability p, the investment succeeds and
  the debtor receives extra 80.
• With probability 1-p, the investment fails and
  the debtor receives nothing.

debt overhang
13
Numerical Example

• Consider a risk-neutral debtor country (contd)
• Incur cost to strive for the extra income.
• The cost function, e(p), is a convex function of
  p.
• Decision needs to be made
• How much effort it is willing to exert (how much
  cost it is willing to incur) in order to generate
  extra 80?
• Decision variable p (or equivalently effort cost
  e)

14
Numerical Example

• Suppose there is no debt relief.
• How much is the expected value of debt repayment
  (EV)?
• EV p (4080)(1-p) 40 4080p
• How much is the debtors expected payoff (EU)?
• EU 4080p-EV-e(p) -e(p)?0
• There is no incentive for the debtor country to
  undertake any investment (political or economic
  reform) when they have to repay a full amount.

15
Numerical Example

• Consider a two-stage game in which creditor
  countries, acting collectively, are willing to
  reduce some debt.
• Stage 1 The representative creditor decides how
  much debt, if any, will be relieved.
• Debt forgiveness decide the amount to be
  forgiven (Flt80)
• Debt buyback decide the price (Plt1) at which the
  creditor is willing to sell for each dollar of
  the debt claims
• Stage 2 The debtor chooses the effort level,
  represented by p, that would generate the extra
  income.

16
Numerical Example

• Debt forgiveness
• How much is the expected value of debt repayment
  (EV)?
• EV p 40(80-F)(1-p) 40 40p (80-F)
• How much is the debtors expected payoff (EU)?
• EU 4080p-EV-e(p) pF -e(p)gtlt0
• e(p)

17
Numerical Example

• Debt forgiveness (contd)
• Expected payoffs

unique Pareto-dominant subgame-perfect equilibrium
18
Numerical Example

• Prediction for debt forgiveness
• F 40 (the amount of relief)
• p 30
• EV 52
• EU 5

19
Numerical Example

• Debt buybacks
• The debtor country benefits by buying back as
  much debt as possible.
• If P is relatively high, then the debtor would
  spend all 40 of its current resources to buy
  back 40/P amount of debt.
• Example If P0.5, then the debtor would be able
  to buy back 40/0.580 at a total price of 40.
• Remaining debt 120-80 40
• Amount of relief 80-40 40
• If P is relatively low, then the debtor would
  spend 120P to buy back all 120 of the debt.
• Example If P0.2, then the debtor would be able
  to buy back all 120 of debt at a total price of
  24.
• Remaining debt 0
• Amount of relief 120-24 96

20
Numerical Example

• Debt buybacks (contd)
• How much is the expected value of debt repayment
  (EV)?
• EV 40 p(120-40/P)
• How much is the debtors expected payoff (EU)?
• EU p80-(120-40/P)-e(p) p(40/P-40) -e(p)gtlt0

21
Numerical Example

• Debt buybacks (contd)
• Expected payoffs

unique subgame-perfect equilibrium
22
Numerical Example

• Prediction for debt buybacks
• P 0.5
• Amount of relief 40 (the same as F under the
  forgiveness scheme)
• p 30
• EV 52
• EU 5

23
Some Experimental Features

• Each session consisted of 16 subjects.
• Randomly assigned 8 subjects to be debtors and 8
  to be creditors.
• Subjects interacted with each other via the
  computer for 20 periods.
• Random matching protocol
• Subjects were re-matched every period.
• Zero probability of being matched with the same
  counterpart for two consecutive periods.

24
Available Data

• 6 sessions (3 for each treatment) which lasted
  about two hours
• 96 subjects
• 960 observations
• Average earnings NZ25.41 (roughly US17.64)
• Creditors NZ40.90
• Debtors NZ 9.93

25
Result 1 Amount of Debt Relief
Forgive 45.54
Buyback 37.11
26
Result 1 Amount of Debt Relief
Panel Data Approach GLS with Random Effects

• There is significantly more debt being relieved
  under the Forgive treatment.

27
Result 2 Project Success Rate (p)
Buyback 35.73
Forgive 36.92
28
Result 2 Project Success Rate (p)
Panel Data Approach GLS with Random Effects

• Debtors effort in terms of the project success
  rate is significantly smaller under the Forgive
  treatment once the amount of debt relief is
  controlled for.
• The more the creditor relieves the debt, the more
  the debtor reciprocates.

29
Result 2 Project Success Rate (p)
Panel Data Approach GLS with Random Effects

• Debtors effort exhibits greater volatility from
  one period to the next under the Forgive
  treatment.

30
Result 3 Expected Payoffs
Buyback 52.64
Forgive 50.53
31
Result 3 Expected Payoffs
Forgive 3.89
Buyback 2.68
32
Result 3 Expected Payoffs
Panel Data Approach GLS with Random Effects

• Given the amount of debt relief, debt forgiveness
  has a significantly negative impact on creditors
  expected payoff but not on debtors.

33
Conclusion

• Creditors tend to relieve more debt under the
  Forgive treatment.
• Debtors do reciprocate, but they dont
  reciprocate significantly more under the Forgive
  than under the Buyback treatment.
• That is, creditors pay more for the same outcome
  under the Forgive treatment.
• Debt forgiveness is a less efficient scheme for
  creditors to relieve the debt.

34
One Creditor
Two Creditors
35
One Creditor
Two Creditors