1. A Gravity Model of Sovereign Lending: Trade, Default and Credit Andrew K. Rose and Mark M. Spiegel 4 th annual I.M.F. Research Conference November 6, 2003

2. Sovereign defaults are still exceptional Direct penalties are elusive Direct penalties are elusive “Gunboat diplomacy” no longer viable “Gunboat diplomacy” no longer viable However, countries largely behave “as if” default penalties were perceived However, countries largely behave “as if” default penalties were perceived Motivation for sovereign debt service therefore remains an important issue Motivation for sovereign debt service therefore remains an important issue

3. One posited penalty in literature is loss in trade Bulow and Rogoff (1989): Trade sanctions as potential penalty Bulow and Rogoff (1989): Trade sanctions as potential penalty Also loss of trade creditAlso loss of trade credit However, unclear whether creditors can levy such penaltiesHowever, unclear whether creditors can levy such penalties Empirical questions about efficacy of default penalty Empirical questions about efficacy of default penalty Creditors may be unable to levy such penaltiesCreditors may be unable to levy such penalties Penalties may not be “renegotiation proof” [Kletzer and Wright (2000)]Penalties may not be “renegotiation proof” [Kletzer and Wright (2000)]

4. Some empirical evidence of trade default penalties Ozler (1993): Evidence of positive, but small premia charged to countries with default histories Ozler (1993): Evidence of positive, but small premia charged to countries with default histories Cline (1987): Bolivia and Peru experienced disruption in trade credits subsequent to Paris Club renegotiation Cline (1987): Bolivia and Peru experienced disruption in trade credits subsequent to Paris Club renegotiation Rose (2002): Sovereign Paris Club reschedulings followed by significant reductions in trade Rose (2002): Sovereign Paris Club reschedulings followed by significant reductions in trade

5. We examine notion of lost trade as enforcement mechanism Harsher penalties in sovereign debt usually improve global welfare by moving closer to first-best outcome Harsher penalties in sovereign debt usually improve global welfare by moving closer to first-best outcome Nations that can threaten heavier trade disruptions therefore have a comparative advantage in lending Nations that can threaten heavier trade disruptions therefore have a comparative advantage in lending We explore that idea We explore that idea

6. Simple borrowing model Sovereign borrowing by small debtor country from two creditor countries Sovereign borrowing by small debtor country from two creditor countries Creditors identical except for bi-lateral trade volumes with debtor country Creditors identical except for bi-lateral trade volumes with debtor country Model predicts that borrowing will be concentrated on country with greater bi- lateral trade Model predicts that borrowing will be concentrated on country with greater bi- lateral trade Then confirm empirically Then confirm empirically

7. Model Assumptions Three countries: borrower country, i, and two creditor countries, a and b Three countries: borrower country, i, and two creditor countries, a and b is a random variable reflecting total trade between country i and country j in the second period is a random variable reflecting total trade between country i and country j in the second period Expectations of are unbiased Expectations of are unbiased where is an i.i.d. disturbance term with expected value 0 on the interval where is an i.i.d. disturbance term with expected value 0 on the interval

8. Model Assumptions (2) Bilateral gains from trade are exogenous and equal to, where is a positive constant Bilateral gains from trade are exogenous and equal to, where is a positive constant r is one plus the world risk-free interest rate, which is exogenous r is one plus the world risk-free interest rate, which is exogenous Lenders are risk-neutral Lenders are risk-neutral If the debtor defaults on country j it suffers a penalty equal to a fraction of its gains from bilateral trade with country j If the debtor defaults on country j it suffers a penalty equal to a fraction of its gains from bilateral trade with country j

9. Extensive form Model has two periods Model has two periods In first period, the representative lender in country j extends a loan of magnitude in return for the promise of a fixed payment in the second period In first period, the representative lender in country j extends a loan of magnitude in return for the promise of a fixed payment in the second period In the second period, is realized and the debtor makes its default decisions In the second period, is realized and the debtor makes its default decisions If the debtor chooses to service its debt it pays If the debtor chooses to service its debt it pays If it defaults, it suffers default penalty If it defaults, it suffers default penalty

10. Agent Characteristics Creditor nations differ only in their expected trade volume with the debtor Creditor nations differ only in their expected trade volume with the debtor Expected debtor utility satisfies Expected debtor utility satisfies where is exogenous 1st-pd consumption satisfies 1st-pd consumption satisfies is exogenous in both periods is exogenous in both periods

11. Default Decision Default decision based on maximizing second period consumption Default decision based on maximizing second period consumption Conditional on debt service, satisfies Conditional on debt service, satisfies where and represents cost of default decision on debt owed to country k Debtor chooses default on country j when Debtor chooses default on country j when

12. Equilibrium Define as minimum realization of that induces debt service. Satisfies Define as minimum realization of that induces debt service. Satisfies Equilibrium is defined as a pair of debt obligations that maximize expected debtor utility subject to both creditors’ zero profit conditions Equilibrium is defined as a pair of debt obligations that maximize expected debtor utility subject to both creditors’ zero profit conditions

13. Borrowing decisions Two decisions: the overall borrowing level, Two decisions: the overall borrowing level, and the allocation across countries Given allocation satisfies Given allocation satisfies Debtor skews borrowing towards nation with lower probability of default with equal borrowing Debtor skews borrowing towards nation with lower probability of default with equal borrowing Doing so increases the default probability in “safe” nation and narrows this difference Doing so increases the default probability in “safe” nation and narrows this difference

14. Result 1: Lending Shares Demonstrate in text that Demonstrate in text that Holding total lending constant, the share of lending originating in country a is increasing in the expected volume of trade with country a Holding total lending constant, the share of lending originating in country a is increasing in the expected volume of trade with country a

15. Result 2: Overall borrowing Maximizing expected utility over the choice of and the optimal allocation rule, and then totally differentiating with respect to and yields Maximizing expected utility over the choice of and the optimal allocation rule, and then totally differentiating with respect to and yields which implies that total borrowing increases with We next test these empirically We next test these empirically

16. Data Set Use annual panel data set of trade and lending Use annual panel data set of trade and lending 20 creditors, 149 debtors, 1986-199920 creditors, 149 debtors, 1986-1999 Bank claims from BISBank claims from BIS Rest from Glick-RoseRest from Glick-Rose

17. Methodology Estimate “gravity” model of lending: Estimate “gravity” model of lending: ln(Cijt) = ln(Xijt) + Zijt + ijt ln(Cijt) = ln(Xijt) + Zijt + ijt where Z are gravity variables (distance, GDP, …)where Z are gravity variables (distance, GDP, …) IV critical because of simultaneity IV critical because of simultaneity Use different instrumental variables from gravity model, especially geographic (landlocked status …)Use different instrumental variables from gravity model, especially geographic (landlocked status …)

18. Miscellany Robust standard errors (clustered by country-pairs) recorded in parentheses. Robust standard errors (clustered by country-pairs) recorded in parentheses. Intercepts and year effects not recorded. Intercepts and year effects not recorded. Instrumental variables for trade are: distance; land border; number landlocked; number island nations; log of area. Instrumental variables for trade are: distance; land border; number landlocked; number island nations; log of area.

19. Table 1: OLS Estimates of Effect of Trade on Claims  Default.54 (.04) Without controls.75 (.02) Levels.0001 (.00003) Levels without controls.0001 (.00003) 1990.51 (.05) 1995.53 (.07) Only industrial debtors.74 (.04)

20. Table 2a: IV Estimates of Effect of Trade on Claims, Geographic Instruments  Default.41 (.07) Without controls.50 (.04) Levels.00006 (.00001) Levels without controls.00007 (.00002) 1990.52 (.10) 1995.40 (.10) Only industrial debtors 1.03 (.07)

21. Table 2b: IV Estimates of Effect of Trade on Claims, Excludable Instruments  Default.80 (.40) Without controls.83 (.07) Levels.00004 (.00001) Levels without controls.00005 (.00001) 1990.59 (.37) 1995 1.13 (.49) Only industrial debtors.79 (.29)

22. Table 3: IV Estimates of Effect of Trade on Claims, Controlling for Total Claims  Default.40 (.07) Without controls.42 (.04) Levels.00005 (.000004) Levels without controls.00005 (.000006) 1990.47 (.10) 1995.37 (.10) Only industrial debtors.48 (.23)

23. Table 3: IV Estimates of Effect of Trade on Claims, Controlling for Total Debt  Default.42 (.07) Without controls.27 (.04) Levels.00006 (.00002) Levels without controls.00006 (.00002) 1990.56 (.09) 1995.42 (.10) Only industrial debtors 1.10 (.20)

24. Table 4: IV Estimates of Effect of Trade on Claims, Panel Estimator: OLS, RE OLS, FE IV, RE Default.31 (.01).19 (.02).52 (.06) Without controls.38 (.01).19 (.01).52 (.03) Levels.00003 (.000001).00002 (.000001).00006 (.00001) Levels no controls.00003 (.000001).00002 (.000001).00007 (.000003) industrial debtors.46 (.06).28 (.07).96 (.19)