External Debt, Capital Flight and Political Risk

The paper addresses the coexistence, in several Latin American countries of the 1970s-80s (most clearly Argentina, Mexico, Venezuela, Uruguay, and to a lesser extent Peru and the Philippines), of a large buildup of public external debt, a large accumulation of private external assets by the same countries’ residents, and declining domestic capital formation (Sec. 1, citing Table 1). The authors note this is “a special sort of intermediation” (quoting Sachs 1984): Western banks lent to Latin American public sectors, and those same countries’ private sectors made offshore deposits that often returned to the same banks. Dornbusch (1986), Diaz Alejandro (1984) and Edwards (1987) are cited for the point that because the private external assets were generally not repatriated, these economies had to undergo large adjustments to generate the foreign exchange needed to service the public debt. The paper frames this as an “apparent form of collective irrationality” to be explained by political polarization and instability, rather than by policy mistakes alone.

The economy has two periods and two groups of agents distinguished by productive role: “workers,” who inelastically supply one unit of labor and cannot hold capital, and “capitalists,” who own the capital stock and hire labor under a constant-returns-to-scale technology y_t = f(k_t) (Sec. 2, eqs. 1-3). Each group has its own political representative (“party”); whichever party is in office sets labor and capital tax rates, makes lump-sum transfers, and can borrow abroad subject to an exogenous credit ceiling d-bar, common to both types. The government in office in period 1 is type “w” (workers’) or “c” (capitalists’) with no presumption about which; in period 2, type “c” holds office with exogenous probability P and type “w” with probability 1-P, interpreted either as the outcome of an election in which rational voters favor whichever party benefits them, or as the probability a non-democratic government is overthrown.

Under a simplifying condition (eq. 14) that ensures full debt repayment is always feasible without taxing a government’s own constituency, each government sets the tax rate on its own group’s income to zero and the tax rate on the opposing group’s income to its maximum (100%), with no transfers to the opposing group (Sec. 3). This follows because, ex post, labor income tax is non-distortionary (labor supply is inelastic) and capital income tax is non-distortionary in period 2 (capital is predetermined) – the classic time-inconsistency-of-capital-taxation result discussed by Fischer (1980) – so a government that places zero weight on its opponents’ welfare has no reason to spare them. The authors note the same qualitative outcome would follow from less extreme preferences as long as the political/administrative cost of changing tax rates exceeds that of changing transfers.

Workers’ first-order condition (eq. 17) shows that whenever P < 1 (there is a positive chance the “c” type takes office and expropriates labor income), workers optimally hold positive external assets, b > 0, as insurance; capitalists’ first-order conditions (eqs. 19a-b) show symmetrically that whenever P > 0, capitalists hold external assets and the domestic marginal return to capital exceeds the world interest rate – i.e., political uncertainty simultaneously produces private capital flight and depresses domestic investment relative to the full-certainty benchmark (Sec. 3). On the government side, Proposition 1 shows that if 0 < P < 1 a type-“c” government in period 1 always borrows the maximum amount d-bar and transfers it entirely to capitalists, who then raise consumption, buy foreign assets, and raise domestic investment; Proposition 2 gives the symmetric result for a type-“w” government. The intuition from the paper’s Lagrange-multiplier derivation (eq. 21): the marginal gain of issuing one more unit of debt today (valued at the governing group’s full marginal utility) is weighed against the expected marginal disutility of repaying it tomorrow, discounted by the probability P of the same type remaining in office – since there is always a chance the opposing type will be in office and the cost of repayment falls on nobody the current government cares about, this expected marginal cost is always strictly less than the marginal benefit whenever P < 1, so the government borrows up to the ceiling (eq. 22 shows the multiplier is strictly positive whenever P < 1).

Proposition 3 (proved in Appendix 3) establishes comparative statics: domestic investment k2 and capital flight b (and b) are increasing in the ceiling d-bar on government borrowing, and – for the capitalists – the effect of political risk is such that a reduction in P (a lower chance capitalists retain power) reduces domestic investment and raises capitalists’ capital flight.* Because workers’ capital flight moves in the opposite direction with P and, the authors argue, is realistically much smaller in magnitude than capitalists’, the model implies that capital flight should be particularly intense when a right-wing regime is expected to collapse, and immediately after a left-wing regime takes office before new capital taxation is implemented (Sec. 3). The proposition also implies that relaxing the government’s external borrowing constraint (e.g., through IMF or World Bank intervention that lets the country borrow more) raises domestic investment but can simultaneously raise capital flight.

Adding a non-discriminatory ceiling q on private foreign-asset holdings (eq. 25) and re-solving, Proposition 4 shows that a type-“c” (capitalists’) government never imposes binding capital controls, while a type-“w” (workers’) government always imposes capital controls binding on both workers and capitalists whenever, absent controls, capitalists’ asset holdings exceed workers’ (Sec. 4). The capitalists’ government gains nothing from controls (they only constrain workers’ intertemporal consumption smoothing, which does not affect capitalists’ welfare) and so would never choose them; the workers’ government trades off the cost of restricting workers’ own consumption-smoothing against the benefit of forcing capitalists to invest more domestically (raising k2 and future tax revenue), formally d(k2)/d(q) < 0. The tightness of the optimal control is inversely related to workers’ risk aversion: highly risk-averse workers would be hurt badly by illiquidity, so even a type-“w” government tolerates substantial capital flight, whereas risk-neutral workers’ government would set q = 0, prohibiting outflows entirely. The authors report this pattern is “probably consistent with” Table 5’s capital-restriction index for Argentina, Chile and Peru (tightening under left-wing/loosening under right-wing shifts), though they note Mexico and Venezuela show less political volatility and correspondingly more stable capital-control policy.

Section 5 allows each government the option of repudiating inherited debt in the final period, at a cost of losing a fraction theta_i (i = w, c) of national output (following Sachs 1985 and Cohen-Sachs 1985) plus the seizure of a fraction mu of the country’s citizens’ external assets by foreign creditors, with the paper assuming 1 > theta_c (cid:18) theta_w > 0 – i.e., right-wing governments are assumed to lose relatively more from the trade/financial disruption of repudiation, since their constituencies rely more on foreign trade and investment (footnote 14 notes the qualitative results survive theta_w > theta_c too, given the paper’s realistic assumption that capitalists’ non-taxable endowment e^c sufficiently exceeds workers’ e^w, eq. 30). Under this assumption, the paper shows the type-“c” government never repudiates in equilibrium (eq. 31), while the type-“w” government repudiates only in the “case (a)” region (defined by condition 32, which holds when P exceeds a threshold P*) if debt exceeds the lower threshold d-bar^w and the “w” type unexpectedly holds office in period 2 (Proposition 5). The authors connect this to the empirical observation that left-wing governments have historically been more vocal critics of IMF adjustment programs designed to guarantee debtor solvency, and note the prediction is consistent with historical repudiations by the Soviet Union, China and Cuba after their revolutions (cited from a companion paper by Alesina).

Via Proposition 3, the paper’s comparative statics imply that if international organizations such as the World Bank or IMF enable a developing-country government to borrow more by relaxing the credit ceiling d-bar, the model predicts this will raise domestic investment but can also raise the volume of capital flight away from the country (Sec. 3) – a cautionary implication about debt-relief-style interventions that do not address the underlying political-uncertainty distortion, though the paper does not model IMF/World Bank behavior explicitly or evaluate welfare effects of such interventions.

The authors state the model is deliberately the “simplest possible version” chosen for readability, and assert (without full derivation in the main text) that the qualitative results generalize along several dimensions: to an infinite-horizon setting for some functional forms (citing their own companion paper, Alesina-Tabellini 1987); to endogenous labor supply; to a setting where workers cannot hold external assets; to a model with public-goods provision; to less extreme government preferences (with caveats); to endogenously derived P from voter preferences; and to upper bounds on tax rates short of full expropriation (Sec. 6, Conclusions). They explicitly flag unaddressed extensions as directions for future research rather than claims already established: conflict between tradeable and nontradeable sectors (rather than capital/labor) as an alternative political cleavage; monetary variables and exchange-rate devaluation as additional channels of capital flight (citing Dornbusch 1985 and Cuddington 1985 as complementary, not competing, explanations); and the politico-economic choice between repudiation and “solvency with adjustment.”