Does compliance matter? Assessing the relationship between sovereign risk and compliance with international monetary law

See Simmons and Hopkins (2005); von Stein (2005); Grieco et al. (2009).

In addition to serving as the empirical fulcrum for the important debate on compliance between von Stein and Simmons and Hopkins, Article VIII is, as a central part of the IMF Treaty, the “first international accord in history to obligate signatories to particular standards of monetary conduct” (Simmons 2000b, 820). Countries that are signatories to Article VIII are required to maintain an open current account. Article VIII signatories are prohibited from placing restrictions on the availability of foreign exchange for goods, services, and “invisibles”—services such as legal and financial advisement, royalties, foreign remittances, etc. Sections 3 and 4 of Article VIII also proscribe states from engaging in, or permitting any of their fiscal agencies to engage in, any discriminatory currency arrangement or multiple currency practices. Although there was discussion within the IMF in the mid-1990s about adding a section to Article VIII requiring capital account convertibility, the requirement only extends to commercial credits granted by exporters or received by importers. See McKinnon (1979, 4–7).

For example, Jensen (2003) and Sobel (1999).

In the statistical analysis, use of Fund credits—a proxy for the Fund’s sanctioning power—is unrelated to the decision to comply.

Italics added for emphasis.

The argument dovetails with Tomz’s (2007) historical analysis of sovereign borrowing. Tomz provides evidence that enforcement through coercive measures (gunboat diplomacy and trade sanctions) cannot explain historical patterns of lending and repayment; rather, governments with good reputations were charged lower interest rates by private lenders, and, preferring to maintain their good reputations, tended to honor their debts in both favorable and adverse economic circumstances.

See North and Weingast (1989). Büthe and Milner (2008) show that membership in GATT/WTO increases foreign direct investment in developing countries.

Downs et al. (1996) suggest that the high rate of compliance identified by Chayes and Chayes (1993) and Mitchell (1994) is a consequence of selection effects.

The literature in political science on the importance of audience costs is large. Some important works include: Fearon (1994); Gaubatz (1996); Lohmann (1997); Mansfield et al. (2002).

See McKinnon (1979, 41) and Simmons (2000b, 820).

I am only excluding wealthy, mature democracies in Western Europe, North America, Oceania, and Japan from the analysis. Since the early 1980s, the only significant incidences of an imposition of current account restrictions by Article VIII members in the OECD occurred in France during Mitterrand’s “U-Turn” in 1983 and in Greece in 1996–97. In addition, the focus of this article is on how compliance may or may not affect reputations, which influence access (and the terms of access) to foreign capital; OECD countries have essentially unlimited access to capital markets, so compliance with Article VIII theoretically should have little effect on the reputation of these countries. There is an additional empirical justification for limiting the sample: Blonigen and Wang (2004) present evidence suggesting that pooling of rich and poor countries is inappropriate in studies of FDI.

Previous work has demonstrated a strong correlation between risk ratings and interest rate spreads (Feder and Ross 1982; Mosley 2006, 98). As an admittedly crude additional test, I used Ahlquist’s (2006) data on portfolio capital inflows as a proportion of GDP to produce a bivariate correlation with my measures of creditworthiness. Unsurprisingly, I found a strongly negative correlation between portfolio inflows/GDP and IIR (ρ = −0.31, p = 0.0000) and the annual size of capital inflows and Euromoney (ρ = −0.37, p = 0.0000).

As discussed in Dreher and Voigt (2008, 25), the construction of the Euromoney rating might pose an endogeneity problem for some of the covariates described in the next section, because factors such as debt level and composition and economic performance are built into the indicator (and thus almost by definition correlated with the country ratings). For this reason, Dreher and Voigt use a modified version of the Euromoney score that extracts three components which are clearly parts rather than determinants of creditworthiness (it is worth noting that Dreher and Voigt report a very high correlation (0.97) between the original and modified ratings). Unfortunately, detailed data that enable the authors to construct a modified rating are only available after 1992; since, following Simmons (2000a, b) and Grieco et al. (2009), my dataset ends in 1997 (and an important robustness check, described below, limits the sample to country-year observations prior to 1992), the Dreher and Voigt solution is too costly for me. However, it is important to note that the explanatory variables I am most interested in—those related to compliance with Article VIII—are not components of either of the two measures of perceptions of credit risk employed in this article.

Reliable data on bond spreads are available for only eight emerging markets from 1994 to the present, and sovereign bond ratings by credit rating agencies are available for a smaller number of countries than either IIR or Euromoney scores (see Mauro et al. 2006, 100). Interest rate differentials are another alternative market-based measure of risk, but this indicator has drawbacks: the availability of data on national interest rates is spotty at best, and until the 1990s, own interest rates in most developing countries were not market-determined (Aizenman and Marion 2004, 575). Nonetheless, I used a simple t-test, relying on Aizenman and Marion’s (2004) construction of the interest rate differential (ln[(1 + i)/(1 + i ^US)], where i ^US is the US T-bill rate and i is the national deposit rate), to see whether countries that fail to comply with Article VIII pay higher relative interest rates. On average, the logged interest rate differential is almost twice as large for countries that are noncompliant with Article VIII (0.41) than it is for countries in which the noncompliance variable equals zero (0.22). The large t-statistic (6.07) indicates that the difference of means between the two groups (compliant and non-compliant) is highly significant (p = 0.0000).

Quinn (1997, 531). See footnote 2 for a description of the actions prohibited by Article VIII.

All member countries of the IMF are, in principle, committed to removing restrictions on the current account. However, upon joining the IMF countries are allowed to retain existing restrictions under Article XIV, which sanctions “transitional” arrangements for countries that are not prepared (or are unwilling) to accept sections 2, 3, and 4 of Article VIII. The IMF attempts to persuade transitional countries to join Article VIII, but some members remained under Article XIV for decades—the Philippines, for example, remained under Article XIV for 50 years (IMF 2006). The Executive Board of the Fund agreed in 1992 that the transitional arrangements had been abused and officials became more forceful in encouraging adoption of Article VIII. It is important to note that once a country notifies the Fund of its acceptance of Article VIII obligations it gives up the right—in perpetuity—to retain existing or impose new current account restrictions. The Fund’s Executive Board has the ability, however, to approve short-term restrictions by Article VIII. The decisions by the Executive Board to approve temporary restrictions are confidential. However, as I discuss in detail below, I take steps to attempt to strip out possible “sanctioned renegers” from the analysis.

16 percent of all country-year observations in the sample are coded as noncompliant.

Biglaiser and DeRouen’s analysis of sovereign bond ratings in Latin America tests the effects of indexes of trade and capital market liberalization, financial liberalization, tax reform, and privatization. The results show that trade reform is the only variable that has a strong (positive) effect on perceptions of creditworthiness (Biglaiser and DeRouen 2007).

The debate on the “catalytic” effect of IMF loans is extensive; see, for example, Bordo et al. (2004); Jensen (2004).

Laevan and Valencia (2008) update the measure of currency crisis originally developed by Frankel and Rose (1996).

The Wooldridge test for autocorrelation is implemented in Stata 11.0 through the xtserial command (Drukker 2003; Wooldridge 2002).

My interest in explaining differences between countries, the relative invariance of the key explanatory variables, and the fact that my data includes many more units (95 countries at most) than observations per unit (18 for panels without missing data on covariates) imply that the fixed effects estimator is inappropriate (Abrevaya 1997). Plümper and Troeger (2007) propose a solution (the fixed effects vector decomposition estimator) that “allows estimating time-invariant variables and that is more efficient than the FE model in estimating variables that have very little longitudinal variance” (2007, 125). When I re-estimated the models in Table 5 with Plümper and Troeger’s xtfevd routine in Stata 11, I obtained very similar findings; in fact, the coefficient on the noncompliance variable is slightly larger in both the fixed effects vector decomposition regressions and OLS regressions with standard errors that assume clustering by country (due to space considerations, the additional results are not reported here, but are available in the online Appendix that supplements the electronic version of this article).

Recall that the dependent variables have been re-scaled so that positive coefficients indicate greater country risk.

The finding is consistent with Reinhart et al.’s (2003) contention that defaults in the past make countries more likely to default on their foreign debts in the future, regardless of the level of indebtedness. This result is also consistent with Archer et al.’s (2007) finding that bond ratings are strongly negatively affected by a history of default.

Non-random selection is at the core of the debate between Simmons and Hopkins (2005) and von Stein (2005). They are interested in the question of whether Article VIII has an independent effect on state behavior, which is a very different question from the one I ask here, and which makes selection processes a much more pressing concern in their debate.

I only report the results from the selection models and additional robustness checks with one-year lags for the variables on the right hand side; full results for models including the three-year lags, as in Table 5, are available in the online Appendix to this article.

Note that the Shift Left variable is omitted from the first stage because, by construction, it is related to Article VIII status: to capture shifts in government partisanship away from the constellation of interests that produced the decision to sign the agreement, it takes a value of 1 if a country has accepted Article VIII obligations and the ideological makeup of the government subsequently moves to the left.

Regional definitions, following Simmons, come from the World Bank’s regional classifications. The level of regional noncompliance should be relatively uncorrelated with a country’s sovereign risk rating: it seems unlikely, for example, that market actors would incorporate information about Peru’s compliance in developing risk assessments for Argentina. Evidence from research on compliance with transparency rules provides support: Glennerster and Shin (2007) examine whether adoption of IMF-led transparency reforms lowers sovereign bond spreads and find that regional adoption of transparency reforms has no effect on borrowing costs.

This is implemented in Stata 11 via the cdsimeq command. The method reports correct standard errors when, in a system of simultaneous equations, one endogenous variable is continuous and the other is dichotomous. See Keshk (2003) for details.

Simmons (2000a, b).

Simmons (2000a, b).

World Bank, Global Development Finance CD-ROM (2003).

World Bank, Global Development Finance CD-ROM (2003).

World Bank, Global Development Finance CD-ROM (2003).

World Bank, World Development Indicators CD-ROM (2004).

World Bank, World Development Indicators CD-ROM (2004).

World Bank, World Development Indicators CD-ROM (2004).

Monty Marshall and Keith Jaggers, “Polity IV Project: Political Regime Characteristics and Regime Transitions, 1800–2002,” Center for International Development and Conflict Management (CIDCM), http://​www.​cidcm.​umd.​edu/​inscr/​polity.

Fearon and Laitin (2003).

Ahlquist (2006).

Wacziarg and Welch (2003, 2008). Expansion and improvement of trade openness measure developed in Sachs and Warner (1995).

Vreeland (2003).

Beck et al. (2001).