94668 MULTILATERAL DEBT RELIEF THROUGH THE EYES OF FINANCIAL MARKETS Claudio Raddatz* Abstract—This paper conducts an event study of the impact of multilateral required to service a massive debt burden may find unprof- debt relief initiatives announcements on the stock prices of South African companies with subsidiaries in countries benefited by these initiatives. itable certain investment projects that they would otherwise It shows that these prices increase significantly above those of other firms, finance, and governments may be reluctant to incur costly especially around the launching of the Multilateral Debt Relief Initiative. reforms if a large part of their returns would go to foreign These price increases are consistent with lower expected levels of future taxation in the benefited countries and provide evidence of the economic creditors. consequences of multilateral debt relief that is robust to reverse causality Since, according to the debt overhang argument, debt between economic performance and the decision to grant debt relief. relief should be associated with increased private and social investment and better macroeconomic performance, many I. Introduction researchers have looked for the impact of debt relief on growth and investment. However, finding causal evidence of O N July 8, 2005, the heads of state and government of the G8 meeting in Gleneagles, U.K., announced the launching of the Multilateral Debt Relief Initiative (MDRI), this impact using aggregate data is difficult because coun- tries that receive debt relief are not random; countries with extremely high levels of debt and bad economic performance whereby they agreed to cancel the historical debt of the are more likely to receive it. Therefore, one could find a neg- world’s poorest countries with the International Monetary ative correlation between debt relief and growth in aggregate Fund, the World Bank, and the African Development Bank. data, even if it actually improves the prospects of countries Since multilateral institutions had become these countries’ that receive it. Of course, it may also be the case that coun- main creditors, the initiative, with an estimated debt write-off tries with good economic prospects get debt relief, in which of $50 billion, or about 70% of these countries’ total stock of case finding a positive correlation between debt relief and debt, was expected to provide substantial debt relief (Inter- economic performance does not provide evidence that the for- national Development Agency and International Monetary mer causes the latter. Moreover, even if there is a causal link, Fund, 2006). However, what the broader public probably does debt relief will likely affect economic performance with a not know is that although drawing broader media coverage delay that makes identifying the impact using time series vari- and celebrity attention than its predecessors, the MDRI is ation difficult. For these reasons, evidence based on aggregate just the latest incarnation of a series of efforts to relieve poor data has to either just report correlations or depend on strong countries’ financial obligations with multilateral institutions identification assumptions. through the Heavily Indebted Poor Countries (HIPC) initia- This paper provides new evidence of the impact of multi- tive, set in motion in 1996 and modified (clearly) in 1999 lateral debt relief initiatives using an event study to determine (which I refer to here as enhanced HIPC). the effect of these initiatives on firms with operations in the Humanitarian motives for debt relief are clearly behind countries they benefited. This approach has several advan- numerous calls from religious leaders, celebrities, and intel- tages over studies based on aggregated data because it is lectuals arguing that it is morally wrong to collect debt from much less likely to be affected by reverse causality and takes countries at the brink of starvation. However, the case for debt advantage of the forward-looking nature of stock prices to relief is also typically argued on the grounds that current debt deal with the timing problems. The approach relies on the burdens trap poor countries in a situation of debt overhang, standard assumption that stock prices quickly reflect the mar- where socially and privately profitable investment opportu- ket’s view of the impact of these initiatives on the value of nities are forgone because of the implicit tax on their returns firms that operate in these countries. So if these firms’ values imposed by previous debt obligations (Krugman, 1988; Calvo improve as a consequence of debt forgiveness, there should & Sachs, 1989). Firms anticipating the high future taxes be an abnormal increase in their stock prices around major debt relief announcements. Received for publication March 20, 2009. Revision accepted for publica- Debt relief can have an impact on the value of firms tion June 14, 2010. operating in benefited countries by reducing their expected * World Bank. taxes or raising their expected gross profits. In the absence I am grateful to Shinsuke Uchida for superb research assistance and to Dani Rodrik and three anonymous referees for valuable comments. I am of severe contractions in public expenditure, repaying large also grateful to those South African multinational companies that gently levels of public debt requires high tax revenues. In poor coun- provided information for this project: ABSA Group, Astral Foods, Bell tries, where income taxes are hard to impose, taxes tend to Equipment, Blue Financial Services, Illovo Sugar, and MTN Group. Com- ments from seminar participants at the University of Chile, LACEA, and fall on companies.1 Therefore, as long as markets assign a the World Bank are also gratefully acknowledged. All remaining errors are my responsibility. Financial support from the Japanese CTF is gratefully acknowledged. The views expressed in this paper are my own and do not 1 According to World Bank (2008), sub-Saharan Africa is the continent necessarily represent those of the World Bank, its executive directors, or with the highest overall business taxes, including the highest average profit the countries they represent. taxes. The Review of Economics and Statistics, November 2011, 93(4): 1262–1288 © 2011 by the President and Fellows of Harvard College and the Massachusetts Institute of Technology MULTILATERAL DEBT RELIEF THROUGH THE EYES OF FINANCIAL MARKETS 1263 positive probability to the debt repayment, the relief should significant, with the announcements resulting in a cumu- reduce expected taxation and directly increase the value of lative abnormal return of about 1 percentage point for the firms operating in these countries. Furthermore, debt relief typical parent company. Considering that the parent compa- may also increase these firms’ gross profits because lower nies are usually at least an order of magnitude larger than expected taxation may lead them to increase their own invest- their affected subsidiaries, these results suggest a two-digit ment or because higher expected aggregate demand resulting cumulative impact on the value of local operations. from higher net resource flows or increased incentives for Evidence comparing the various stages of multilateral debt public and aggregate private investment may lead to an relief programs indicates that the MDRI, and to a lesser extent increase in their own demand. If any of these channels is in the enhanced HIPC, had a larger impact on HIPC-connected operation, an event study of the stock price response of pub- multinational companies than the original HIPC initiative. licly traded companies operating in HIPC countries around Also, country-specific announcements of the achievement of dates of debt relief announcements provides an indirect test the various milestones of the broad HIPC program (deci- of the hypothesis that debt relief has a positive effect on these sion and completion points) have little impact on firms’ countries’ economic prospects. returns, although there is some evidence that reports of a Since HIPC countries typically lack well-functioning and country reaching completion point have a positive impact on liquid stock markets, this paper follows Guidolin and La the stock prices of related multinationals. Furthermore, the Ferrara (2007) and studies the response of stock prices of stock price response takes place mainly around the formal multinational firms with subsidiaries and operations on HIPC announcements of the launching of the major initiatives in countries, but that are traded in foreign, more developed the G8 summit meetings, with little evidence of stock reac- financial markets. In particular, it focuses on South African tions around the annual meetings of the World Bank and multinational companies traded on the Johannesburg Stock International Monetary Fund, where the final details of the Exchange (JSE), one of the largest and most active emerg- implementation are disclosed. ing stock markets that also provides meaningful price data Consistent with the interpretation that the increase in stock daily. Focusing on South African multinationals has the addi- prices is related to financial markets’ reassessment of the tional advantage that these companies are smaller than global value of firms operating in countries benefited by debt relief multinationals operating in African HIPC countries. There- programs, the increase in stock returns is larger among South fore, their operations in these countries are relatively more African multinational companies that are relatively more important and their share prices more likely to respond to exposed to the events, as measured by the total employ- events affecting their subsidiaries.2 ment in subsidiaries located in eligible countries as a fraction To implement this approach, I build a detailed chronology of the employment of the parent company. This compar- of the multilateral debt relief initiatives that allows me to iden- ison strengthens the causal interpretation of the findings, tify the dates of different announcements related to the three since it implicitly controls for any potential common effect major initiatives implemented since 1996 (HIPC, enhanced affecting all companies with operations in HIPC countries or HIPC, and MDRI), including the dates when individual coun- multinational firms in general. tries reached any of the milestones considered in the HIPC The stock price response to major debt relief announce- framework (decision points and completion points). I also ments of parent companies in different industries suggests gather stock price data for a sample of 35 South African multi- that the underlying increase in value is mainly related to national companies with 187 subsidiaries and operations in the expectation of lower future taxes rather than improved eighteen African HIPC countries during the period 1995 to economic prospects. Companies in resource extraction indus- 2006, when the various stages of the initiatives took place. tries, which are more likely to be the target of taxation and I use these data to estimate a two-factor return model and are less dependent on local economic conditions, exhibit a measure the abnormal returns of these companies around the significantly larger stock price increase than companies in dates of the announcements, and to formally test the hypoth- service industries, where the relevance of local economic con- esis that these abnormal returns are equal to 0 in applying ditions in relation to taxes reverses. This finding also indicates various parametric and nonparametric procedures. that the stock price response is not a mechanical response to The results show that stock prices of South African multi- potential real exchange rate appreciation associated with debt nationals with operations in HIPC-eligible African countries forgiveness (Rajan & Subramanian, 2005a, 2005b). exhibit an abnormal and statistically significant increase Overall, the evidence presented in this paper suggests that around the announcement dates of major debt relief initia- financial markets view the announcements of major debt tives. The magnitude of the increase is also economically relief initiatives as positive news for firms operating in bene- fited countries, resulting mainly from a reduction in expected 2 Large global multinationals such as Unilever have operations in several taxes or the probability of expropriation, which is consistent African HIPC countries. However, it is unlikely that Unilever’s stock price with the debt overhang hypothesis of the costs of excessive on the London Stock Exchange would vary importantly as a result of events debt. It must be noted that the perceived improvement of affecting a marginal operation in Kenya. In contrast, operations in HIPC countries, while small, are not marginal for South African multinationals economic prospects of firms operating in countries receiving (see table 2). debt relief does not necessarily signal an improvement in the 1264 THE REVIEW OF ECONOMICS AND STATISTICS economic prospects of these countries because the binding Plan to a particular country, which may be correlated with constraint for investment in these countries may not be debt that country’s economic prospects (Kovrijnykh & Szentes, overhang but poor institutions (Arslanalp & Henry, 2006). 2007). Nevertheless, the reduction of one barrier to investment can This paper contributes to this literature by providing indi- still be interpreted as an improvement in economic condi- rect evidence from the firm-level performance of the impact tions, even if that barrier is not the most pressing one at the of multilateral debt relief on HIPC countries that is less moment. Furthermore, the paper shows some anecdotal evi- likely to be contaminated by endogeneity concerns for sev- dence that domestic aggregate demand was also expected to eral reasons. First, it focuses on the announcement of major expand and that foreign investment in African HIPC coun- initiatives that benefited a large set of countries and are tries increased around the announcement of the MDRI, and it unlikely to be motivated by any specific country’s economic did it faster than in the rest of the sub-Saharan African coun- prospects. Second, it exploits the forward-looking informa- tries, which is consistent with debt overhang having some tion contained in the variation in stock prices around specific impact on investment in these countries. event dates and relies on local variation of stock prices at a This paper relates to the empirical literature on sovereign daily frequency to identify the effect of debt relief. Therefore, debt overhang and the impact of debt relief. Several papers the findings are not driven by any existing information on a in this literature have used aggregate data to test the debt country’s economic prospects that was available a few days overhang hypothesis by looking at the relation between debt before the announcements. Finally, the use of firm-level data levels and growth or investment (Claessens, 1990; Desh- also differentiates among firms with ex ante different expo- pande, 1997; Cordella, Ricci, & Ruiz-Arranz, 2005; Imbs sure to the events to provide a further test that controls for & Ranciere, 2005) and have identified different thresholds common shocks. over which debt burden is negatively correlated with growth, From a methodological standpoint, this paper is closely and in some cases further thresholds above which debt has related to Guidolin and La Ferrara (2007), who also use an no growth effect (Cordella, Ricci, & Ruiz-Arranz, 2005). event study to estimate the impact of civil conflict on multina- A slightly different line of research has been followed by tional firms operating diamond mines in Angola. In contrast Depetris-Chauvin and Kraay (2005), who instead of look- to Guidolin and La Ferrara (2007), this paper concentrates on ing at the relation between debt levels and macroeconomic a completely different question and focuses on multination- performance directly study the growth effect of debt relief, als in an indirect manner, using them to gauge the impact of finding little evidence that countries experiencing relatively debt relief on local firms. larger reductions in their stock of debt tend to grow faster. The rest of the paper is structured as follows. Section II While these papers provide interesting results, their reliance gives an overview of multilateral debt relief to poor coun- on aggregate data exposes them to the econometric problems tries and presents a chronology of the initiatives that is used arising from reverse causality and from debt relief not being to identify the dates of various announcements. Section III randomly assigned to countries, and that they have to rely describes the methodological approach and data. Section IV on various econometric techniques and strong identification presents the results. Section V concludes. assumptions to move from correlation to causality. Indirect but stronger evidence on the impact of debt relief was provided by Arslanalp and Henry (2005), who also use an event study approach to show that the stock market II. Multilateral Debt Relief to Poor Countries indexes of countries benefited by the Brady Plan significantly A. A Brief History of Multilateral Debt Relief to Poor increased relative to a control group after the announcement Countries of the plan.3 However, since HIPC countries typically lack stock markets, Arslanalp and Henry (2005) do not apply Historically, countries that became part of the group of their methodology to estimate the impact of the HIPC ini- HIPC had little access to commercial lending and relied tiative and rely instead on indirect arguments to conjecture instead on official financing in the form of bilateral loans that HIPC countries are unlikely to benefit from the type of from industrial countries and multilateral loans from institu- debt relief provided by the Brady Plan because investment tions such as the IMF, the World Bank, and various regional in these countries is constrained not by debt overhang but by development banks. Official loans to these countries grad- bad institutions (Arslanalp and Henry, 2006). An additional ually increased until the early 1980s, when many started concern with this paper is that they look only at aggregate having problems servicing their debts, at the same time country-level indicators, so their estimates may be contami- as several middle-income countries did. However, whereas nated by the endogeneity of the decision to extend the Brady middle-income countries that defaulted on their commercial loans were shut out of international capital markets, industrial 3 The Brady Plan was a program for restructuring the debt of Latin Amer- countries’ governments and multilateral institutions reacted ican countries after the debt crisis of the 1980s. The plan was articulated to the debt problems of low-income countries by reschedul- by U.S. Treasury Secretary Nicholas F. Brady and roughly consisted of the conversion and rescheduling of syndicated bank loans into bands—the ing payments and extending bilateral and multilateral loans so-called Brady Bonds. See Arslanalp and Henry (2005) for further details. that would allow these countries to avoid defaulting. For MULTILATERAL DEBT RELIEF THROUGH THE EYES OF FINANCIAL MARKETS 1265 this reason, and in contrast to most middle-income coun- criticism from international aid groups and African govern- tries, low-income countries maintained a positive net resource ments that were calling for substantial modifications to the transfer during the 1980s (Birdsall & Williamson, 2002). initiative, and a consensus emerged among industrial coun- This additional lending to poor countries was not typically tries for faster implementation of debt relief. As a result, accompanied by growth, resulting in further debt accumula- the leaders of the G7 countries, meeting in Cologne in June tion as a fraction of GDP. By the late 1980s, most low-income 1999, announced a comprehensive review on the HIPC ini- countries exhibited symptoms of unsustainable debt levels, tiative to provide faster, deeper, and broader debt reduction with debt-to-export and debt-to-GDP ratios close to 400% in what became known as the Enhanced HIPC Initiative and 150%, respectively. At that stage, it became evident that (E-HIPC). The main changes were a broadening of the eligi- low-income countries were unable to fully serve their offi- bility criteria by reducing the debt-to-export ratio to 150%, cial debt and that some form of broad debt forgiveness was and shortening the time required to reach the decision point required. to three years. Moreover, a country reaching a decision point Systematic debt relief to poor countries initially took place under the enhanced HIPC would immediately receive some by bilateral loans to Paris Club creditors under what became debt relief in the form of reduced debt service. Debt stock known as the Toronto Terms, the Trinidad Terms, the London reductions would take place once the country reached the Terms, and the Naples Terms, all which provided reschedul- completion point. The enhanced initiative also put emphasis ing under concessional (below-market) interest rates equiva- on a country’s commitment to poverty reduction in two ways: lent to a reduction in the net present value of the debt stock first, in addition to a good policy track record, a country had of about $30 billion.4 At the same time, new bilateral flows to submit a sustainable poverty-reduction strategy to become increasingly started taking the form of grants. As a result, eligible, in the form of a poverty-reduction strategy paper an increasing fraction of the debt of low-income countries (PRSP), written with the participation of nongovernmental was owed to multilateral institutions, and it was apparent that organizations, social movements, labor unions, and similar helping these countries achieve debt sustainability required other organizations; second, countries had to commit to use some form of relief from multilateral debt, which had histor- the resources freed by debt relief to achieve the goals set in ically being treated as senior to all other claims and repaid in the PRSP. Because of the broader eligibility criteria and pub- full, even if by rolling over old loans.5 lic pressure, sixteen additional countries were approved for Multilateral debt relief to poor countries started with the a decision point and started receiving debt relief in the year launching of the Heavily Indebted Poor Countries (HIPC) in 2000. September 1996 at the 22nd meeting of the G8 countries in The HIPC initiative emphasized the reduction of debt bur- Lyon, France. The goal of this initiative was to reduce the dens to sustainable levels to help benefit countries fighting debt burden of eligible countries to levels considered man- poverty, but the view that multilateral debt cancellation was ageable, conditional on satisfactory policy performance, and the only possible solution to the problems of HIPC countries involved cooperation among multilateral and bilateral cred- became increasingly popular shortly after the announcement itors. Under the initiative, debtor countries with per capita of the enhanced version of the initiative. As a result, a income under $695 and ratios of net present value of debt broad campaign was launched to convince leaders of indus- to exports above 200% or 250% (depending on country trial countries—those with most voting rights on multilateral characteristics) would qualify for the program.6 Qualifying institutions—to provide some form of debt forgiveness. This countries with six years of stable macroeconomic conditions campaign culminated in the announcement of the Multilat- under an IMF program would reach a decision point, whereby eral Debt Relief Initiative (MDRI) at the Gleneagles Summit creditors would arrange a debt relief package, and after no Meeting of the heads of state and government of G8 coun- more than three additional years of successful policy imple- tries in July 2005. The goal of this initiative is to further mentation, they would reach a completion point, when they reduce the debts of HIPC countries and help them achieve the would start receiving debt relief. Contrary to initial expec- Millennium Development Goals set by world leaders in Sep- tations, only six of the forty countries that were eligible for tember 2000, during the United Nations Millennium Summit. HIPC relief had reached a decision point in 1999, and only Although the initiative operates similarly to the HIPC, its one, Uganda, had reached completion point. main difference is that it contemplates that once a coun- The slow advances of debt relief under the original HIPC try completes the HIPC process (that is, reaches completion initiative, mainly the result of eligibility conditions, led to point), all debt it contracted with the IMF, the World Bank, and the African Development Bank before 2003–2004 would be forgiven.7 4 See Daseking and Powell (1999). 5 There is no consensus on whether multilateral institutions engage in defensive lending. While some authors argue that this is the case (Bulow & Rogoff, 2005), others have found no robust evidence of such behavior in 7 Unlike the HIPC Initiative, the MDRI is not comprehensive in its creditor the data (Geginat & Kraay, 2007). coverage and does not involve participation by official bilateral or commer- 6 The ratio was replaced by 280% of government revenue for very open cial creditor, or of multilateral institutions other than the above-mentioned economies. three. 1266 THE REVIEW OF ECONOMICS AND STATISTICS Table 1.—Chronology of HIPC Initiatives and MDRI June 27, 1996 G8 summit in Lyon, France. The proposal of the HIPC initiative was accepted, except the idea of IMF’s gold sales to finance debt relief, which Germany strongly opposed. Options for financing are yet to be finalized. September 29–30, 1996 Annual meetings of the IMF and World Bank. The HIPC initiative is approved by the IMF and the World Bank. Agreement by the G7 nations to increase debt cancellation from 67% (decided in Naples Terms in 1994) to a maximum of 80% is followed by the approval of the initiative. June 18, 1999 G8 Summit in Cologne, Germany. An amendment of the HIPC Initiative is agreed. September 26, 1999 Annual meetings of the IMF and World Bank. Enhanced HIPC approved by the IMF and the World Bank. July 8, 2005 G8 Summit in Gleneagles, Scotland. As much as 100% debt cancellation for the HIPCs owed to the World Bank, the IMF, and the African Development Bank is committed by G8 nations. September 26, 2005 Annual meetings of the IMF and World Bank. The IMF and the World Bank agree to endorse 100% debt cancellation for the HIPCs. Sources: Deutsche Presse-Agentur, June 28, 1996; Financial Times (London), September 30, 1996; Independent (London), June 19, 1999; World Markets Analysis, July 11 and September 27, 2005; Agence France Presse, September 25, 2005. B. Multilateral Debt Relief Event Dates and July 8, 2005, for the MDRI. (Results considering the September dates will be discussed in the robustness analysis This paper’s analysis separately considers two types of in Appendix A.) multilateral debt relief events: (a) those common to all eligible The process for countries reaching decision and comple- countries, such as announcements of the different stages of tion points under the HIPC initiative also entails various steps, the HIPC initiative and the MDRI (henceforth labeled major such as the preparation of a debt sustainability analysis and initiatives); and (b) those that benefit an individual HIPC a series of World Bank and IMF discussions of whether a country, such as the announcement that a nation has reached country meets or is progressing toward the conditions for each a decision or completion point. milestone. However, since most of these discussions are tech- The brief historical discussion evidences that major ini- nical and take place within multilateral institutions, I consider tiatives are typically a matter of lengthy discussions, so a as event dates for these country-specific events the day when choice has to be made regarding the relevant announcement the board of governors of the World Bank and International dates. For all three major initiatives, the announcement pro- Monetary Fund officially announces that a country reaches cess typically entails three stages. During the first stage, there any of these milestones, as documented in the Country Report are numerous requests for debt relief that place the discussion Documents of the HIPC initiative.8 of an initiative on the agenda of a G8 summit meeting. In the Figure 2 shows the distribution of multilateral debt relief second stage, which takes place during or shortly before the events, including the announcement of major initiatives as summit, the finance ministers of the G8 countries agree on well as decision and completion points for benefited coun- the details of the forthcoming initiative. Finally, the heads of tries. Just a few countries reached decision or completion state and government of G8 countries formally announce the points between 1996 (the year of the launching of the origi- initiative during the summit meeting, with the exact details of nal HIPC initiative) and 1999, but a clear cluster of countries financing, implementation, eligibility, and so on, to be worked reached decision point shortly after the announcement of the out in the coming months and disclosed during the annual enhanced HIPC initiative. meetings of boards of governors of the World Bank and the International Monetary Fund. The chronology of these stages for the three major multilateral debt relief initiatives is sum- III. Methodology and Data marized in table 1. (A detailed chronology of the initiatives, Under the assumption that the stock returns of multina- including all the discussion meetings, is available in the Table tional parent companies operating in HIPC countries respond Appendix.) to events affecting their subsidiaries and that the value of For each major debt relief initiative, the event dates are these subsidiaries is not negatively correlated with the eco- selected following a semi–de facto approach by choosing nomic performance of the host country,9 a standard event either the day of the G8 finance ministers’ meeting or the study that quantifies the impact of multilateral debt relief day of heads of state meeting, depending on which day has announcements on the stock prices of multinational compa- the higher amount of news related to the initiative reported nies operating in HIPC-eligible countries provides indirect by the international press according to Factiva Newsplus. The amount of this news around those days and around the days of 8 Available at http://www.imf.org/external/np/hipc/index.asp. the annual meetings of the World Bank and IMF is presented 9 The correlation between the growth in market value of parent companies in figure 1. It shows that the announcements made following and the average growth of countries where there are subsidiaries is in fact positive and significant (0.16, significant at the 1% level). Furthermore, a the meetings of the heads of state and government of G8 coun- regression of the annual growth of a firm’s market capitalization against the tries are those that receive the most press coverage. Therefore, average growth rate of African HIPC countries where it has subsidiaries, the following days are considered benchmark announcement controlling for firm fixed effects, yields a coefficient of 3.26, also signifi- cant at the 1% level. The sign and significance even survive after adding dates for the major initiatives: June 27, 1997, for the origi- year fixed effects to the specification in growth rates (coefficient of 2.57, nal HIPC initiative; June 18, 1999, for the enhanced HIPC; significant at 5%). These results are not reported but are available on request. MULTILATERAL DEBT RELIEF THROUGH THE EYES OF FINANCIAL MARKETS 1267 Figure 1.—Distribution of News Related to Major Debt Relief Announcements The different panels exhibit the number of press articles on the various multilateral debt relief announcements during the days around the G7 (G8) summit meetings where the initiatives were launched and around the annual meetings of the World Bank and IMF where the details of implementation where released. The exact days of announcements are shown by the vertical lines (June 27, 1996, June 18, 1999, and July 8, 2005). evidence of the overall impact of multilateral debt relief on dummy that indicates whether a multilateral debt relief event the economic prospects of receiving countries. This paper benefiting a subsidiary of a parent company i occurs at time t . implements such a study focusing on the stock prices of South It takes the value 1 when t equals τ and 0 otherwise. The time African multinationals operating in African HIPC-eligible index τ takes values between t1 (greater than t0 ) and t2 , the countries and testing the hypothesis that these announce- beginning and end of the event window in calendar time. The ments convey positive news for these companies. To this end, variable i,t is an error term correlated across firms on a given the parameters of the following augmented two-factor return day but assumed independent across days. The parameters α, model are estimated by OLS: β, γ, θ, and δ are coefficients to be estimated. The coefficients of interest are the δτ associated with the event time dummies, L which capture the average abnormal return across companies Ri,t = αi + βi RtM + γi RiI,t + θli Sil,t for each day of the event window. The abnormal returns for l=1 each firm during the event and estimation windows corre- t2 spond to the difference between the actual returns and those + δτ Dτ,t + i,t , t ∈ [t0 , t2 ], (1) predicted by the return model (1): τ=t1 L where Ri,t is the stock return of company i between trading ARi,t = Ri,t − α ˆi RM − γ ˆi + β ˆi RiI,t − ˆl S l . θ (2) t i i,t days t −1 and t ; RtM and RiI,t are the market return and the return l =1 of company i’s industry during the same period, respectively; Sil,t is a dummy variable that controls for the impact of cor- Under the hypothesis that the event under study has a pos- porate events and takes the value 1 if corporate event type l itive impact on parent companies’ returns, the δτ coefficients affected company i on day t . The variable Dτ,t is an event time should be significantly positive around or immediately after 1268 THE REVIEW OF ECONOMICS AND STATISTICS Figure 2.—Distribution of HIPC Event Dates by Country The markers in the figure show the dates of announcements of decision and completion points for African HIPC-eligible countries with investment from South African multinationals, as well as the announcements of the three major debt relief initiatives (HIPC, enhanced HIPC, and MDRI). The dates of the events affecting each of the countries listed in the y-axis are represented by the markers that appear in the row corresponding to each country. Decision and completion points under the original HIPC initiative are represented by + and ◦, while × and • represent, respectively, decision and completion points under the enhanced HIPC initiative. Vertical lines correspond to the announcements of the three major initiatives. The solid lines correspond to the dates of the G7 (G8) summit announcements and the dashed lines to the final endorsement dates. the event date, and the cumulative abnormal return (CAR), akin to the estimator suggested by Donald and Lang (2007) defined as for clustered panel data. Other than correcting for this cluster- t ing and for potential heteroskedasticity, all parameters across the paper are estimated by OLS. CARt = δ ˆ τ , τ ∈ [t1 , t2 ], (3) In the benchmark results, the parameters of the model are τ=t1 obtained using estimation and event windows of 180 and should be significantly increasing during the event window. 15 calendar days before each event, roughly corresponding The event study literature typically privileges the analysis of to 112 and 10 market trading days, respectively, and using the CAR when there is uncertainty about the exact date of lumped returns,11 but results for different estimation win- the event; this convention is followed in the rest of the paper. dows, event windows, and using trade-to-trade returns to Another advantage is that the cumulative impact of the events control for thin trading were also computed as robustness is easier to visualize.10 checks. Furthermore, the results were also checked com- The econometric model described in equation (1) directly puting the abnormal returns using the three-factor model deals with the clustering of the events under consideration of Fama and French (1993) that replaces the industry-level in calendar time (and the potential cross-firm correlation return RiI,t with the differential return between small and of returns) by allowing the error term i,t to have a calen- large companies and the differential return of high versus low dar time component, exploiting the large number of clusters book-to-market companies (the so-called SMB and HML risk (calendar days) in the data to obtain asymptotically correct factors). inference. Nevertheless, the standard errors of the average Nonparametric tests of the hypothesis that the announce- abnormal returns are also estimated using the alternative port- ments have a positive impact on parent company return have folio approach that exploits only their time series variation better power when abnormal returns are not normally dis- over the estimation window (Brown & Warner, 1985) and is tributed, as may be the case when trading is infrequent, also called thin-trading (Corrado, 1989; Campbell & Wasley, 10 As it is standard in the literature, inference is conducted under the 1996). Two nonparametric tests frequently used in the litera- assumption that abnormal returns are serially uncorrelated. From a the- ture are the Corrado (1989) rank test, and the Cowan (1992) oretical perspective, this assumption is grounded on the unpredictability of returns in efficient financial markets. From an empirical perspective, stan- dard tests of serial independence of average abnormal returns do not reject 11 This means that during periods of inactivity, all returns are assigned to this hypothesis in the data used in this paper. the first day in which there is new trading. MULTILATERAL DEBT RELIEF THROUGH THE EYES OF FINANCIAL MARKETS 1269 generalized sign test. The Corrado rank test is based on the ˆ is the empirical probability of a nonnegative abnormal p following statistic Tt : return; and σ(G ¯ ) is the standard deviation of the mean G ¯t , which should be 0 under the null hypothesis that the event K¯t Tt = has no impact on returns. This statistic is less sensitive than σ(K¯) Corrado’s rank test to changes in variance within the event 1 N Kit − 0.5 × (t2 − t0 + 1) window and to thin trading (Cowan, 1992) and comple- i=1 = N ments the evidence obtained from the former. The cumulative 2 1 t2 1 N Kit − 0.5 × (t2 − t0 + 1 mean sign deviations and their standard deviations reported (t2 −t0 ) t =t0 N i=1 in various figures are computed in the same manner as the N (0, 1), cumulative mean rank deviations. Ki,t = rank (ARi,t ) t ∈ [t0 , t2 ], Three aspects of the empirical strategy deserve further dis- (4) cussion. First, the impact that an event on a subsidiary has on the value of its parent company depends on the fraction of where 0.5 × (t2 − t0 + 1) is the expected value of the rank. the parent’s value represented by the subsidiary. If the sub- The numerator K ¯ t is the mean rank deviation of abnormal sidiary is small relative to the parent, even a large change in returns at event time t , and σ(K ¯ ) is the standard deviation its value will result in a small change in the parent’s returns of this mean. Under the assumption that abnormal returns that may be hard to separate from normal return fluctuations are independent across time, this statistic follows a standard using statistical procedures. This means that tests based on normal distribution, and the hypothesis that the median rank the response of parents’ returns could have low power, which deviation of the abnormal returns at a given time τ, K ¯ τ , is would tend to bias the results against finding a significant statistically different from 0 can be tested by applying stan- abnormal return as a result of the events under considera- dard normal critical values to the statistic Tτ .12 In addition to tion. This is precisely the reason for this paper’s focus on reporting the mean rank deviations and their significance, the South African multinationals with operations in HIPC coun- figures that follow will show the cumulative mean rank devia- tries, which are smaller than other multinationals operating tion computed by adding the mean rank deviations during the in these countries. For instance, the median assets of South event window and whose standard deviation is obtained under African multinationals operating in Ghana were about $8 bil- the assumption that mean rank deviations are i.i.d. (Campbell lion in 2006. In contrast, the largest multinationals operating and Wasley, 1996).13 in Ghana in various industries are Royal Dutch Shell, Bar- The Cowan (1992) generalized sign test follows a similar clays, and Nestle, all with hundreds of billions of dollars in logic as the rank test, but focuses instead on the mean sign assets. Although data on the value of assets of subsidiaries are deviations of the abnormal returns G ¯ t to build a normally typically unavailable, total employment in HIPC countries is distributed statistic St as about 7% of the employment in South African parent compa- ¯t Nt nies on average (median of 3%). Thus, these subsidiaries are G i=1 Git − Nt pˆ St = = √ N (0, 1), (5) about one order of magnitude smaller than their parents but σ(G¯) ˆ (1 − p Nt p ˆ) not negligible (see table 2). Furthermore, the reduced power 1 if ARi,t > 0 of standard estimation can be significantly improved by com- Gi,t = , plementing the standard results of the nonparametric tests. 0 otherwise Second, event studies rely on the efficient market assumption N t1 1 1 that news that has an impact on the value of the firm is quickly ˆ= p Gi,t , (6) incorporated in stock prices. This assumption requires trans- N i =1 Mi t =t0 parent and liquid stock markets. While the JSE is smaller where N and Nt are the total number of firms and the number in absolute terms and more illiquid than stock markets in of firms active at time t , respectively; Mi is the number of developed countries, it is one of the largest emerging stock nonmissing returns of firm i during the estimation period; markets, with a market capitalization of 1.6 times its GDP— much larger than that of countries like the United States, and 12 The results reported below follow the modification of Campbell, Cowan also one of the most liquid, with a market turnover value and Salotti (2009) modification to the Corrado (1989) test to allow for similar to that of Singapore. Moreover, South Africa fares missing returns. well among emerging markets in terms of investor rights and 13 Inference based on the i.i.d. assumption may fail because at the firm level, ranks are serially dependent by construction. A recent literature has corporate governance indexes, with a creditor rights index built some tests that address this issue (Kolari & Pynnonen, 2008). However, of 3 out of 4 according to Djankov, McLiesh, and Shleifer the serial correlation at the individual level does not need to translate in serial (2007) and an active program to improve corporate gover- correlation for the average ranks. In fact, in the results reported below, the serial correlation of average ranks is negligible, and the hypothesis of serial nance. Finally, from a more technical perspective, all the tests independence cannot be rejected at conventional levels, so the multiday discussed rely on asymptotic distributions of the statistics, so tests based on serial independence (such as those for the cumulative mean they require a large number of event firms to be valid. This rank deviations) have only a small bias. Moreover, this potential problem does not apply to the single day tests reported below or to the sign statistics should not be a problem when pooling all events together, but discussed next. may become relevant when trying to disentangle the impact 1270 THE REVIEW OF ECONOMICS AND STATISTICS Table 2.—Summary of Investment of Parent Companies by that have been eligible for different stages of the HIPC and HIPC Countries MDRI, the number of parent companies with subsidiaries in (1) (2) (3) each of these countries, the median assets of these parent Median Average Assets of Number of companies, and the average number of subsidiaries per par- Number of Parent Subsidiaries ent in each of these countries are reported in table 2. The Parent Companies per Parent four countries with the largest number of parent companies HIPC Country Companies (US$ million) Company investing are Zambia, Mozambique, Malawi, and Tanzania. Angola 7 1, 491 1.43 For instance, 24 of the 35 parent companies have subsidiaries Ivory Coast 1 14, 435 1.00 Cameroon 1 14, 435 1.00 in Zambia. However, this does not mean that Zambia could Congo, Democratic Republic 4 45, 262 1.50 drive all results since there are only 4 parent companies that Congo, Republic 1 14, 435 1.00 invest only in this country. The parent companies have median Equatorial Guinea 1 15, 365 1.00 Ethiopia 1 1, 852 1.00 assets of $2 billion, but these vary substantially across coun- Ghana 8 3, 058 1.38 tries. There is a negative correlation between the number Guinea 2 5, 409 1.00 of parent companies and the median values, since only the Kenya 7 4, 039 2.14 Madagascar 1 1, 481 1.00 largest multinationals have operations in smaller and poorer Malawi 16 1, 667 1.25 countries. On average, each of the parents has about 5.34 Mali 3 1, 491 2.00 subsidiaries per country. Mozambique 22 1, 672 1.77 Rwanda 1 14, 435 1.00 The date of initial investment in subsidiaries located in Tanzania 16 1, 672 1.25 African HIPC countries reveals whether a parent company Uganda 6 9, 237 1.33 is affected by a specific debt relief event. For instance, only Zambia 24 1, 486 1.79 All countries 35 2, 078 5.34 companies with active subsidiaries in African HIPC-eligible Sources: McGregor (2006). countries in June 1996 are considered as affected by the orig- inal HIPC initiative announcement. Table 3 lists the parent of single events, especially the original HIPC when there are companies with investments in African countries that have only ten parent companies treated. Thus, while the results been eligible for multilateral debt relief since 1996, the num- for single events are informative, the inference conducted in ber of their subsidiaries in African HIPC countries, total those with a small number of treated parent firms (HIPC and, employment in these subsidiaries, and the ratio of this total to some extent, the enhanced HIPC) should be taken with employment to that of the parent company for each of the caution.14 three years with an announcement of a major initiative (1996, Information on parent-subsidiary relations comes mainly 1999, and 2005). The last three columns show the median from McGregor (2006), which reports all South African com- size of the parents’ assets and sales at the end of 2006, panies with operations in other African countries at the end and the main industry of the parent company, respectively. of 2006. This publication enumerates the subsidiaries and The three bottom rows of the table display the average of operations of each South African multinational and reports each measure across companies for each announcement year, partial information on the date of initial investment, hold- across all years (labeled “Overall”), and the medians across ings, and number of employees in each subsidiary. From it, all years, respectively. Both assets and sales of parents are all firms listed in the JSE, and with operations or subsidiaries about $2 billion, and the average parent company has about in African HIPC-eligible countries during the period of the four subsidiaries. Parent companies with interests in HIPC initiatives (1995–2006) were selected, to obtain a sample of countries are also homogeneously distributed across indus- 35 companies with 187 subsidiaries in eighteen countries. tries (column 12). Utilities are the only industry where no This information was complemented by and checked with South African company has affiliates in other African HIPC data from United Nations Conference on Trade and Devel- countries.16 opment (1993, 2004), Graham and Whiteside (1997–2004), Stock returns and corporate-event data for the selected par- Lexis-Nexis (2007), business press reports relating parent and ent companies were obtained from Bloomberg. The market subsidiaries in Lexis-Nexis and Factiva-Newsplus, and infor- return is based on the JSE All Shares Index, and the industry mation requests sent directly to South African companies return associated to each parent is that of the FT JSE Index identified as having affiliates in eligible African HIPC coun- of the industry of the primary activity of the parent company, tries.15 The list of the eighteen sub-Saharan African countries both also obtained from Bloomberg. The fraction of parent companies with available return data increased during the period, but even in 1996, there are return data available for 14 A recent literature focuses on methods to produce valid inference in more than 70% of the firms. Corporate event data included cases with even a single treated firm that rely on the use of bootstrapping to compute small sample statistics. Since the point estimates are not encour- in Sil,t comprise the following corporate action types: capital aging for the case in which small samples may be an issue (the original HIPC), this approach is not further pursued in this paper. 15 A letter was sent to each parent company with missing information requesting data on initial investment date and size of subsidiaries, reaching 16 The spatial distribution of parent companies by country is reported in a response rate of about 30%. the Table Appendix. Table 3.—Summary Statistics for Parent Companies (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) 1996 1999 2005 No. Subs. No. Subs No. Subs. Annual African Employ. African Employ. African Employ. Assets Sales HIPC African Ratio HIPC African Ratio HIPC African Ratio (US$ (US$ Parent Countries HIPC Subs. Employ. Countries HIPC Subs. Employ. Countries HIPC Subs. Employ. million) million) Industry Aveng 0 – – 0 – – 9 – – 1,491 2,240 Basic industries Alexander Forbes 0 – – 3 48 0.01 7 123 0.02 19,528 813 Financials Aeci 2 187 0.01 2 187 0.02 6 303 0.04 1,161 1,126 Basic industries African Life Assurance Co. 0 – – 0 – – 9 461 0.15 2,078 484 Financials Afgri 0 – – 0 – – 1 47 0.01 915 794 Noncyclical consumer goods African Oxygen 0 – – 2 131 0.01 4 151 0.05 590 505 Basic industries Anglogold Ashanti 0 – – 1 473 0.01 8 14,593 0.23 9,326 2,999 Resource African Rainbow Minerals 0 – – 9 – – 13 – – 2,176 688 Resource Astral Foods 0 – – 0 – – 1 85 0.01 324 772 Noncyclical consumer goods Absa Group 0 – – 1 441 0.01 2 1,005 0.03 73,743 8,109 Financials Barloworld 3 380 0.01 5 503 0.02 8 550 0.02 5,310 6,359 General industries Business Connexion Group 0 – – 0 – – 3 – – 356 478 Information technology Bell Equipment 1 55 0.03 2 55 0.03 3 95 0.04 304 526 General industries Bhp Billiton 0 – – 1 542 0.02 1 542 0.02 7,226 4,789 Resource Bidvest Groupco. 0 – – 9 481 0.01 11 481 0.01 4,170 11,510 Cyclical services First Rand 0 – – 0 – – 2 97 0.00 86,355 10,719 Financials Gold Fields 1 1,939 0.04 1 1,939 0.04 2 2,616 0.05 4,712 2,175 Resource Grindrod 0 – – 7 13 0.01 9 13 0.00 1,083 1,863 Cyclical services Illovo Sugar 0 – – 3 5,621 0.29 4 7,267 0.50 690 888 Noncyclical consumer goods Imperial Hld. 4 – – 4 – – 4 – – 5,594 8,059 General industries Massmart Hld. 0 – – 0 – – 3 375 0.02 1,433 4,463 Cyclical services Mtn Group 0 – – 2 293 0.11 5 631 0.08 14,435 4,598 Noncyclical services Metorex 0 – – 1 452 0.41 1 452 0.11 242 227 Resource Nedbank Group 0 – – 0 – – 1 82 0.00 63,287 7,060 Financials Nampak 3 284 0.02 5 620 0.04 12 1,366 0.08 1,852 2,311 Cyclical Services Sabmiller 6 1,553 0.02 9 1,812 0.03 14 2,060 0.03 4,039 1,844 Noncyclical consumer goods Standard Bank Group 6 1,688 0.05 7 2,169 0.07 8 2,255 0.06 144,324 12,216 Financials Steinhoff International Hld. 0 – – 0 – – 2 – – 4,745 4,802 Cyclical consumer goods Shoprite Hld. 1 – – 2 125 0.00 10 895 0.01 1,481 4,991 Noncyclical services Santam 0 – – 1 – – 2 99 0.04 2,505 1,728 Financials Sasol 0 – – 2 447 0.02 5 739 0.02 15,365 12,272 Resource Sun International 0 – – 0 – – 1 379 0.05 1,119 886 Cyclical services Tongaat-Hulett Groupco. 1 3,372 0.16 1 3,372 0.28 1 3,372 0.37 1,349 1,118 Noncyclical consumer goods Tourism Investment 0 – – 0 – – 1 – – 138 196 Cyclical services Unitrans 0 – – 0 – – 3 – – 756 1,996 Cyclical services MULTILATERAL DEBT RELIEF THROUGH THE EYES OF FINANCIAL MARKETS Average 3 1,182 0.04 3 986 0.07 5 1,469 0.07 Overall average 4 1,256 0.07 13,834 3,617 Overall median 3 467 0.03 2,078 1,996 The table reports summary statistics for each parent company in each of the years of a major announcement of a debt relief intiative. Columns 1, 4, and 7 report the number of subsidiaries in African HIPC countries in 1996, 1999, and 2005, respectively. Columns 2, 5, and 8 report the total employment in these subsidiaries, adjusted for the holdings of the parent company, for the same three years. The adjustment is such that if the subsidiary is fully owned, all employment is counted, but if it is partly owned, the number of employees imputed are the total in the subsidiary times the fraction owned by the parent. Columns 3, 6, and 9 present the ratio of total employment in African HIPC countries to total parent employment. Columns 10 and 11 report total assets and annual sales of the parent companies from the latest available of the 2005 or 2006 fiscal year. Column 12 shows the industry group to which each parent company belongs. Source: Bloomberg and Worldscope databases, and McGregor (2006). 1271 1272 THE REVIEW OF ECONOMICS AND STATISTICS Table 4.—Evolution of Cumulative Abnormal Returns: Three Major Initiative Events All Major Events HIPC Initiative Enh. HIPC Initiative MDRI Standard Standard Standard Standard Event Day CAR Error CAR Error CAR Error CAR Error −10 0.18 0.27 1.68∗∗∗ 0.09 −0.16 0.14 −0.05 0.07 −9 0.18 0.29 2.07∗∗∗ 0.17 −0.12 0.17 −0.19∗ 0.11 −8 0.15 0.31 2.39∗∗∗ 0.20 −0.35 0.29 −0.19 0.13 −7 0.03 0.42 2.24∗∗∗ 0.23 −1.02∗∗∗ 0.35 0.06 0.14 −6 0.11 0.43 2.27∗∗∗ 0.28 −1.02∗∗∗ 0.40 0.21 0.17 −5 0.32 0.59 2.10∗∗∗ 0.32 0.10 0.47 −0.05 0.19 −4 0.21 0.61 1.79∗∗∗ 0.36 −0.05 0.57 −0.07 0.23 −3 0.09 0.62 1.92∗∗∗ 0.40 −0.31 0.60 −0.16 0.25 −2 0.57 0.67 1.68∗∗∗ 0.46 0.71 0.65 0.17 0.28 −1 0.99 0.73 1.61∗∗∗ 0.51 1.78∗∗∗ 0.71 0.31 0.31 0 0.93 0.76 0.94∗ 0.55 1.54∗∗ 0.75 0.53 0.34 1 1.39∗ 0.80 1.07∗ 0.60 1.61∗ 0.83 1.35∗∗∗ 0.36 2 1.32 0.83 1.07∗ 0.63 1.21 0.87 1.48∗∗∗ 0.39 3 1.12 0.84 1.45∗∗ 0.70 0.85 0.91 1.27∗∗∗ 0.42 4 1.43∗ 0.86 1.02 0.75 1.25 0.96 1.76∗∗∗ 0.45 5 1.35 0.88 1.37∗ 0.80 0.99 1.02 1.68∗∗∗ 0.48 6 1.26 0.90 1.62∗ 0.84 1.01 1.09 1.42∗∗∗ 0.51 7 1.39 0.92 1.73∗∗ 0.88 1.26 1.14 1.50∗∗∗ 0.54 8 1.17 1.00 1.74∗ 0.94 0.26 1.21 1.70∗∗∗ 0.57 9 1.11 1.02 1.89∗ 0.99 −0.13 1.30 1.79∗∗∗ 0.60 10 1.17 1.05 1.04 1.02 0.14 1.37 1.96∗∗∗ 0.63 Number of firms 68 10 23 35 Columns labeled CAR report the cumulative abnormal returns of South African multinational companies with operations in countries eligible for each stage of the HIPC and MDRI initiatives for each trading day within a ten-day event window. CAR are based on abnormal returns estimated from a two-factor model using an estimation window of 112 trading days. Standard errors correct for the clustering of the events in calendar time. Column 1 reports results obtained by pooling all major initiatives together, and columns 2 to 4 separately present similar results for the HIPC initiative, the enhanced HIPC initiative, and the MDRI. Event dates considered for each of these three major initiatives correspond to the dates of the G8 summits held on June 27, 1996, June 18, 1999, and July 8, 2005. “Number of firms” reports the number of companies over which abnormal returns are computed in each occasion. ∗∗∗ , ∗∗ , and ∗ denote 1%, 5%, and 10% significance levels, respectively. changes, corporate events, and distributions, all as defined by of about 1 percentage point (figure 3A). While cumulative Bloomberg.17 abnormal returns are 10 basis points three days before the announcement, they increase to 140 basis points one day IV. Results after the announcement (column 1 of table 4). This three- day increase is equivalent to more than 200% on an annual This section presents the results of the event study analy- basis. Considering that parent companies are large relative sis of the impact of multilateral debt relief announcements. to their subsidiaries, this increase is economically signifi- It first describes the results obtained for the major debt cant and suggests a much larger increase in the underlying relief initiatives, followed by those obtained for country- value of the affiliates. For instance, if a parent company’s level debt relief announcements (decision and completion operations in all HIPC countries represent 10% of its over- points). Results from parametric and nonparametric tests and all value (roughly the average ratio of subsidiary to parent robustness analysis are discussed in each case. employment in the data), a 1% increase in the value of the parent company is consistent with a 10% increase in the value A. Major Debt Relief Initiatives of those operations.18 The results therefore indicate that the announcement of major debt relief initiatives conveys pos- The evolution of the estimated cumulative abnormal return itive news for South African multinational companies with CARτ for the three major debt relief initiatives, in an event affiliates in eligible African HIPC countries that translate to window of ten trading days around the event date is reported an abnormal increase in their share returns. in table 4 and depicted in the various panels of figure 3, Figures 3B to 3D show heterogeneity in the response to along their 90%, confidence bands. Figure 3A shows the the different major debt relief initiatives. The announcement CAR obtained by pooling the three major initiatives, and of the original HIPC initiative does not have a significant Figures 3B to 3D display the CAR separately estimated for impact on the returns of parent companies (figure 3B). While the HIPC initiative, enhanced HIPC initiative, and MDRI, these companies had significantly positive abnormal returns respectively. at the beginning of the event window (two weeks before The announcement of a major debt relief initiative is asso- the event), there is no upward break in the CAR around ciated with a statistically significant increase in the CAR 17 The following specific corporate actions are included: Acquired, Acqui- sition, Cash Dividend, Corporate Meeting, Debt Offering-New Issue, 18 This is because the total value of the multinational equals the sum of De-listing, Divestiture, Equity Offering, Fiscal Year End, Change ID, Num- the value of their subsidiaries. Thus, a percentage increase in the value of ber Change, Listing Name, Change Par Value, Change Rights Offerings, a subsidiary creates a percentage increase in the value of the parent that is Stock Buyback, Stock Dividend, Stock Split, and Ticker Symbol Change. proportional to the fraction of the total value represented by the subsidiary. MULTILATERAL DEBT RELIEF THROUGH THE EYES OF FINANCIAL MARKETS 1273 Figure 3.—Evolution of Cumulative Abnormal Returns for Three Major Debt Relief Initiatives Figures in each panel display the evolution of CARs of South African multinational companies with operations in countries eligible for each stage of the HIPC and MDRI initiatives for each trading trading day within a 10-day event-window around the date of the announcements (solid lines) and their 90% confidence intervals (broken lines). Each CAR is based on abnormal returns estimated from a two-factor model using an estimation window of 112 trading days. A: the CAR obtained by pooling all major initiatives together. B–D: evolution of the CAR estimated for the HIPC initiative, the Enhanced HIPC initiative, and the MDRI. Event dates considered for each of these three major initiatives correspond to the dates of the G8 summit held on June 27, 1996, June 18, 1999, and July 8, 2005. the event date. Moreover, the positive CAR at the begin- points three days before the event to 140 basis points the day ning of the event window cannot be attributed to the meeting after it, and climbed to 200 basis points ten days after the of the finance minsters of G7 countries because on this announcement of the initiative, also remaining statistically occasion, the meeting took place simultaneously with the significant throughout the window. These results support the summit. Apparently financial markets did not perceive this hypothesis that markets considered the latest MDRI, and to announcement as affecting the valuation of parent companies a lesser extent the enhanced HIPC initiative, as significantly with ongoing operations in HIPC-eligible countries. How- positive news for parent companies with operations in eli- ever, this evidence has to be taken with caution because there gible countries. The conclusions about the impact of major were ten South African parent companies with activities in debt relief initiatives on the pattern of cumulative abnormal HIPC countries in 1996 in the data. The announcement of returns are robust to changes in the length of the estimation the enhanced HIPC initiative and the MDRI resulted in an and event windows, as well as changes in the sample of parent increase in cumulative abnormal returns of about two per- companies (see Appendix A). centage points (figures 3C and 3D), although the increase In addition to being economically meaningful in terms of was consistently significant at conventional levels only for abnormal returns, the announcements resulted in increases in the MDRI. The CAR increased from −30 basis points three firm value that were nontrivial but commensurate with the days before the announcement of the enhanced HIPC to 150 amount of debt relief provided. For instance, the average par- basis points the day of the announcement, although it started ent firm in the sample during the MDRI announcement had declining consistently and losing statistical significance two a market capitalization of about $3 billion and a cumulative days later. Thus, the impact on stock prices, while signifi- abnormal return of about 2%, which translates to a $2 billion cant and economically meaningful, seems to have been short value increase when considering the thirty-five firms in the lived. In the case of the MDRI, the CAR rose from −20 basis sample. This figure is economically meaningful and, being 1274 THE REVIEW OF ECONOMICS AND STATISTICS Figure 4.—Evolution of Cumulative Abnormal Returns for Three Major Debt Relief Initiatives: Portfolio Approach Figures in each panel display the evolution of CARs of South African multinational companies with operations in countries eligible for each stage of the HIPC and MDRI initiatives for each trading trading day within a 10-day event window around the date of the announcements (solid lines) and their 90% confidence intervals (broken lines). Each CAR is based on abnormal returns estimated from a two-factor model using an estimation window of 112 trading days. A: the CAR obtained by pooling all major initiatives together. B–D: the evolution of the CAR estimated for the HIPC initiative, the Enhanced HIPC initiative, and the MDRI. Event dates considered for each of these three major initiatives correspond to the dates of the G8 summit held on June 27, 1996, June 18, 1999, and July 8, 2005. one order of magnitude smaller than the $50 billion esti- CAR after the enhanced HIPC, which is notoriously reduced mated amount of debt relief provided, is consistent with these and is not significant at conventional levels.19 parent companies’ capturing a fraction of the benefits of debt The baseline results were derived using an augmented relief. two-factor return model that controls for market and industry The inference in the baseline results was conducted using risk factors. This choice of return model is not crucial for a regression approach to cluster the standard errors of the the results, and very similar but statistically stronger conclu- average abnormal returns at the calendar level. This is an sions are reached when using the three-factor return model appropriate procedure when the number of clusters is large, introduced by Fama and French (1993) or when computing as is the case in this paper (about 130 calendar days), but in the abnormal returns of a parent company as the differ- the event study literature, it is also common to deal with the ence between its returns and those of the rest of the South clustering of events by using only the time series variation African firms in the same size and book-to-market groups of abnormal returns to compute their standard errors (Brown (not reported). and Warner, 1985). This is also akin to the estimator proposed Although the JSE is now a liquid emerging stock mar- by Donald and Lang (2007) for inference with clustered data. ket, it was considerably smaller and less liquid in 1996, just The results reported in figure 4 show that the main findings 19 The point estimates for the CAR may slightly differ from those in the are robust to using this approach for computing the standard baseline estimation because the estimation windows of some firms with errors of the CAR. While the confidence bands are wider, few observations were extended for the estimation of the firm-level return especially when considering individual events, figures 3A model. The standard errors reported were obtained using only the time and 3D show that the CAR are still significant at the 10% series variation of the abnormal returns, as indicated in Brown & Warner (1985), but including only days in the estimation window where the number level for the pooling of all major initiatives and for MDRI, of traded firms was equal to the median number of firms present during the respectively. The most affected result is the significance of the event window, as a crude adjustment for the illiquidity of some stocks. MULTILATERAL DEBT RELIEF THROUGH THE EYES OF FINANCIAL MARKETS 1275 Table 5.—Nonparametric Rank and Sign Tests: Three Major Initiative Events A. Rank Test B. Generalized Sign Test Enhanced Enhanced All Major Events HIPC Initiative HIPC Initiative MDRI All Major Events HIPC Initiative HIPC Initiative MDRI Event Day (1) (2) (3) (4) (5) (6) (7) (8) −10 2.9 38.8∗∗∗ −1.6 −3.4 3.8 3.9∗∗∗ 0.3 −0.3 −9 4.9 11.8 0.6 5.3 4.8 1.9 0.3 2.7 −8 −1.1 5.8 −5.2 0.5 −1.2 0.9 0.3 −2.3 −7 −4.1 −0.4 −16.6∗∗ 2.6 −4.2 −1.1 −4.7∗∗ 1.7 −6 1.9 6.4 1.2 1.1 4.8 0.9 2.3 1.7 −5 −1.1 −9.2 9.7 −5.5 −5.2 −3.1∗∗ 1.3 −3.3 −4 −1.1 −8.1 −4.6 4.1 3.8 −1.1 1.3 3.7 −3 1.9 7.5 2.8 −0.2 0.8 1.9 3.3 −4.3 −2 1.9 −6.2 −1.2 6.0 −0.2 −1.1 −0.7 1.7 −1 6.9 −6.1 12.0 6.9 8.8∗∗ −0.1 4.3∗ 4.7 0 1.9 −17.9 4.5 5.1 −1.2 −3.1∗∗ −0.7 2.7 1 10.9∗∗ −3.5 5.6 18.4∗∗∗ 9.8∗∗ −0.1 2.3 7.7∗∗∗ 2 6.9 6.7 8.6 5.2 7.8∗ 0.9 4.3∗ 2.7 3 −5.1 3.3 −7.4 −6.8 −3.2 0.9 −1.7 −2.3 4 4.9 −15.1 5.7 10.3 4.8 −1.1 2.3 3.7 5 3.9 15.7 0.8 2.5 2.8 1.9 1.3 −0.3 6 −1.1 2.9 0.4 −4.0 0.8 −0.1 −0.7 1.7 7 3.9 −1.5 5.0 5.0 6.8∗ −1.1 2.3 5.7∗∗ 8 −2.1 7.7 −14.7∗ 3.5 2.8 0.9 −2.7 4.7 9 2.9 10.5 −5.4 5.7 −0.2 0.9 −1.7 0.7 10 −1.1 −18.7 2.2 1.9 −5.2 −3.1∗∗ 0.3 −2.3 Standard deviation 4.9 13.5 7.8 7.3 4.1 1.6 2.4 3.0 Number of firms 68 10 23 35 68 10 23 35 Multiday tests Average days 1–2 8.9∗∗∗ 1.6 7.1 11.8∗∗ 8.8∗∗∗ 0.4 3.3∗∗ 5.2∗∗∗ Average days 1–5 4.3∗∗ 1.4 2.7 5.9∗ 4.4∗∗ 0.5 1.7 2.3∗ Panel A: each column reports the mean rank deviations of abnormal returns during a 10-trading day window around the announcement of major debt relief initiatives and indicators of the significance of the test that each of those differences is equal to 0. Panel B: reports the mean sign deviations for the generalized sign test. Mean rank and sign deviations were computed based on abnormal returns estimated over an estimation window of 112 trading days before the beginning of the 10-day event window. Event dates considered for the HIPC initiative, the enhanced HIPC initiative and the MDRI correspond to the G8 summits held on June 27, 1996, June 18, 1999, and July 8, 2005, respectively. ∗∗∗ , ∗∗ , and ∗ denote 1%, 5%, and 10% significance levels, respectively. a few years after the end of the apartheid and international multiday tests reported at the bottom of the table that test the and domestic restrictions to capital flows. This means that hypothesis that the average mean rank deviation at different thin-trading and the resulting volatility and nonnormality of horizons is different from 0 when considering the first two returns may be a problem in earlier years, although the list- and five days after the event. The magnitude of the rank devi- ing of smaller companies in later years may also result in ations, between 11 and 18, is also economically significant. It the presence of illiquid stocks in the final part of the sam- indicates that the ranks of the abnormal returns in those dates ple. Nevertheless, the positive impact of major debt relief are at least ten places higher than what would be expected announcements indicated by the evolution of cumulative by chance, which corresponds to 20% of the expected rank abnormal returns is supported by the results of nonparametric value.21 These changes in the rank of abnormal returns are rank and sign tests with better power under nonnormal returns clearly displayed in figure 5, which exhibits the cumulative than standard parametric tests.20 The evolution of the mean mean rank deviations during the event window and their 90% rank deviations K ¯ t during each day of the event window is confidence bands. In all cases but the original HIPC initiative, reported in columns 1 to 4 of table 5 for all major initiatives, there is a positive break in the cumulative values around the HIPC, enhanced HIPC, and MDRI, respectively. Similar to event day, although the cumulative statistics are statistically the parametric tests based on the CAR, the mean rank devia- significant only for the pool of major initiatives (figure 5A) tions take a positive and statistically significant value the day and for the MDRI (figure 5D). after the event when looking at the pooling of all major initia- Stronger results are obtained for the median sign deviations tives (column 1) and at the MDRI (column 4). Furthermore, G¯ τ , whose evolution within the event window is reported in in these two cases, there is no other positive significant rank columns 5 to 8 of table 5 and depicted in figure 6 in cumu- deviation within the event window. This time coincidence fur- lative form, and which is more robust to thin trading than ther supports the association of these abnormally high returns the rank test (Cowan, 1992). Except for the original HIPC and the event under consideration. This is corroborated by initiative, the results exhibit positive and significant statis- tics in a window of two days around the event date. After 20 The results are also supported by evidence from CAR obtained from the announcement, abnormal returns are much more likely to estimating equation (1) using trade-to-trade returns (returns computed over be positive than expected by chance. For instance, there are two consecutive days of trading, not over two consecutive calendar days), an alternative and popular manner of dealing with thin trading first suggested by Maynes and Rumsey (1993) (not reported). 21 Based on the length of the windows, the expected rank is 66. 1276 THE REVIEW OF ECONOMICS AND STATISTICS Figure 5.—Evolution of Cumulative Median Rank Deviations of Cumulative Abnormal Returns for Three Major Debt Relief Initiatives Figures in each panel display the evolution of the cumulative mean rank deviations of abnormal returns (CAR) of South African multinational companies with operations in countries eligible for each stage of the HIPC and MDRI initiatives for each trading trading day within a 10-day event window around the date of the announcements (solid lines) and their 90% confidence intervals (broken lines). Each CAR is based on abnormal returns estimated from a two-factor model using an estimation window of 112 trading days, and the same windows are used to determine the ranks. A: the median rank deviations of the CAR obtained by pooling all major initiatives together. B–D: the evolution of the median rank deviations separately estimated for the HIPC initiative, the Enhanced HIPC initiative, and the MDRI. Event dates considered for each of these three major initiatives correspond to the dates of the G8 summit held on June 27, 1996, June 18, 1999, and July 8, 2005. on average 10 more firms with positive abnormal returns the parent companies indirectly affected. One could conjecture day after the announcement of a major initiative (out of 68) that this lack of impact could arise from the requirements and 8 firms more after the announcement of MDRI (out of that eligible countries had to meet under the original HIPC 35). This is confirmed by the multiday tests reported at the initiative to begin receiving some form of relief, which the bottom of the table that show that the average median sign markets could have anticipated were not going to be met by deviations remain significant several days after the event date a broad set of countries, although, of course, this remains to and confirmed as well in figure 6, which shows that cumu- be proven. lative mean sign deviations are statistically significant after The baseline results reported above consider as event dates the event day for the pooling of all events, for the MDRI, the announcements of debt relief initiative taking place in the and marginally so for the enhanced HIPC. As in the case of summit meetings of heads of state and government of the G8. parametric tests, these findings are robust to variations in the Nevertheless, as discussed in section II, the details of the ini- estimation windows (not reported). tiatives, including the eligibility criteria, were typically sorted The lack of significant results for the original HIPC out during the annual meetings of the board of governors of initiative using nonparametric tests with better power than the World Bank and IMF taking place in September of the parametric tests under nonnormality, as shown by Corrado same year. It is therefore possible that some valuable infor- (1989) and Cowan (1992), makes unlikely that the lack of mation affecting the returns of companies with interests in impact of the original HIPC on parent companies’ returns HIPC countries could be released around these dates instead. stems from the illiquidity of the JSE during this period or the To check for this possibility, the evolution of the CAR was small number of firms, and supports instead the view that the also estimated around the dates of the World Bank and IMF original initiative did not have a noticeable impact on these annual meetings reported in table 1. The results, presented in MULTILATERAL DEBT RELIEF THROUGH THE EYES OF FINANCIAL MARKETS 1277 Figure 6.—Evolution of Cumulative Median Sign Deviations of Cumulative Abnormal Returns for Three Major Debt Relief Initiatives Figures in each panel display the evolution of the cumulative mean sign deviations of abnormal returns (CAR) of South African multinational companies with operations in countries eligible for each stage of the HIPC and MDRI initiatives for each trading trading day within a 10-day event window around the date of the announcements (solid lines) and their 90% confidence intervals (broken lines). Each CAR is based on abnormal returns estimated from a two-factor model using an estimation window of 112 trading days. A: the median sign deviations of the CAR obtained by pooling all major initiatives together. B–D: the evolution of the median sign deviations separately estimated for the HIPC initiative, the Enhanced HIPC initiative, and the MDRI. Event dates considered for each of these three major initiatives correspond to the dates of the G8 summit held on June 27, 1996, June 18, 1999, and July 8, 2005. figure 7, do not show clear evidence of a break in the pattern of impact after the launching of the initial HIPC in 1996. The abnormal returns around these alternative event dates, except evidence on the consequences of the announcement of the in the case of the enhanced HIPC initiative. When pooling enhanced version of the HIPC in 1999 is less clear. The point all major initiatives together, as well as when looking at the estimates for CAR and nonparametric statistics show an ini- MDRI, there is no significant impact on the sequence of CAR tial stock price response that is similar to that following the around the dates of the annual meetings. The results, there- MDRI but much shorter lived, and the statistical significance fore, tend to indicate that the announcements of the launching of this initial response varies across tests from marginally sig- of the various initiatives by the heads of state of industrial nificant to not significant at all. This lack of robust inference countries are the ones that are considered good news for may be due to the relatively small number of firms affected firms with operations in HIPC countries by financial markets, by this event (23) and the high volatility of the JSE during suggesting that for the most part, financial markets assume this year (the JSE all share index was much more volatile in that once those announcements are made, the details of the 1999 than in 1996 despite the larger number of listed firms). implementation are of second-order importance. According to this interpretation, from a statistical perspec- Overall the results consistently show that the announce- tive, this episode would be better considered as part of the ment of major debt relief initiatives had on average a positive pool of major debt relief initiatives than in isolation. How- impact on the stock market returns of parent companies with ever, the comparison of the stark results for the MDRI and the ongoing operations in benefited countries. When looking at weaker enhanced HIPC results also suggests the possibility individual announcements, there is also strong evidence of that there was something different about the episodes. It could an abnormally positive stock price response following the be that the announcement of the MDRI was a sharper event, announcement of the MDRI and no evidence of a positive in that it was more immediately clear what the implications 1278 THE REVIEW OF ECONOMICS AND STATISTICS Figure 7.—Evolution of Cumulative Abnormal Returns around Final Implementation Dates of Three Major Debt Relief Initiatives Figures in each panel display the evolution of CARs of South African multinational companies with operations in countries eligible for each stage of the HIPC and MDRI initiatives for each trading trading day within a 10-day event window around the date of the announcements (solid lines) and their 90% confidence intervals (broken lines). Each CAR is based on abnormal returns estimated from a two-factor model using an estimation window of 112 trading days. A: the CAR obtained by pooling all major initiatives together. B–D: the evolution of the CAR estimated for the HIPC initiative, the Enhanced HIPC initiative, and the MDRI. Event dates considered for each of these three major initiatives correspond to the dates of the final agreements on implementation of the initatives during the annual meetings of the World Bank and International Monetary Fund on September 29, 1996, September 26, 1999, and September 26, 2005, respectively. were for individual countries, while it may have taken more ten-trading-day event window, along with their 90% confi- time for these implications to be clear in HIPC and enhanced dence bands, are depicted in figure 8. Figures 8A and 8B show HIPC cases, blurring the results for these events. no significantly positive CAR around the event dates for deci- sion or completion points. However, in both cases, the CAR B. Country-Level Announcements exhibits a preevent decreasing trend that stops and slightly reverses around the announcement dates and could indicate The impact of the announcements of major debt relief ini- some market response to the announcements. Looking sep- tiatives on parent companies with operations in HIPC-eligible arately at decision and completion points reached under the countries indicates that markets believe these countries will HIPC and enhanced HIPC does not significantly affect the eventually get debt relief. However, at the time of these results (not reported). In sum, parametric tests offer little sup- announcements, there was still uncertainty about whether and port to the hypothesis that the resolution of the uncertainty when a specific country will meet the conditions for HIPC and about a country getting debt relief has an impact on the stock MDRI relief. This uncertainty resolves as countries advance returns of parent companies with affiliates in that country. through the HIPC process and reach decision and comple- Nonparametric tests, however, are more supportive of tion points. So the formal announcement that a country has the hypothesis that these country-level announcements have reached one of these milestones could in principle affect the some effect on stock returns. The estimated mean rank devi- returns of a related parent company. The following results ations and mean sign deviations of the rank and sign tests test for this possibility. statistics for decision and completion points are reported The estimated sequences of cumulative abnormal returns in table 6. Both tests produce positive and statistically sig- for decision and completion point events during a nificant statistics immediately after the announcement of a MULTILATERAL DEBT RELIEF THROUGH THE EYES OF FINANCIAL MARKETS 1279 Figure 8.—Evolution of Cumulative Abnormal Returns for Decision and Completion Point Events Figures in each panel display the evolution of CARs of South African multinational companies with affiliates in African HIPC countries benefited by decision and completion points under the HIPC initiative during a 10-trading-day window around the announcement of each decision and completion point (solid lines) and their 90% confidence intervals (broken lines). Each CAR is based on abnormal returns estimated from a two-factor model using an estimation window of 112 trading days. A: CAR obtained around decision point announcements. B: CAR obtained around completion point announcements. completion point (between dates 0 and 2; see columns 2 and Table 6.—Rank and Generalized Sign Tests: Decision and Completion Point Events 4), with average statistics that are still significant at con- ventional levels five days after the event. Since these tests Mean Rank Deviation Mean Sign Deviation are better able to deal with nonnormality, they provide more Decision Completion Decision Completion robust evidence than parametric tests based on the estimated Points Points Points Points Event Day (1) (2) (3) (4) CAR. The magnitude of the increase in abnormal returns is small, however, reaching a maximum cumulative value of 40 −10 0.25 0.9 1.4 3.6 −9 −1.75 −3.1 −3.6 −8.4∗ basis points five days after the event (relative to its preevent −8 1.25 −1.1 2.4 −2.4 level). −7 3.25 −5.1 1.4 −5.4 In summary, the evidence cannot fully reject the hypoth- −6 −1.75 −4.1 0.4 −5.4 −5 −1.75 −1.1 −3.6 0.6 esis that the announcement of a completion point has some −4 −0.75 −3.1 2.4 −4.4 positive impact on the returns of parent companies with affil- −3 −4.75 2.9 −3.6 2.6 iates in the benefited country, but the size of the impact is −2 −5.75 2.9 −5.6 3.6 −1 −5.75 −2.1 −3.6 −3.4 small compared with that of major initiatives. This is not sur- 0 0.25 8.9∗∗ 2.4 10.6∗∗ prising considering that country-level announcements affect 1 −5.75 −1.1 −0.6 −1.4 parent companies only through their operations in that indi- 2 5.25 6.9∗ 5.4 7.6 3 1.25 −1.1 −1.6 1.6 vidual country, while major initiatives affect them through 4 1.25 −0.1 5.4 2.6 their operations in all eligible countries. It seems that the 5 −0.75 4.9 −2.6 3.6 breadth of the impact of major initiatives compensates for the 6 4.25 −0.1 5.4 0.6 7 −2.75 −5.1 −2.6 −7.4 remaining uncertainty about the timing of debt relief to spe- 8 3.25 1.9 3.4 1.6 cific countries to induce movements in the returns of parent 9 2.25 −2.1 3.4 −2.4 companies. 10 0.25 −0.1 −2.6 1.6 Standard deviation 3.6 3.8 4.0 4.8 Number of firms 65 92 65 92 Multiday tests C. Does the Intensity of Exposure Matter? Average 0–2 −0.1 4.9∗∗ 2.4 5.6∗∗ Average 0–5 0.8 2.6∗ 2.0 3.6∗ The baseline results are based on identifying abnormal Each column reports the sequence of mean rank deviations of abnormal returns of South African multi- returns of parent companies with operations in HIPC coun- national companies with affiliates in benefited HIPC countries during a 10-trading-day window around each decision and completion point events, as well as indicators of the significance of the test that each of tries controlling for market and industry movements but do those differences is equal to 0. Mean rank deviations were computed based on abnormal returns estimated over a window of 112 trading days before the beginning of the 10-day event window. ∗∗∗ , ∗∗ , and ∗ denote not exploit variation in their exposure to the events. Any two 1%, 5%, and 10% significance levels, respectively. South African multinationals with operations in an HIPC country at the time of an announcement are considered exposed to the event regardless the size of their operations. with larger interests in HIPC-eligible countries to benefit rel- However, to the extent that the abnormal returns already atively more from the events. Therefore, differences in the documented result from an improvement in the prospects intensity of exposure to the event can be used to build an addi- of exposed companies, one would expect parent companies tional test of the hypothesis that the stock price response to 1280 THE REVIEW OF ECONOMICS AND STATISTICS the announcements comes from their significance for firms’ D. Aggregate Demand versus Future Taxation prospects and to provide further support to the mechanism behind the results. Debt relief can increase the value of firms with operations To build this test, I construct a measure of the expo- in benefited countries by reducing their expected future tax- sure of a parent company to an HIPC event by computing ation or raising their gross profits as a consequence of the the ratio of the total employment of its affected affiliates increase in demand resulting from an increase in its own or (adjusted for parent company holdings) to the parent’s total in aggregate investment. This section presents results aimed number of employees. The focus on employment as a size to disentangle which of these channels is most likely to be measure instead of assets is for data availability reasons; responsible for the stock return response of parent compa- while data on a parent company’s assets are easy to obtain, nies. To this end, it separately looks at the impact of the most subsidiaries report employment data only to McGre- announcements of major initiatives on the returns of par- gor (2006). Employment data for those subsidiaries with ent companies operating in industries that are likely to have some available information are typically obtainable only for different sensitivities to local taxation and aggregate demand. 2005–2006, so subsidiary employment is considered a fixed Resource extraction is typically considered an enclave subsidiary characteristic and is normalized by parent total industry, whose rents are readily taxed or even expropriated, employment data for the year of each event. Based on these especially if they are foreign owned (Engel & Fischer, 2010) data, a parent’s company exposure to an announcement at and whose investment and production is sensitive to owner- time t is computed as the ratio of the total employment of ship risk (Bohn & Deacon, 2000). Also, resource extraction all its subsidiaries operating in benefited countries at that produces commodities sold in global markets, so gross prof- moment, to the parent’s total employment.22 Recognizing that its in this industry depend little on the growth of aggregate this measure is likely to contain an important amount of noise demand in the host country. Thus, it can be expected that because of the assumptions required to extend the sample this industry would benefit relatively more by a reduction coverage, the final measure does not exploit the continuous in expected future taxes than by an improvement in the nature of this exposure indicator but instead classifies parent host country’s future economic conditions. On the contrary, companies in two groups depending on whether its exposure the performance of firms in nontradable service industries indicator is above or below the median level, and compare the depends mainly on local economic conditions, and at the behavior of the abnormal returns across these two groups.23 same time, for political economy reasons, these industries The evolution of the CAR for parent companies with rela- may be a less tempting target for taxes because they tend to tively high and low exposure is reported in figure 9. Figure 9C be more labor intensive. Therefore, the differential response reports the evolution of the differential effect of the exposure of stock returns of parent companies operating in resource on these CARs and their 90% confidence bands. The response extraction versus companies operating in services can be of CAR in parent companies with relatively higher exposure used to test whether the positive stock market return of major to the events is clearly larger than that of less exposed com- announcements comes mainly from the market’s assessment panies (see figures 9A and 9B). Moreover, this difference is of an expansion of aggregate demand on the benefited coun- significantly different from 0 at conventional levels immedi- tries or from a reduction in expected taxes paid by parent ately after the event. As shown in figure 9C, which exhibits companies. the evolution of the difference in CAR for firms with rela- The evolution of cumulative abnormal returns and their tively high exposure, the CAR of these firms increases clearly corresponding 90% confidence bands for companies in more than that of firms with low exposure. resource extraction and service industries, as well as the The results reported in this section show that returns of difference in CAR between these two industries during a multinational firms with relatively larger interests in countries ten-day window around the announcement of major initia- benefited by multilateral debt relief increase relatively more tives is reported in figure 10. The figure shows that the CAR after the announcements of these initiatives. This finding experiences a larger and more significant increase around the provides strong support to the hypothesis that the abnormal event day for parent companies in resource extraction than in returns for parent companies exposed to debt relief events services. Although the CAR of resource companies reaches in the baseline estimates are a consequence of the market’s 5 percentage points and is statistically significant at the 10% assessment that the events convey positive news for the value level since day 0, those of service companies climb up to of these companies. Results in the next section explore the only 1 percentage point and are statistically significant at reasons behind these assessments. 12%. This differential response can be clearly seen in Figure 10C, which shows that the CAR of companies in resource 22 Assuming that the capital labor ratio of a firm is constant across sub- industries is higher than those of companies in service indus- sidiaries and headquarters, the ratio of employment in subsidiaries to the parent’s employment is a correct measure of the fraction of the value of the tries, and that this difference is significantly different from 0 parent represented by these subsidiaries. in various postevent dates (days 3, 7, and 8). 23 The difference between the abnormal returns of firms with high and low Thus, while the CAR of firms in nontradable industries exposure was obtained by adding an interaction term between the event time dummies and a dummy that takes the value of 1 for parent companies with increases after a major announcement, that of resource indus- exposure above the median. tries increases significantly more. Under the assumption that MULTILATERAL DEBT RELIEF THROUGH THE EYES OF FINANCIAL MARKETS 1281 Figure 9.—Evolution of Cumulative Abnormal Returns around Three Major Debt Relief Initiatives: Firms with Relatively Small and Large Investments in HIPC Countries Figures in each panel display the evolution of CARs of South African multinational companies with operations in countries eligible for each stage of the HIPC and MDRI initiatives for each trading day within a 10-day event window around the date of the announcements (solid lines) and their 90% confidence intervals (broken lines). Each CAR is based on abnormal returns estimated from a two-factor model using an estimation window of 112 trading days. Event dates considered for each of these three major initiatives correspond to the dates of the G8 summit held on June 27, 1996; June 18, 1999; and July 8, 2005. stock prices of companies in resource extraction industries are that this is not the case. First, subsidiaries in sectors that are relatively more responsive to expected taxation than those of more sensitive to local conditions also experience an increase companies in the service sector and that the opposite is true for in value. Service firms are also subject to taxation, so part of the responsiveness to local economic prospects, the stronger the increase must also be driven by lower expected taxes, positive stock returns’ response of the former relative to the but among service industries, firms in noncyclical services latter suggests that the expected reduction in future taxation (food retailers and telecommunications) that are likely more after a major debt relief announcement has a larger impact closely related to local economic conditions than other ser- on stock market prices than expected increases in aggregate vice firms experienced the largest increase (about 3% CAR, demand. significant at the 5% level). This seems to indicate some Although an expected tax reduction could be a pure transfer expected increase in aggregate demand but has to be taken to shareholders without any benefit for the countries receiving with caution because there are only five event firms in this sec- debt relief, as would be the case if public and private invest- tor. Second, private FDI inflows to African HIPC countries ment was not immediately constrained by high expected increased on average from 3.9% to 5.8% of GDP between future taxes but for other factors (such as institutions, see 2004 and 2006 (the year before and after the MDRI), while Arslanalp & Henry, 2006), there is some anecdotal evidence the increase for non-HIPC African countries was from 3.5 to 1282 THE REVIEW OF ECONOMICS AND STATISTICS Figure 10.—Evolution of Cumulative Abnormal Returns around Three Major Debt Relief Initiatives: Firms in Resource Extraction versus Firms in Services Figures in each panel display the evolution of CARs of South African multinational companies with operations in countries eligible for each stage of the HIPC and MDRI initiatives for each trading day within a 10-day event window around the date of the announcements (solid lines) and their 90% confidence intervals (broken lines). Each CAR is based on abnormal returns estimated from a two-factor model using an estimation window of 112 trading days. A: the CAR of parent companies in resource extraction industries; B: the CAR of parent companies in the service sector. Event dates considered for each of these three major initiatives correspond to the dates of the G8 summit held on June 27, 1996, June 18, 1999, and July 8, 2005. 4.1 of GDP during the same period, suggesting that more While not conclusive, this anecdotal evidence suggests that resources have been devoted to these countries by inter- investment and economic activity may have been constrained national investors.24 Furthermore, as shown in table 3, the by debt overhang and that the decline in expected taxes number of subsidiaries of South African multinationals in is unlikely to be a pure transfer to the shareholders of African HIPC countries also increased during this period, multinationals. and a more detailed comparison of the dates of new invest- Finally, the evidence in figure 10 also helps to dispel the ments reported in McGregor (2006) shows that the average possibility that the abnormal returns could come from firms annual number of new South African subsidiaries per country in nontradable industries benefiting from a real appreciation in African HIPC countries experienced a statistically signif- resulting from the increase in net aid inflows associated with icant increase of 0.24 between 1990–1995 and 1996–2006 debt relief (Rajan & Subramanian, 2005a, 2005b). If this (from 0.31 to 0.55). Among non-HIPC African countries, mechanism were driving the results, the resource-producing the increase was a nonsignificant 0.15 (from 0.89 to 1.04).25 industries, which are tradable, should experience a smaller abnormal stock price response than nontradable industries, 24 This is not only due to a decline in GDP since the inflows in levels also exactly the opposite of what is found in the data. duplicated from an average of $172 to $366 million during the same period (World Bank, 2009). 25 The average number of new subsidiaries per country is the number of new subsidiaries created or acquired in a given year divided by the number eighteen sub-Saharan African HIPC countries, and the eleven non-HIPC of countries. The number reported is the average of these ratios across two countries where South African companies have any investment according time periods: pre-HIPC, 1990–1995 and post-HIPC, 1996–2006, among the to McGregor (2006). MULTILATERAL DEBT RELIEF THROUGH THE EYES OF FINANCIAL MARKETS 1283 V. Conclusion relief. Absent direct transfers, this would most likely occur through lower taxation, as shown in the previous section, There has been considerable attention in recent years to which is the type of improvement in the business environ- the debt problems of poor countries and numerous calls for ment behind the debt overhang argument. It still may be the debt relief coming from all sectors of society and from actors case that investment may not be constrained by debt over- ranging from the religious leaders to rock stars. These calls hang but by poor institutions, as discussed in Arslanalp & have been based on arguments ranging from the moral wrong Henry (2006), and the decline in taxation may lead not to of rich countries’ collecting debt from people at the brink of further investment but just higher dividends for sharehold- starvation to the economic costs of debt overhang, and have ers. However, the scarce available anecdotal evidence points encompassed issues such as the fight against poverty and the toward expansions in foreign direct investment inflows in irrationality of permanently rolling over multilateral loans. African HIPC countries relative to other African countries This paper has focused on the economic rationale for debt around the launching of the MDRI and the broader HIPC relief and looked for evidence that multilateral debt relief initiative, consistent with the debt overhang argument. With- initiatives improve the economic prospects of benefited coun- out constituting a formal proof, this suggests that incentives tries. In contrast to the existing literature, this is done by to invest in benefited countries improved, which is probably conducting an event study to overcome the standard econo- partly responsible for the increase in stock prices following metric problems of estimating aggregate relations between the announcements, as suggested from limited evidence from debt relief and macroeconomic performance. noncyclical service firms. The evidence presented in the paper overwhelmingly Second, the results in this paper indicate that the benefit indicates that the announcement of multilateral debt relief is quantitatively meaningful from the point of view of par- initiatives conveys good news for multinational firms operat- ents and subsidiaries. However, the size of the effect for the ing in receiving countries according to the views of financial receiving country will surely differ from my estimation. The market. The stock return of firms with affiliates in these coun- evidence presented here has to be considered highly indica- tries experiences a statistically and economically significant tive of the presence of the positive impact of debt relief rather abnormal increase around the dates of the formal announce- than a quantification of that impact. As is almost always the ments of the initiatives, especially after the launching of the case, the benefits of the methodological approach in terms MDRI. This increase is larger for firms with relatively larger of cleaner identification are weighed by the cost of deviating interests in receiving countries and seems to come mainly from the immediate object of interest: the aggregate growth from the perception that future taxes will be lower as a result effect of these initiatives. of debt relief. This indirect evidence is consistent with the Finally, an event study methodology exploits, by construc- debt overhang argument for debt relief, where the expectation tion, the local variation of the data to identify the conse- of future taxation depresses the value of future investment quences of a specific event; everything occurring outside the projects, and it suggests that the announcements of the initia- event windows is considered as potentially contaminated by tives were perceived as potentially beneficial for the receiving other developments. This is, of course, an untested identifi- countries. cation assumption. It is therefore impossible to know within As in any other empirical exercise, there are caveats to this framework whether the financial market’s view of the keep in mind despite the strength of the evidence. First, the consequences of these initiatives persisted in time beyond results directly show that multinationals operating in bene- the windows studied in this paper or reversed as a result of fited countries benefit most likely because their subsidiaries further news. benefit. The interpretation that the countries receiving debt REFERENCES relief also benefit relies on the increases in parent companies’ values resulting from an improvement in the host coun- Arslanalp, Serkan, and Peter Blair Henry, “Is Debt Relief Efficient?” Journal of Finance 60 (2005), 1017–1051. tries’ economic prospects. This can plausibly result from an ——— “Policy Watch: Debt Relief,” Journal of Economic Perspectives 20 expected increase in public and private investments resulting (2006), 207–220. from lower expected taxes that rises with aggregate demand Birdsall, Nancy, and John Williamson, Delivering on Debt Relief: From IMF Gold to a New Aid Architecture (Washington, DC: Peterson and productivity or from an expansion in net resource trans- Institute for International Economics, 2002). fers. There are other channels, however, for the value of Bohn, Henning, and Robert T. Deacon, “Ownership Risk, Investment, multinational to increase without an improvement in host and the Use of Natural Resources,” American Economic Review 90 (2000), 526–549. countries’ economic prospects. One is that the value of the Brown, Stephen J., and Jerold B. Warner, “Using Daily Stock Returns: The firms under study behaved countercyclically. If that were the Case of Event Studies,” Journal of Financial Economics 14 (1985), case, an improvement in the value of multinationals may sig- 3–31. Bulow, Jeremy, and Kenneth Rogoff, “Grants versus Loans for Development nal a decline in expected local conditions, but this is not the Banks,” American Economic Review 95 (2005), 393–397. case for the firms in this sample. A second possibility, dis- Calvo, Guillermo, and Jeffrey Sachs, “The Debt Overhang of Developing cussed in section IVD, is that debt relief may simply be a Countries” (pp. 80–102), in Guillermo A. Calvo, Ronald Findlay, Pentti Kouri, and Jorge Braga De Macedo (Eds.), Debt, Stabiliza- transfer from creditors to the shareholders of these multi- tion and Development: Essays in Memory of Carlos Diaz-Alejandro national firms without any benefit for the country receiving (Cambridge, MA: Blackwell, 1989). 1284 THE REVIEW OF ECONOMICS AND STATISTICS Campbell, Cynthia J., Arnold R. Cowan, and Valentina Salotti, “Multi- Kovrijnykh, Natalia, and Balázs Szentes, “Equilibrium Default Cycles,” Country Event Study Methods,” unpublished manuscript (2009). Journal of Political Economy 115 (2007), 403–446. Campbell, Cynthia J. and Charles E. Wasley, “Measuring Abnormal Daily Krugman, Paul, “Financing vs. Forgiving a Debt Overhang,” Journal of Trading Volume for Samples of NYSE/ASE and NASDAQ Securi- Development Economics 29 (1988), 253–268. ties Using Parametric and Nonparametric Test Statistics,” Review of Lexis-Nexis, Corporate Affiliations (Lexis-Nexis, 2007). Quantitative Finance and Accounting 6 (1996), 309–326. Maynes, Elizabeth, and Rumsey, John, “Conducting Event Studies with Claessens, Stijn, “The Debt Laffer Curve: Some Estimates,” World Devel- Thinly Traded Stocks,” Journal of Banking and Finance 17 (1993), opment 18 (1990), 1671–1677. 145–157. Cordella, Tito, Luca Antonio Ricci, and Marta Ruiz-Arranz, “Debt Over- McGregor, Robert, McGregor’s Who Owns Whom in South Africa (vol. 26, hang or Debt Irrelevance? Revisiting the Debt Growth Link,” IMF McGregor Publishers, 2006). working papers no. 05/223 (2005). Rajan, Raghuram, and Arvind Subramanian, “Aid and Growth: What Does Corrado, Charles J., “A Nonparametric Test for Abnormal Security-Price the Cross-Country Evidence Really Show?” IMF working papers Performance in Event Studies,” Journal of Financial Economics 2 05/127 (2005a). (1989), 385–395. ——— “What Undermines Aid’s Impact on Growth?” NBER working Cowan, Arnold R., “Nonparametric Event Study Tests,” Review of Quanti- papers no. 11657 (2005b). tative Finance and Accounting 2 (1992), 343–358. United Nations Conference on Trade and Development, World Investment Daseking, Christina, and Robert Powell, “From Toronto Terms to the Directory (New York: United Nations, 1993, 2004). HIPC Initiative—A Brief History of Debt Relief for Low-Income World Bank, “Paying Taxes 2008—The Global Picture,” technical report Countries,” IMF working papers no. 99/142 (1999). (2008). Depetris-Chauvin, Nicolas, and Aart Kraay, “What Has 100 Billion Dollars ——— “World Development Indicators” (Washington, DC: World Bank, Worth of Debt Relief Done for Low-Income Countries?” Eco- 2009). nomic Working Paper Archive, International Finance, no. 0510001 (October 2005). Deshpande, Ashwini, “The Debt Overhang and the Disincentive to Invest,” Journal of Development Economics 52 (1997), 169–187. Djankov, Simeon, Caralee McLiesh, and Andrei Shleifer, “Private Credit APPENDIX in 129 Countries,” Journal of Financial Economics 84 (2007), 299– 329. Robustness to Changes in the Windows and Samples Donald, Stephen G., and Kevin Lang, “Inference with Difference-in- Differences and Other Panel Data,” this review 89 (2007), The conclusions about the impact of major debt relief initiatives on the 221–233. pattern of cumulative abnormal returns are robust to changes in the length of Engel, Eduardo, and Ronald Fischer, “Optimal Resource Extraction Con- the estimation and event windows, as well as changes in the sample of parent tracts under Threat of Expropriation,” in W. Hogan and F. Sturzeneg- companies, as shown in the different panels of figure A1. Figure A1 summa- ger (Eds.), The Natural Resource Trap (Cambridge, MA: MIT Press, rizes the findings only for the pooling of major debt relief announcements, 2010). but similar results are obtained for individual announcements (available Fama, Eugene F., and Kenneth French, “Common Risk Factors in the upon request). Panel A shows results for different estimation windows, Returns on Stocks and Bonds,” Journal of Financial Economics 33 where the bold lines depict the evolution of the baseline estimates of the (1993), 3–56. CAR, obtained with an estimation window of 112 trading days, and the Geginat, Carolin, and Aart Kraay, “Does IDA Engage in Defensive Lend- thin lines display the evolution of the CAR for smaller estimation windows ing?” World Bank Policy Research working paper series 4328 ranging from 52 to 102 trading days, in increments of 10. The shaded area (2007). corresponds to the envelope of the estimated patterns. It is clear in the figure Graham and Whiteside, Major Companies of Africa South of the Sahara that the pattern of the CAR does not vary importantly with the reduction of (London: Graham and Whiteside, 1997–2004). the estimation window and that the range spanned by the different estimates Guidolin, Massimo, and Eliana La Ferrara, “Diamonds Are Forever, Wars does not deviate importantly from the baseline results. Panel B summarizes Are Not: Is Conflict Bad for Private Firms?” American Economic the CAR obtained after changing the length of the event window from the Review 97 (2007), 1978–1993. baseline level of 10 trading days around the event to 5 and 20 trading days. Imbs, Jean, and Romain Ranciere, “The Overhang Hangover,” World Bank As before, the pattern and level of CAR are largely unaffected by these Policy Research working paper series 3673 (2005). changes in length. Panel C shows that the baseline results are also robust International Development Agency and International Monetary Fund, to changes in the sample of parent companies by summarizing the distribu- “Heavily Indebted Poor Countries (HIPC) Initiative and Multi- tion of the sequence of CAR obtained after dropping one parent company lateral Debt Relief Initiative (MDRI)” (Washington, Dc: World at a time. At each event time, the figure shows the mean, 25th percentile, Bank and International Monetary Fund, 2006). http://www.imf 75th percentile, and minimum and maximum CAR estimated for that date, .org/external/np/pp/eng/2006/082106.pdf. and marks in gray the area spanned between the minimum and maximum Kolari, James W., and Seppo Pynnonen, “Generalized Rank Test for Test- estimated CAR. The distribution of the CAR at each event day is tightly con- ing Cumulative Abnormal Returns in Event Studies,” mimeograph centrated around the mean value, which indicates that the pattern depicted (2008). in the baseline results is not driven by any individual parent company. MULTILATERAL DEBT RELIEF THROUGH THE EYES OF FINANCIAL MARKETS 1285 Figure A1.—Evolution of Cumulative Abnormal Returns under Different Estimation Windows Figures in each panel display the evolution CARs of South African multinational companies with operations in countries eligible for each stage of the HIPC and MDRI initiatives for each trading trading day within a ten-day event window around the date of the announcements. All figures correspond to the CAR obtaining by pooling all major initiatives together. A: different lines correspond to CAR estimated using estimation windows of different length; the thick line corresponds to the baseline values of the CAR obtained with an estimation window of 112 trading days, and the other lines depict the CAR obtained with shorter estimation windows ranging from 102 to 52 trading days, in intervals of 10. The gray area is the envelope spanned by the different sequences of CAR reported. B: similar results for various event windows around the date of the announcements. The thick line corresponds to the CAR obtained with an event window of 20 trading days around the announcements, the dashed line those for 10 trading days, and the dotted line those for 5 trading days. In all cases, the estimation window is 112 trading days before the beginning of the 10-day estimation window plus (minus) the difference corresponding to the varying length of the event window. C: the distribution of CAR at each day in the benchmark 10-day event window obtained after sequentially dropping one of the affected firms from the sample. The gray area corresponds to the range in which the different estimated CARs fall, and is enclosed by the minimum and the maximum estimated value. Within the gray area, the solid line is the mean of the estimated CAR, and the broken lines represent the 25th and 75th percentiles values of the CAR. Event dates considered for each of these three major initiatives correspond to the dates of the G8 summit held on June 27, 1996, June 18, 1999, and July 8, 2005. Table A1.—Detailed Chronology of Events Related to the HIPC Initiatives and the MDRI Date Event Source September 26, 1994 Commonwealth finance ministers meeting. U.K. chancellor Financial Times, June 17, 1995 of the exchequer, Kenneth Clarke, proposes IMF’s gold sales to finance debt cancellation for the poorest countries. April 25, 1995 IMF and World Bank spring meetings. Britain, Canada, Gazette, April 26, 1995 and the United States spearhead a proposal to sell as much as $4 billion worth of gold from world gold stockpiles to provide debt relief for the poorest nations. June 16, 1995 G7 meeting in Halifax, Canada. Canadian finance minister Financial Times, June 17, 1995 Paul Martin endorses Clarke’s proposal. October 9, 1995 World Bank’s Policy-Setting Development Committee. The Financial Times, October 10, 1995 committee calls for the bank and the IMF to make proposals, in time for its next meeting in April, for dealing with the debts owed to multilateral institutions by a handful of poor countries. 1286 THE REVIEW OF ECONOMICS AND STATISTICS Table A1.—(Continued) Date Event Source March 14, 1996 World Bank board meeting. A joint proposal with the IMF, New York Times, June 10, 1996; Financial Times, March 14, which aims to provide a comprehensive resolution of 1996 the external debt problems faced by the world’s poorest countries over a six-year span, is approved by the World Bank. This is the first attempt by these multilateral creditors on their debt relief proposal. However, the use of IMF’s reserved gold is not considered. April 21, 1996 IMF and World Bank spring meeting. G7 financial Deutsche Presse-Agentur, April 23, June 10, 1996 ministers reject the joint proposal. June 4, 1996 Revised proposal by the World Bank. The revised plan Guardian (London), June 10, 1996 proposes to sell the IMF’s gold as well as to cut the waiting time for countries qualifying for multilateral debt relief from six to three years. June 27, 1996 G7 finance meeting and G8 summit in Lyon, France. The Deutsche Presse-Agentur, June 28, 1996 proposal is accepted, except the idea of IMF’s gold sales to finance debt relief, which Germany strongly opposes. Options for financing is yet to be finalized. September 29–30, 1996 Annual meetings of the IMF and the World Bank. The Financial Times (London), September 30, 1996 HIPC initiative is approved by the IMF and World Bank. Agreement by the G7 nations to increase debt cancellation from 67% (decided in Naples Terms in 1994) to a maximum of 80% is followed by the approval of the initiative. April 1, 1997 Meeting at the United Nations Economic Commission for Africa News, April 1, 1997 Africa. Leading African finance policymakers with top policymakers from the World Bank, the IMF and the African Development Bank discuss the African debt problem. Criticism on slow and insufficient process of the initiative is raised. September 16, 1997 Commonwealth finance ministers meeting in Mauritius. Financial Times (London) September, 20, 1997 U.K. chancellor of the exchequer, Gordon Brown, urges more rapid and generous implementation of the current initiative. May 17, 1998 G8 summit in Birmingham, U.K. The members issue a Financial Times (London), June 2, 1998; Deutsche Presse- statement “supporting the speedy and determined Agentur, May 17, 1998 extension of debt relief to more countries within the terms of the Heavily Indebted Poor Countries (HIPC) initiative. . . . We encourage all eligible countries to take policy measures needed to embark on the process as soon as possible, so that all can be in the process by the year 2000.” September 8 and 10, 1998 World Bank and IMF executive boards. A plan by the Economist, September 12, 1998 bank and IMF is agreed to make more countries eligible for relief. September 16, 1998 UNCTAD Report. The report says debt relief for Africa is Business Day (South Africa), September 17, 1998, UNCTAD still insufficient and too slow, although it welcomes the report (http://www.unctad.org/en/docs/tdr1998_en.pdf) HIPC initiative. It calls for an independent body consisting of eminent persons from the finance and development fields, agreed on by creditors and debtors, to be formed to assess the sustainability of debt. March 16, 1999 U.S. State Department conference. In the presence of Deutsche Presse-Agentur, March 16, 1999 representatives from 46 African nations, President Clinton proposes moves to forgive $70 billion in global debt to those countries advocated by the G7 countries, the IMF, and the World Bank. April 28, 1999 IMF and World Bank spring meeting. The bank and IMF Deutsche Presse-Agentur, April 28, 1999 called for “broader, deeper and faster” debt relief for the HIPCs, but indicated clearly they would take instructions from the industrialized nations on exactly how to do that. MULTILATERAL DEBT RELIEF THROUGH THE EYES OF FINANCIAL MARKETS 1287 Table A1.—(Continued) Date Event Source June 12, 1999 G7 finance meeting in Frankfurt, Germany. An amendment International Herald Tribune, June 14, 1999 of the HIPC initiative is preagreed June 18, 1999 G8 summit in Cologne, Germany. An amendment of the Independent (London), June 19, 1999 HIPC initiative is agreed September 26, 1999 Annual meetings of the IMF and the World Bank Enhanced HIPC is approved by the IMF and the World Bank. September 1, 2000 UN’s Millennium Summit. The statement says it is to help Africa News, February 8, 2005 the HIPC countries achieve the Millennium Development Goals without damaging the financing capacity of the main multilateral finance bodies. February 4, 2005 G7 finance meeting London; The G7 nations are willing to Financial Times (London), February 7, 2005; Africa consider as much as 100% debt relief for the HIPCs, News, February 8, 2005 but disagree on how to finance debt relief, focusing on who should pay, whether additional aid resources should be provided, and how any new mechanisms would influence bilateral and international financial creditors. June 11, 2005 G7 finance meeting in London: The G7 finance ministers Africa News, July 6, 2006; Sunday Independent (Ireland), June agree that as much as 100% debt cancellation for the 12, 2005; Economist, June 18, 2005 HIPCs owed to the World Bank, the IMF, and the AFDB. July 8, 2005 G8 summit in Gleneagles, Scotland. The G8 reaffirmed its World Markets Analysis, July 11, 2005 plan, unveiled last month at the G7 meeting. September 26, 2005 Annual meetings of the IMF and the World Bank. The World Markets Analysis, September 27, 2005; Agence France IMF and the World Bank agree to endorse 100% debt Presse, September 25, 2005 cancellation for the HIPCs. November 7, 2005 IMF executive board. The board reaches consensus on the Public information notice (PIN) no. 05/164, December implementation modalities of the proposal advanced by 8, 2005, http://www.imf.org/external/np/sec/pn/2005/ the G8 and decides to call it the Multilateral Debt pn05164.htm Relief Initiative (MDRI). November 23, 2005 IMF executive board. The board approves the requisite Public information notice, no. 05/164, December 8, 2005, decisions to implement the MDRI in January 2006. http://www.imf.org/external/np/sec/pn/2005/pn05164.htm 1288 Table A2.—Spatial Presence of Parents across African HIPC Countries lic pub Re a tic ic ne c ra e bl ui e o or pu lG ar qu e m ’Iv oon Re a ria asc bi i a a D a la er o pi a to a ag a m da ni nd go bi l go t eD m ng io an ea ua ny i aw an nza a n m ta n h h in ad al oz al g Parent/Country A Co Ca Co Et G G Eq Ke M M M M Rw Ta U Co Za To Aveng 1 — — — — — 1 — — — 2 2 — — 1 — — 2 9 Alexander Forbes — — — — — — — — 2 — — 1 1 — 1 1 — 1 7 Aeci — — — — — 1 — — — — 1 1 1 — 1 — — 2 7 African Life Assurance Company — — — — — 1 — — 6 — — — — — 1 — — 2 10 Afgri — — — — — — — — — — — — — — — — — 1 1 African Oxygen 1 — — — — — — — — — — 1 1 — — — — 1 4 Anglogold Ashanti — — — — — 3 1 — — — 3 — — — 1 — — — 8 African Rainbow Minerals — — — — — — — — — — — — — — — — 3 10 13 Astral Foods — — — — — — — — — — — — — — — — — 1 1 Absa Group 1 — — — — — — — — — — 1 — — 1 — — — 3 Barloworld 1 — — — — — — — — — — 3 2 — — — — 2 8 Business Connexion Group — — — — — — — — — — — 1 — — 1 — — 1 3 Bell Equipment — — — — — — — — — — — 1 1 — — — — 1 3 Bhp Billiton — — — — — — — — — — — 1 — — — — — — 1 Bidvest Group — — — — — — — — — — — 4 3 — 1 — 1 2 11 First Rand — — — — — — — — — — — — — — — — 1 1 2 Gold Fields — — — — — 2 — — — — — — — — — — — — 2 Grindrod 1 — — — — — — — 1 — — 5 1 — 2 — — — 10 Illovo Sugar — — — — — — — — — — — 1 1 — 1 — — 1 4 Imperial Holdings — — — — — — — — — — — 2 — — — — — 2 4 Massmart Holdings — — — — — 1 — — — — — 1 1 — 1 1 — 1 6 Mtn Group — 1 1 1 — — — — — — — — — 1 — 3 — 1 8 Metorex — — — — — — — — — — — — — — — — — 1 1 Nedbank Group — — — — — — — — — — — — 1 — — — — — 1 Nampak — — — — 1 — — — 3 — — 2 1 — 2 — — 3 12 Sabmiller 3 — — — — 1 — — 1 — — 2 1 — 2 1 — 3 14 THE REVIEW OF ECONOMICS AND STATISTICS Standard Bank Group — — — — — 1 — — 1 — — 1 1 — 1 1 1 1 8 Steinhoff International Holdings — — — — — — — — 1 — — 1 — — — — — — 2 Shoprite Holdings 2 — — — — 1 — — — 1 — 1 1 — 2 1 — 1 10 Santam — — — — — — — — — — — — 2 — — — — — 2 Sasol — — — — — — — 1 — — — 5 — — — — — — 6 Sun International — — — — — — — — — — — — — — — — — 1 1 Tongaat-Hulett Group — — — — — — — — — — — 1 — — — — — — 1 Tourism Investment Corporation — — — — — — — — — — — — — — — — — 1 1 Unitrans — — — — — — — — — — — 1 1 — 1 — — — 3 Total 10 1 1 1 1 11 2 1 15 1 6 39 20 1 20 8 6 43 187