By Deborah Brautigam and Yufan Huang On September 29, AidData, a research lab at William & Mary, released a detailed overview of their new data on China’s global lending, Banking on the Belt and Road. The report has generated much commentary. Yet most people will likely read only the headlines: that US$ 385 billion of a purported US$ 676 billion in Chinese loans made to developing countries between 2000 and 2017 was not being reported to the World Bank. “The average government,” the authors contend, “is underreporting its actual and potential repayment obligations to China by an amount that is equivalent to 5.8% of its GDP.” How worried should observers be about this claim? The AidData team has released their data publicly, allowing others to examine the underlying data. In this briefing paper, we examine the data underpinning the AidData conclusions. While we agree with many of the points in their paper, our own analysis of the data puts a very different spin on their headline conclusions. By providing averages, not discussing outliers, and allocating the entire Chinese loan for joint ventures to only the host government partner, the report is unduly alarmist. Three countries account for 61 percent of the debt not disclosed to the World Bank. Out of the US$ 385 billion that AidData calculated as Chinese loans not being disclosed to the World Bank, 61 percent — US$ 235 billion — was committed to just three countries: Venezuela, Russia, and Kazakhstan. Venezuela is not a World Bank borrower and is therefore not required to report its borrowing. AidData’s methodology includes the full value of a Chinese loan to a joint venture investment as the host government’s “underreported debt” contrary to World Bank guidelines. Approximately US$ 153 billion was loaned to joint ventures (JVs) and special purpose vehicles (SPVs), limited liability companies set up for specific projects. Some of these are majority owned by host governments, but in several significant cases they are controlled by private companies — ExxonMobil, for example — with modest shareholdings by host governments. The World Bank does not classify these latter loans as public debt (unless it has a government guarantee), and does not require borrowing governments to report it. Finally, the median figure for “underreported debt” is 1.8 percent of GDP. AidData reports, instead, the average figure of 5.8 percent of GDP. This is distorted by several very large outliers such as 49.7 percent for Equatorial Guinea, 30.6 percent for the Republic of Congo, and 21.1 percent for Venezuela. THE NEW DATA A loan-by-loan dataset on Chinese global lending like the one compiled by AidData is a major contribution. The data (for Africa, at least), appears to be very high quality, reflecting the four years it took to collect and clean it. This data release follows the decision by the World Bank in July 2020 to publish, for the first time, information detailing what each of its borrowers owes to official and private creditors, by creditor country. This means that the World Bank’s International Debt Statistics (IDS), an online repository and annual publication of information about debt held by all the World Bank’s borrower countries, now contains data on loan commitments, outstanding debt, and debt service by creditor country. Yet the World Bank’s aggregated data gives no details about how the loans are used, which banks issued them, or which entity borrowed them. The AidData report includes these details for 165 borrowing countries and territories, and is an important boost to much needed transparency about Chinese development finance.We at the China-Africa Research Initiative (CARI) at Johns Hopkins University and researchers at Boston University maintain two similar loan-level public databases of Chinese lending, but each with a narrower focus than AidData. WHAT ARE “UNDERREPORTED” AND “HIDDEN” LOANS? AidData has introduced two separate concepts in this paper. The first, “underreported” loans, refers to loans that are not reported to the World Bank’s Debtor Reporting System, the source of the data in the IDS. Most of the headline takeaways from the report refer to these “underreported” loans. The second, “hidden” loans, refers to the entire set of loans to host country state-owned enterprises (SOEs) and state-owned banks and SPVs with some degree of host government ownership but without state guarantees. We discuss these separately below. It is important to emphasize that despite its reputation for secrecy, it is not the Chinese government that is “underreporting” or “hiding” these loans. There is no global creditor reporting system for China to use. While members of the Paris-based OECD, an organization of 38 wealthy democracies, are required to report their lending to the OECD’s Creditor Reporting System, China is not an OECD member. “Underreported” Loans. AidData counted US$ 385 billion in “underreported” debt. “Underreported” means that the loans are, by AidData’s assessment, not properly reported by the borrowers to the only global repository for debt information, the World Bank’s IDS. Since 1951, all of the World Bank’s borrowers have been required to provide loan-by-loan data on all loans acquired by public agencies or by the private sector, if guaranteed by a public agency. In turn, countries that do not borrow from the World Bank have no obligation to disclose their debt statistics to the IDS repository. The World Bank has specific rules about the debt data that should be reported. For example, public sector debt includes the central government and all its departments, and regional and municipal governments, and also all borrowing by SOEs (emphasis added) including railways, utilities, and so on. In some cases, host governments have set up SPVs that limit their legal liability for repayment, and take the debt off the central government balance sheet. This mode is increasingly used for public private partnership (PPP) projects that bring private sector investors to partner with the public sector to build and manage large infrastructure projects. The World Bank has cautioned that, “current account practice permits governments to keep the costs and liabilities associated with PPPs off-balance sheet, thus circumventing budgetary constraints and obfuscating scrutiny by the national legislature.” Nevertheless, the World Bank specifies that JV debt or SPV debt should still be reported to the IDS repository, where “the government owns more than 50% of the voting stock” [or] “if, in case of default, the state would become liable for the debt of the institution”. In other words, if the government is the major shareholder or if there is a public guarantee, it should be reported as public debt. Otherwise, it is considered private sector debt. Underreporting can be significant for some countries. Our own work on Zambia describes how the Zambian government has not reported to the World Bank the Chinese debts of its SOEs like ZESCO, the state electricity company, or SPVs like the Kafue Gorge Lower hydropower project, even though the Zambian public sector is the sole shareholder. Yet AidData’s methodology (“Whenever an official sector loan from China is issued to a joint venture, AidData records the entire face value of the loan”) overcounts by including the entire Chinese loan to a JV as a host country debt, even when the host country owns as little as 10 percent of the investment. Since AidData compares their data on Chinese loan commitments to the World Bank’s data, it is important that apples be compared with apples. “Hidden” Loans. AidData’s “hidden” loan definition includes all Chinese lending into SOEs, JVs, and SPVs that are “wholly or partially owned by the host government with implicit host government repayment guarantees”. In other words, this category (US$309 billion) captures borrowing with some degree of host government involvement, but without a formal guarantee. AidData has grouped all Chinese lending to this category of borrower as “hidden”, without regard as to whether the borrowing government is required to report this borrowing to the Debtor Reporting System or not. This leads to some problematic conclusions. For example, referring to Chinese loans mostly to Angola’s state-owned oil company Sonangol, AidData reports that Angola has significant “hidden debt” exposure to China: 11.8 percent of GDP, or US$ 12.4billion. However, the World Bank’s data suggest that Angola has included all of the Chinese Sonangol loans as part of the US$ 59.8 billion in Chinese loan commitments reported to the World Bank. This suggests Angola has not been “underreporting” its significant Chinese borrowing, and it seems inappropriate to call this “hidden” lending simply by definition. HOW SIGNIFICANT IS THE “UNDERREPORTED” LOANS PROBLEM? First, we note that the report emphasized the average “underreporting” without noting how this is affected by outliers. “The average government,” the researchers contend, “is underreporting its actual and potential repayment obligations to China by an amount that is equivalent to 5.8% of its GDP.”However, the median for the same index of countries in the sample is only 1.8 percent. This shows that the sample is skewed by cases like the Republic of Congo (30.6 percent) and Equatorial Guinea (49.7 percent). Second, because the AidData researchers have chosen to use a different methodology from the World Bank, their data on “underreported loans” naturally includes significant numbers of loans that the World Bank’s own guidelines rule out as public and publicly guaranteed lending. Many of these are hard to characterize as cases where the host government will feel pressured to take over the loan, i.e., significant contingent liabilities for the host government, as the cases below make clear. The Coral South Liquid Natural Gas Project in Mozambique. Take the case of a US$ 4.66 billion syndicated loan to the US$ 8.89 billion Coral South floating liquid natural gas (LNG) project in Mozambican waters. The project is a JV led by the American oil giant Exxon Mobil and the Italian multinational energy company Eni. Mozambique’s national energy company owns just 10 percent. In 2017, a syndicate of 15 international banks and five export credit agencies provided a US$ 4.66 billion loan for the Coral South LNG project. The Chinese portion of this was US$ 1.55billion. AidData includes this figure of US$ 1.55 billion in their totals for Mozambique’s Chinese loans. Since Mozambique has not reported this to the World Bank (following World Bank rules), it becomes, in AidData’s analysis, “underreported debt”. In this case, there is no evidence that the Mozambique government guaranteed any part of this loan, so by definition it also becomes “hidden” debt. If the government didn’t provide any sovereign guarantee for the loan, then none of it should be captured by the World Bank database since the state’s equity is less than 50 percent. The Husab Uranium Mine in Namibia. In 2012, China Development Bank provided a US$ 1.175 billion loan to Taurus Mineral Ltd., a Chinese company, which Taurus used to buy the Husab Uranium Mine in Namibia from an Australian firm. The Namibian government later bought 10 percent of the Husab Uranium Mine. It is unclear why AidData included this loan as a liability of Namibia since the borrower was clearly a Chinese entity, but that individual loan made up 69 percent of what AidData claimed as the Namibian government’s “underreported” debt exposure to China. The Democratic Republic of the Congo: Sicomines. In 2008, the Democratic Republic of the Congo’s stateowned mining company Gécamines and a consortium of Chinese companies set up a jointly owned mining venture, Sicomines. The Chinese hold 68 percent of the Sicomines shares, while Gécamines and other Congolese interests hold 32 percent. China Eximbank and the Chinese company consortium provided US$2.9 billion in loans to Sicomines to bring an abandoned copper mine back into production. There is no DRC government guarantee for these investment loans. For that reason, the DRC wouldn’t need to report this loan to the World Bank as its liability. However, AidData included US$ 2.83 billion of these loans as “underreported” Chinese commitments for the Democratic Republic of Congo. Kazakhstan-China Gas and Oil Pipelines. In Kazakhstan a JV, Asia Gas Pipeline LLP, owned 50 percent by China National Petroleum Corporation (CNPC) and 50 percent by the Kazakhstan SOE borrowed US$ 12.2billion from China Development Bank and Bank of China to build sections of the Turkmenistan-China Gas Pipeline. According to the World Bank’s rules, since Kazakhstan does not have a controlling share of this JV (the shares are split evenly, at 50:50), and the state didn’t provide a sovereign guarantee for these loans, they are not required to report this debt to the IDS. Yet AidData included this as a large portion of Kazakhstan’s “underreported” debt. The China-Laos Railway Project. AidData’s rationale for these decisions is clear from their detailed discussion of the US$ 5.9 billion China-Laos Railway Project. A JV between the Laos government, which holds 30percent of the equity, and three Chinese companies, the railway borrowed US$ 3.54 billion from China Eximbank to build one of the flagship Belt and Road Initiative projects: a railway that is planned to extend from China’s Yunnan Province through Southeast Asia. Again, the Laos government is not required to report this loan to the World Bank since it did not provide a guarantee for it, but AidData included the whole US$3.54 billion as Laos’ liability and therefore part of its “hidden” and “underreported” debt. The AidData researchers argue that they included the whole JV loan as Laos’ liability because if the railway is unable to repay the loan, the Laos government will likely face pressure to cover all of the losses, despite the majority ownership by the three Chinese firms and its significant role as a transport link between China’s southern provinces and Southeast Asia. This strikes us as an unreasonable assumption. WHAT IS DRIVING “UNDERREPORTED DEBT”? There are several reasons why countries have not reported debts to the World Bank: they are not a World Bank borrower, geopolitics, weaknesses in capacity, and local accounting rules. Understanding these factors helps provide important context for the discussion of “underreported” debt. First, the World Bank rules for the Debtor Reporting System only apply to countries that are actually borrowing from the World Bank. Under leftist President Hugo Chavez, Venezuela pulled out of the Washington-based World Bank for ideological and political reasons in 2007, a5er paying off its loans early. AidData has identified US$ 90 billion in underreported Chinese lending to Venezuela, or 21.1percent of its GDP. Yet because it is not a borrower, Venezuela is not required to report its Chinese loans to the World Bank. This is also probably why Equatorial Guinea, which is not borrowing from the World Bank and was recorded by AidData as having 49.7 percent of its GDP in “underreported” debt, does not report its loans to the World Bank. Among the 154 mid- and low-income countries and territories that AidData included in the “underreported debt” calculation, 36 apparently do not report any borrowing from any source to the World Bank. Their Chinese loans, all labeled as “underreported” by AidData, amount to US$ 31 billion. Second, geopolitics may also affect whether countries report to the World Bank. This clearly affected the Venezuela decision under the Chavez government to withdraw from the World Bank. For unknown reasons, Russia also does not report fully to the World Bank (this was the case both before Russia’s annexation of Crimea, which, given sanctions, would have affected its ability to borrow from the World Bank, and afterwards). In 2021, the World Bank’s IDS noted that World Bank staff had to estimate Russia’s debt obligations. Third, countries with weak government capacities or beset by civil wars are o5en unable to report their borrowing in a timely and accurate fashion. For example Sudan, which ranked near the bottom of the World Bank’s “debt policy index,” a measure of a government’s capacity to manage its debt, stopped reporting loan commitments by China after 2010. Fourth, some countries, Mozambique most notoriously, have deliberately hidden foreign borrowing, in this case guaranteed loans to SOEs from Swiss and Russian banks. However, other countries may not report their SOE loans because their own domestic accounting systems keep SOE debt distinct from central government public debt, and their debt management offices may not even have access to SOE borrowing data. As the World Bank noted: [I]n most instances any omission in reporting does not signal any unwillingness to report, but rather, reflects the legal framework that governs contracting, measuring and monitoring public debt at the national level, and the mandate of the public debt office. For example, in relatively well-governed Botswana, the Ministry of Finance and Economic Development does not report borrowing by its state-owned Botswana Power Corporation to the World Bank. Finally, a recent paper by a group of researchers from AidData, the Center for Global Development, and elsewhere concluded that confidentiality clauses in Chinese loan contracts could prevent borrower governments from “revealing the terms or even the existence of the debt.” But we need more evidence for that claim. For instance, 45 of the 83 loan contracts containing confidentiality clauses were from Cameroon. That doesn’t seem to have prevented Cameroon from reporting its Chinese loan commitments fully to the World Bank. CONCLUSION So-called “underreported” or “hidden” debt (as noted, AidData defines the two terms differently) is a complex issue that involves borrower governments’ capacities, their relations with the World Bank, and concern about potential liabilities that may not be included in the Bank’s reporting guidelines. “Hidden debt” or “underreported debt” isn’t necessarily hidden from the public in the borrower country or even from the IMF. After all, researchers at AidData, CARI, and Boston University have been able to identify these loans using government websites, IMF country reports, and other sources. At the end of the day, what drives “underreported debt” is the general failure of borrowing countries to report consistently to the World Bank, and the World Bank’s lack of power to force them to do so, particularly if they are not borrowing from the World Bank. However, a significant amount of what AidData calls “underreported debt” is due to the researchers’ decision to include the entire amount of a Chinese loan to a JV where the host government is not a controlling shareholder, in AidData’s totals for Chinese loans to that government. This goes against the World Bank’s own rules. “Hidden debt” is another matter. As a World Bank blog post noted in 2017, political leaders around the world, including in wealthy countries, “are jumping on the PPP bandwagon and pressuring public sector servants to launch PPPs.” AidData is right to be concerned about cases where critical public infrastructure is being built as a PPP and financed through an SPV company without a public guarantee but with the strong expectation that the host government will bail out the company if it is in trouble. These contingent liabilities are challenges for accountability. However, defining any debt held by a company with some government ownership but that has no government guarantee as “hidden” on the assumption that the government could still be pressured to pay may be pushing the notion of contingent liability too far. In Mozambique, it is unlikely that Mozambique’s state-owned oil company will be required to bail out Exxonmobil and Italy’s ENI if the Coral South LNG project fails. For transparency to be a global norm, perhaps a different data repository system is required. The World Bank’s data has a serious flaw in that only low and middle-income World Bank borrowers are obliged to report. At present some 128 countries are included, compared with 193 United Nations member countries. Furthermore, if a country is not a borrower, or goes for long periods without borrowing, the World Bank has little leverage to compel transparency. An alternative could be to evolve a norm of transparency by requiring all 190 members of the International Monetary Fund, as a uniform condition of membership, to transparently report all external borrowing, perhaps as part of their annual Article IV consultations. The Chinese government should publicly encourage countries to report all their Chinese borrowing to the World Bank’s Debtor Reporting System, and make it clear that the confidentiality clauses in some loan contracts do not prohibit borrowers from doing so. Having Beijing come down firmly in favor of debt transparency at least at an aggregate level, as reported in the IDS, will help all lenders, including Chinese banks, to better assess the debt risks. AidData has produced a rich source of data on China’s overseas development finance that can be mined for a host of analytical purposes. We applaud their hard work and the high quality of this year’s release. Yet their presentation of their data without important context, using averages without highlighting outliers, and choosing to include Chinese loans to JVs as full liabilities of the host government, despite those government’s minority shares, and against World Bank rules, were unfortunate choices that could hamper greater understanding of this important issue. DEBORAH BRAUTIGAM is the Bernard L. Schwartz Professor of International Political Economy and the Director of the International Development Program and founding director of the China Africa Research Initiative (CARI) at Johns Hopkins SAIS. Her Ph.D. is from the Fletcher School of Law and Diplomacy, Tu!s University. YUFAN HUANG is a Ph.D. student at the Government Department of Cornell University. From 2018 to 2020, he was a researcher with CARI. Before that, he was a researcher on foreign and defense a”airs at the New York Times Beijing Bureau. Yufan has an M.A. in China Studies from Johns Hopkins SAIS, and a B.A. in International Politics from Renmin University of China.