The Discipline of Saying No
China’s Credit Rating Crackdown and the Historical Limits of Evaluative Authority
Executive Summary
China’s recent crackdown on credit rating inflation has been widely interpreted as a technical effort to improve the quality of domestic credit ratings. This essay argues that the reforms reveal a deeper institutional question.
Drawing on new evidence from China’s bond market alongside the historical transition to issuer-pays in the United States during the 1970s, the essay examines why two remarkably similar episodes of rating inflation produced fundamentally different regulatory responses. While Chinese authorities are actively reviewing AAA ratings, encouraging greater differentiation and experimenting with alternative institutional arrangements, the United States embedded Moody’s and S&P more deeply into the regulatory architecture despite evidence that issuer-pays had also inflated ratings.
The explanation proposed here lies in the historical development of evaluative institutions. The credit rating agencies in the United States (mostly) entered the issuer-pays era immediately after the collapse of the National Credit Office, whose failure during the Penn Central crisis demonstrated the systemic consequences of losing a functioning evaluator. China, by contrast, retained the institutional capacity to reconstruct its domestic rating system before confidence had collapsed completely.
The essay argues that the value of a credit rating agency extends beyond statistical accuracy. Evaluative institutions derive authority from the market’s confidence that they remain willing to issue adverse judgements when the evidence requires them. Where that capacity weakens, the informational value of ratings deteriorates regardless of the formal rating scale. China’s reforms can therefore be understood as an attempt to restore confidence in the evaluative function of its domestic credit rating agencies while preserving state oversight of the financial system.
The comparison ultimately raises a broader question that extends well beyond credit ratings: under what conditions do states tolerate weaknesses in evaluative institutions, and when do they decide those institutions must instead be rebuilt?
The Discipline of Saying No
Chinese regulators spent the early summer of 2026 opening the files of every issuer holding the highest available credit rating and asking whether it still deserves the label. Caixin reported on 8 July that regulators had begun broad financial reviews of companies awarded AAA grades, screening out issuers that lack sufficient grounds for such ratings, and pressing China’s credit rating agencies toward stricter standards. Authorities are examining any company whose bond yields sit more than 200 basis points above government debt of comparable maturity, a group that includes many property developers and local government financing vehicles. As of the end of June, the Chinese bond market had recorded twenty-eight rating downgrades this year, already well above the total of nine for all of 2025. More than 220 companies withdrew rating requests as the risk of a downgrade rose, and CSC Financial has projected that as much as 494.6 billion yuan of corporate bonds could face downgrade pressure in the second half of the year.
The scale of the problem being corrected is not modest. As of the first quarter of 2026, 27 per cent of China’s roughly 6,000 bond issuers carried AAA ratings, and a further 32 per cent carried AA+, meaning well over half of all rated issuers in the country sat in the top two tiers of a scale meant to differentiate risk. In the first half of last year, 90 per cent of newly rated corporate bonds received AAA grades, up from under half a decade earlier and the highest share on record dating back to 2008 (Hale and Wang, 2025). Kaihua Deng of Renmin University, whose research covers China’s rating practices, put the underlying mechanism plainly to the Financial Times: companies shop across multiple agencies for better grades, and borrowing costs have stayed roughly flat relative to government debt even as the share of top ratings climbed, which is not what falling risk would produce. Defaults by AAA-rated issuers, including Henan’s Yongcheng Coal and Electricity Holding Group and Tsinghua Unigroup, shook confidence in a market that had assumed the top grade meant something close to zero risk.
The academic literature China’s own reformers draw on treats this pattern as a textbook case of ratings inflation produced by the issuer-pays model. Zhou, Wang and Zhang (2024) document that ratings and default outcomes moved in opposite directions before regulators intervened: 585 of 920 defaulting corporate bonds carried a rating of AA or higher, a default rate among supposedly safe issuers of 63.59 per cent. Liu and Wang (2021), using default-probability-implied equivalents instead of simple notch comparisons with global scales, find that Chinese domestic agency ratings sit roughly ten notches above what the same default probabilities would earn from S&P Global, and that firms rated AAA, AA, A and BBB domestically default at rates of 0.41, 0.76, 12.83 and 18.59 per cent respectively, a monotonic relationship so steep that the middle categories are barely differentiated from each other at all. Liu and Wang (2022) trace a continuous loosening of issuer-paid rating standards between 2006 and 2020, with average ratings rising by 1.73 notches for firms of identical underlying strength, concentrated in the years of heaviest regulatory reliance on ratings as gatekeeping devices. Bush (2022) summarises the mechanism bluntly: China’s credit rating industry combines acute conflicts of interest, weak independence, an oligopolistic structure, and until recently, limited relevance to actual bond pricing.
A rival account exists, and it deserves a hearing before the argument proceeds. Zelezetskii and Rutledge, writing shortly after the Financial Times published its own account of the same phenomenon, argue that comparing China’s 90 per cent AAA share against the roughly 2 per cent of American corporates rated global AAA mistakes a scale difference for a quality difference. China Chengxin International, one of China’s largest domestic credit rating agencies, rates the same companies on both national and global scales, and the gap runs a consistent six to seven notches: China Three Gorges Corporation earns AAA locally and A+ on the global scale, and State Grid, China National Petroleum, and China State Construction Engineering all show the identical pattern. On this reading, Chinese domestic AAA maps to roughly global BBB- through A+, and the FT’s comparison resembles comparing Celsius to Fahrenheit without conversion. Jiang and Packer (2017), whose BIS working paper is the most careful empirical treatment of the scale question, confirm the gap: the mean domestic rating exceeds the mean global rating by more than seven notches for jointly rated firms, and 81 per cent of domestic ratings cluster at AAA, AA- or AA, against roughly 25 per cent for the same firms’ Greater China ratings from global credit rating agencies and zero per cent for their pure global ratings.
The scale-difference argument and the inflation argument are not, on inspection, mutually exclusive, and the distinction between them is where the interesting question actually starts. A stable calibration gap of six or seven notches, if it held constant over time, would be a fact about how two rating systems translate into each other, not evidence of an eroding standard. What Liu and Wang (2021) and Liu and Wang (2022) document is not a constant gap but a widening one: ratings rose 1.73 notches for firms of unchanged fundamental strength over fourteen years, and default probabilities within each domestic rating band spread out instead of staying tight, meaning firms facing genuinely different risks of default were being pooled into the same letter grade. A translation problem does not produce that pattern. A translation problem is static. What the data shows is dynamic, and the dynamic is the one that regulators are now acting against.
That the Chinese state itself has settled this debate in practice, whatever the merits of the scholarly dispute, is worth considering strongly. If domestic AAA merely occupied a different position on a stable, well-understood scale, there would be no reason for the PBOC and CSRC to launch financial reviews of the companies holding the grade, no reason for issuers to withdraw 220 rating requests in fear of a downgrade, and no reason for CSC Financial to model half a trillion yuan in exposure. Regulators are not behaving like officials who discovered a labelling convention. They are behaving like officials who discovered that a signal load-bearing enough to determine bank capital treatment, insurance eligibility, and repo access had stopped tracking the thing it was supposed to track. Whatever one makes of the calibration argument, the crackdown itself is evidence of a judgement already reached inside the system that produces the ratings.
A Familiar Problem, A Different Response
This is the point at which the essay’s actual subject comes into view, because rating inflation of a very similar character, driven by a nearly identical mechanism, has already happened once before, in a market regulators did not choose to reconstruct. This is not to say it has not happened across the history of the credit rating sector of course, but to say it has been documented and proven only in very particular circumstances. The most revealing is perhaps around the late 1960s/early 1970s. Jiang, Stanford and Xie (2012) study the moment Standard & Poor’s switched from charging investors to charging issuers for corporate bond ratings, in July 1974, four years after Moody’s made the identical switch (Moody’s started with municipal ratings first). Using a sample of 797 corporate bonds rated by both credit rating agencies between 1971 and 1978, they exploit the fact that Moody’s revenue model held constant across the whole period while S&P’s did not, which lets Moody’s ratings serve as a benchmark unaffected by the change under study. Before July 1974, when Moody’s alone charged issuers, Moody’s ratings for the same bond ran higher than S&P’s by roughly 11 per cent of a full rating grade. After S&P adopted issuer-pay, that gap closed, and the two agencies’ ratings became statistically indistinguishable. The authors are careful to rule out a general drift in credit conditions: because the comparison is bond by bond against a Moody’s rating that did not change its own revenue model during the window, the convergence has to come from S&P moving up, not from some market-wide shift both agencies were tracking.
The cross-sectional evidence is the part of the paper that matters most for the comparison with China, because it shows the inflation was not diffuse. S&P’s ratings rose by approximately 20 per cent of a rating grade overall, but that increase concentrated almost entirely in bonds subject to the greatest potential conflicts of interest, measured two ways: issue size and issue frequency above the sample median (a proxy for bargaining power over the credit rating agency), and low credit quality within a given Moody’s rating band (a proxy for issuers with the most to gain from a nudge upward). For bonds expected to generate high fees, S&P’s rating relative to Moody’s moved from minus 0.20 to near zero, a full 20 per cent of a grade. For bonds of comparatively weak credit quality within their category, the shift ran even larger, 25 per cent of a grade. The magnitude translated into real money: roughly 10 basis points of yield spread, or about $51,000 a year in reduced interest costs for the average issuer in the sample, worth over $222,000 in 2010 dollars. This is the mechanism the Chinese literature describes under different names. Zhou et al. (2024) call it rating shopping. Bush (2022) calls it commercial pressure overriding professional integrity. Jiang, Stanford and Xie call it, more simply, bargaining power that issuers acquire the moment they become the ones paying the bill.
The 1974 switch was not, however, the American market’s first experience of what an issuer-paid evaluator without a serious competitor could do, and the earlier episode did more to shape the regulatory response than the literature on issuer-pays inflation has generally acknowledged. Cash’s (2024) account of Moody’s history traces the commercial paper market, adjacent to the corporate bonds Jiang, Stanford and Xie studied but governed by an entirely separate evaluator, back to a single national gatekeeper: the National Credit Office (NCO), a subsidiary of Dun & Bradstreet that had rated commercial paper since 1920. A Congressional investigation conducted by the SEC staff after 1970 found that dealers required issuers to obtain an NCO rating at the issuer’s own expense, because many institutional buyers were bound by statute or by their own governing boards to purchase only paper NCO had rated prime (U.S. Senate, 1972). NCO operated its own issuer-pays subscription model decades before Moody’s or S&P adopted the structure for bonds: issuers signed an agreement, paid an annual fee, and supplied NCO with the same financial information they gave their banks (U.S. Senate, 1972). The market NCO alone rated was growing fast enough to strain any evaluator, with total issuances rising from 227 in 1967 to 615 in 1970 and the market as a whole reaching roughly $40 billion (U.S. Senate, 1972).
By the middle of 1970, NCO had 651 outstanding ratings, and all but 34 of them carried the top grade, Prime (U.S. Senate, 1972), a concentration at the summit of a five-tier scale running Prime, Desirable, Satisfactory, Fair, and No Rating that sits remarkably close to the 90 per cent AAA share that later drew attention to Chinese domestic ratings. Eugene Schenk, NCO’s president, told the Congressional investigators that virtually every prospective buyer in the market relied on NCO’s appraisals (U.S. Senate, 1972), a claim of authority that sat awkwardly against the resources behind it. NCO’s commercial paper department consisted of three or four analysts. Its head from September 1969, Rudolph Merker, had spent forty-two years at NCO’s retail credit reporting arm, held no college degree, was not a chartered financial analyst, and received no training or written guidance when he inherited the role (U.S. Senate, 1972). His predecessor had been physically present in the office only twice in the preceding four years.
This was the evaluator standing between investors and the Penn Central Transportation Company, the country’s largest railroad and sixth-largest nonfinancial corporation, as its finances deteriorated through 1969 and 1970. Merker’s principal source of reassurance throughout was Jack Vogel of Goldman, Sachs, the dealer selling Penn Central’s commercial paper to investors, and Merker told the Congressional investigators he saw no conflict of interest in relying on Vogel’s word despite knowing Goldman was the exclusive seller of the paper NCO was rating (U.S. Senate, 1972). Goldman had quietly sold down its own inventory of Penn Central paper by February 1970 while continuing to tell NCO the company remained sound. On 15 May 1970, Standard & Poor’s downgraded Penn Central’s bond rating outright. NCO’s response was to telephone Vogel and ask whether he still felt the same way; when he said yes, NCO kept its Prime rating in place. Only on 1 June 1970, after Penn Central’s own management stopped answering NCO’s requests for information, did NCO act, and even then it did not downgrade the paper. It simply stopped rating it. Penn Central filed for reorganisation three weeks later, on 21 June 1970, holding a record $82 million of commercial paper it could no longer roll over.
The fallout was systemic, not confined to Penn Central’s own creditors. Investors who had trusted an NCO prime rating on other issuers lost confidence across the board, and nonbank commercial paper outstanding contracted by roughly $3 billion, close to ten per cent of the market, in the first three weeks of July 1970, a run serious enough that the Federal Reserve intervened directly to prevent it becoming a wider credit collapse (Handal, 1972). On 23 August 1971, Dun & Bradstreet transferred NCO’s commercial paper rating function to Moody’s, its other rating subsidiary, and NCO effectively ceased to exist as an evaluator (U.S. Senate, 1972). Moody’s changed the rating symbols to distance its new offering from NCO’s discredited ones, and independent research subsequently found that standards in the commercial paper market improved considerably once Moody’s took over (Handal, 1972).
The detail worth pausing on is the timeline, because it changes what the SEC’s later tolerance of issuer-pay inflation actually represents. Moody’s had already adopted the issuer-pays model for its own corporate and municipal bond ratings, just months before Penn Central’s bankruptcy and roughly a year before it absorbed NCO’s commercial paper function. The market’s replacement evaluator was another issuer-paid institution, not a clean alternative to the conflict that had just destroyed NCO, distinguished from NCO chiefly by scale, by decades of accumulated expertise in bond analysis, and by a staff and structure NCO’s three-person commercial paper unit had never possessed. When the Securities and Exchange Commission created the ‘Nationally Recognized Statistical Rating Organization’ designation in 1975, an internal category devised for the narrow purpose of grading securities under the net capital rule (Securities and Exchange Commission, 1997), it did so five years after a national market had relearned, at the cost of a multi-billion-dollar run and direct Federal Reserve intervention, what happens when its sole evaluator loses the capacity to say no. NCO itself was never considered for the designation; it no longer existed to be considered. What existed instead was Moody’s, an issuer-paid institution but a demonstrably functioning one, freshly proven capable of absorbing NCO’s discredited role and tightening standards where NCO had let them collapse.
Read against that history, what happened to Moody’s and S&P after 1974 looks considerably less puzzling than it does in isolation. Regulators did not launch a review of every AAA-rated bond retroactively assigned during the years of documented inflation. Congress did not force a return to investor-pay. Instead, NRSRO recognition fed directly into bank capital rules, money market fund eligibility, and a wide range of federal regulation, with the practical effect of hard-wiring the two conflicted incumbents into the architecture of American finance instead of displacing them. Smaller, newer credit rating agencies working on the investor-pay model, such as Duff and Phelps, continued charging investors into the 1980s, but NRSRO recognition gave Moody’s and S&P an initial regulatory moat that made competing on an uninflated standard commercially irrelevant (Duff and Phelps would gain NRSRO status in 1982). The documented conflict of interest became, if anything, more consequential over the following three decades, as the two agencies expanded from corporate and municipal bonds into structured products, where issuer-pay fees eventually made up the largest share of Moody’s revenue by 2008, the same year the ratings assigned to mortgage-backed securities collapsed in value on a scale nobody in 1974, or in 1970, could have modelled.
The Authority to Refuse
The puzzle, stated plainly, is this. Measurable rating inflation, driven by an issuer-pays fee structure and concentrated among the issuers with the most bargaining power, is documented in NCO’s commercial paper market in 1970, in S&P’s corporate bond ratings after 1974, and in China’s domestic bond market before 2026. In two of the three cases, the state’s response was to embed the conflicted institution more deeply into its own regulatory machinery, or to make no institutional response at all beyond absorbing the wreckage. In the third, the state’s response is retroactive review, licence suspension, forced downgrades, and an active experiment in an alternative payment model. Rating inflation does not, on this evidence, inevitably destroy the credibility of the institution that produces it. Something else determines whether inflation becomes tolerable background noise, a catastrophe absorbed and then repeated in modified form, or an emergency requiring reconstruction.
One candidate answer treats credibility as a function of average accuracy, the proportion of ratings that turn out, in retrospect, to have matched what actually happened to the rated debt. On that account, an institution earns authority by being right most of the time and forfeits it by being wrong often enough. The comparison assembled so far does not support this account particularly well. Most AAA-rated bonds did not default in the United States after 1974, and most did not default in China before the current crackdown either; even NCO, judged purely on the proportion of its Prime ratings that avoided default, would not look catastrophic. What separated the case NCO could not survive, the case the American state chose to tolerate afterward, and the case the Chinese state is choosing to reconstruct looks closer to a capacity, exercised, withdrawn too late, or left undemonstrated in each of the three settings, to issue a judgement the rated party did not want and could not buy off, than to any difference in the frequency of error.
Consider what a credit rating actually promises a market, and consider it against Eugene Schenk’s own words on the eve of Penn Central’s collapse: that NCO was the agency on which virtually all buyers relied, and that its authoritative appraisals had been of material assistance in making the market for these notes (U.S. Senate, 1972). A rating promises that the institution assigning it is willing, when the evidence warrants it, to publish a conclusion the issuer will resent, may punish by withdrawing future business, and cannot prevent. Much of the institutional architecture surrounding ratings ultimately exists to sustain confidence that an evaluator remains willing to issue an unwelcome judgement when the evidence demands it. NCO’s own record shows exactly what happens when the promise goes unfulfilled: a rating held at Prime through a first-quarter loss of seventeen million dollars, held through a second agency’s downgrade of the same borrower’s bonds, and only withdrawn, not lowered, three weeks before bankruptcy. The same pattern recurs in China, on an incomparably larger market and half a century later. Liu and Wang (2021) find that the median domestic rating on eventually-defaulting Chinese issuers held at AA until roughly eight months before default, slipped to AA- with six months remaining, and did not fall to BB until one month before the event itself. A system that recognises distress a single month in advance, in Shanghai in the 2010s or in New York in 1970, has done more than simply misprice risk. It has demonstrated, issuer after issuer, that its capacity to say no early enough to matter does not function. What the Chinese state is now correcting is a set of grades sitting too high on some abstract ladder, and beneath that, a documented absence of the one thing a rating system exists to supply, the same absence NCO demonstrated first.
This reframing is offered as a proposition to be tested against the full comparison, not asserted from one side of it, and it should not be pushed further than the evidence allows. If the capacity to refuse is what evaluative institutions are actually selling, the market’s own history explains why Moody’s and S&P’s compromised version of that capacity survived where NCO’s had not. NCO lost the capacity to refuse outright: no downgrade preceded the default it existed to warn against, only a last-minute suspension issued after the borrower stopped answering the telephone. Moody’s, by contrast, retained and exercised the capacity throughout the 1970s. Jiang, Stanford and Xie’s evidence shows S&P declining to exercise it only for a defined, traceable population of high-fee and low-quality bonds, worth roughly twenty per cent of a rating grade, while continuing to exercise it everywhere else in the portfolio. The refusal capacity was compromised at the margin in the 1970s bond market. It had been extinguished at the centre in the 1970 commercial paper market, five years earlier, in a run the market had not yet forgotten. Seen against that recent history, the SEC’s 1975 decision to grant Moody’s and S&P NRSRO status looks less like a regulator overlooking a demonstrated failure and more like a calculation, explicit or not, that a partially compromised refusal capacity, housed inside an institution that had just absorbed and stabilised a fully collapsed one, was worth considerably more than the alternative the market had already tested and rejected. Whether that calculation was correct is a separate question from whether it was made, and the essay returns to it once the Chinese evidence is in view.
Rebuilding Evaluative Authority
China’s own regulatory history offers a first clue, because the state’s relationship to its domestic credit rating agencies long predates the current crackdown and has never been a hands-off one. Kennedy (2008) traces the industry back to 1987, when the PBOC’s provincial branches began spinning off credit rating departments, and shows that the number of licensed credit rating agencies swelled past ninety before a wave of defaults in the mid-1990s, concentrated in Liaoning and Jilin, forced local governments and PBOC branches to absorb bailout costs estimated between RMB 3 and 8 billion. The PBOC’s response in December 1997 was to license only nine of the existing fifty firms to rate publicly issued bonds, and further consolidation after the Asian financial crisis cut the number to five accredited credit rating agencies by the early 2000s: China Chengxin, Dagong, China Lianhe, Shanghai Brilliance, and Shanghai Far East. Kennedy’s central argument is that Chinese credit rating agencies never developed independent market authority because the state’s own approval architecture made ratings largely beside the point. Corporate bond issuance required NDRC approval keyed to national investment priorities, and issuers additionally needed a state-linked guarantor obliged to cover any default, a structure that removed both the risk and the informational function ratings exist to serve. One credit rating agency executive, quoted by Kennedy, called the system what it was: a planned economy in which there was no role for credit rating agencies at all.
That history matters because it means the poor reputation of Chinese credit rating agencies, going into the 2000s bond market expansion, was substantially a product of decades of state policy, not conventional market failure. Kennedy’s argument that state policy is the ultimate source of the industry’s weak standing is worth taking at face value, because it reframes the current crackdown as the state attempting to repair damage its own prior architecture caused. The PBOC’s 2005 reforms, introducing short-term commercial paper that any qualified firm could issue without a state guarantor, and stipulating for the first time that a credit rating was required to trade the instrument on the interbank market, created the actual demand for ratings that had never previously existed. Once regulators made rating-contingent access the norm across repo eligibility, insurance investment, and money market fund purchases, exactly the conditions Jiang and Packer (2017) describe as generating a hard AAA threshold for public bond issuance, they built the regulatory floor that later produced the clustering at the top of the scale. The state that is now cracking down on inflation is, to a considerable degree, cracking down on an outcome its own earlier rules made close to inevitable.
The 2010 creation of China Bond Rating, established by the National Association of Financial Market Institutional Investors under PBOC guidance as the country’s first investor-paid agency, was the first serious attempt to build an alternative to that structure from within. Lu et al. (2026) compare CBR’s rating actions against those of the incumbent issuer-paid credit rating agencies across 2012 to 2024 and find three consistent differences: CBR issues negative rating actions more often, particularly for state-owned enterprises, non-listed issuers, and firms under short-term debt pressure; when CBR does downgrade, the adjustment tends to be more severe; and bond markets react more strongly to CBR’s negative actions, producing larger abnormal price declines than comparable moves from issuer-paid credit rating agencies. The authors’ conclusion identifies exactly the capacity this essay has argued is the real asset at stake. Investor-paid ratings in China carry their advantage through a demonstrated willingness to reveal adverse credit information more readily, more forcefully, and with greater market credibility than the issuer-paid alternative, a willingness that shows up in the severity of the downgrades and the market’s reaction to them far more than in any systematic gap between the average ratings the two models assign. CBR is, in other words, the institution China built specifically to demonstrate the refusal that the incumbent agencies had stopped demonstrating.
The state layered a second mechanism on top of CBR in August 2021, when the PBOC and six other departments jointly issued guidance encouraging issuers to select two or more credit rating agencies, explicitly to draw on the cross-validation effects of dual and multiple ratings. Zhou, Wang and Zhang (2024) find that the dual-rating system had a measurable disciplining effect: it reduced the frequency of rating upgrades, increased the frequency of downgrades, and widened the gap between how far agencies were willing to move ratings in each direction, all consistent with a system in which agencies could no longer count on a single relationship with an issuer insulating them from a competing agency’s more accurate assessment. The same notice reduced default rates among previously high-rated bonds, from an implied pre-notice rate near 63.59 per cent of defaults occurring among AA-and-above issuers to a measurably lower share afterward, evidence that the reform changed outcomes and did not merely relabel risk.
Layered again on top of the domestic reforms is a slow, carefully bounded opening to foreign entry. Fitch and China Chengxin’s joint venture in the late 1990s collapsed in 2003; Moody’s took a 49 per cent stake in Chengxin from 2006, and Fitch a 49 per cent stake in Lianhe from 2007, arrangements that gave the global credit rating agencies a minority financial interest without operational control, since foreign firms have been barred since 2007 from holding controlling stakes in domestic agencies under the NDRC’s restricted-industries catalogue. S&P Global received approval in January 2019 to establish the first wholly foreign-owned credit rating operation permitted to rate domestic Chinese bonds directly, described at the time by S&P’s China chief executive Simon Jin as necessarily using a China-specific national scale instead of S&P’s global scale, in recognition, in his words, of the size and complexity of China’s capital markets. The distinction Jin drew, deliberately, is the same distinction Zelezetskii and Rutledge later leaned on to defend the FT’s numbers: a national-scale AAA and a global-scale AAA were never meant to be interchangeable, even when the same institution assigns both.
The direction of travel became even clearer in July 2026. The National Association of Financial Market Institutional Investors announced that credit rating agencies seeking registration for panda bonds would, from 1 August, be required to publish explicit mappings between their domestic rating scales and internationally recognised credit-rating scales, alongside clearer disclosure of their rating definitions. Reports failing to include the mapping would no longer be accepted for registration (Yang, 2026). The measure is notable because it neither abandons China’s national-scale approach nor accepts that domestic and international scales should converge. Instead, it requires agencies to make the relationship between them transparent. At precisely the moment regulators are tightening standards within the domestic market, they are also making domestic evaluations more interpretable to international investors. The objective appears to be neither standardisation nor replacement, but greater intelligibility across evaluative systems as China's capital markets become increasingly international.
None of this amounts to China simply adopting a more market-oriented rating regime, and treating it that way would miss what makes the case analytically interesting. CBR, the dual-rating requirement, and foreign entry on terms including a China-specific scale were each initiated from within the state, not extracted from it by investor pressure, and the AAA reviews of 2026 are being conducted by the PBOC and securities regulators directly, not delegated to an independent body insulated from political direction. Zerlina Zeng of CreditSights captured the underlying logic when she told the Financial Times that Beijing’s central objective is to prevent local government financing vehicles and weaker issuers from tapping the market at all, because the government does not want to see defaults. The state is manufacturing a refusal capacity it can direct, sorting issuers the way it wants them sorted while retaining the authority to decide who is actually permitted to fail.
The tension is worth stating without resolving it, because resolving it prematurely would flatten what is actually a genuinely difficult institutional problem. A state that wants credible differentiation between strong and weak borrowers needs an evaluative institution willing to assign a low rating to a politically connected state-owned enterprise or a local government financing vehicle, precisely the issuers Kennedy (2008) and Lee (2016) both identify as the ones Chinese credit rating agencies have historically been least willing to downgrade, given the reliance on political relationships, or guanxi, that both scholars describe as central to how domestic agencies have competed for business. Yet the same state has never given up the power to decide, through the guarantor requirements, implicit guarantees, and NDRC-level approval processes that Kennedy documents as still partially intact, which issuers actually get to default. An evaluative institution cannot generate a fully credible negative judgement about an entity whose ultimate fate the evaluator’s own political principal still controls. China’s reform programme is an attempt to build genuine downgrade capacity inside a system that has not surrendered, and shows no sign of intending to surrender, the final word on which downgrades are allowed to matter.
A second, complementary factor sits alongside the continuity argument developed from NCO’s collapse, though it applies chiefly to the Chinese side of the comparison and should not be mistaken for the primary explanation. Return to the American case with the tension over China’s own reform programme in view: the SEC in the 1970s had no equivalent stake in whether any particular corporate issuer defaulted. Its interest lay in the orderly function of capital markets broadly, and Moody’s and S&P, whatever their conflicts, had decades of accumulated reputational capital, reinforced, not undermined, by Moody’s recent, visible success in stabilising the market NCO had wrecked, that made displacing them costlier than tolerating a documented but bounded inflation of roughly 20 per cent of a rating grade concentrated among high-fee issuers. NRSRO status in 1975 was not a judgement that the issuer-pay conflict was benign. The SEC’s own 2003 report conceded plainly that the issuer-fee model naturally creates the potential for conflicts of interest and ratings inflation, while concluding that the conflict was, in its words, manageable. What made it manageable, in practice, was that no American regulator’s own solvency, fiscal capacity, or political legitimacy depended on any single corporate issuer’s rating being accurate, in the way China’s does depend on the fate of its local government financing vehicles. A ratings failure in the U.S. corporate bond market of the 1970s was a cost borne by bondholders and, eventually, by taxpayers through episodes like the savings and loan crisis, not a direct threat to the state’s own balance sheet or its capacity to manage social stability.
China’s position is different in a way that goes beyond scale or recency. Local government financing vehicles, the sector regulators are now scrutinising hardest, are liabilities the state cannot treat as arm’s length corporate risk, because their debt is entangled with the fiscal capacity of the local governments that stand behind them and, by extension, with the central government’s own exposure to a systemic local-debt crisis. Property developers, the other sector under heaviest review, sit at the centre of a four-year contraction that Hale and Wang (2025) describe as already testing Beijing’s tolerance for instability. An evaluative institution that continues issuing AAA grades to LGFVs and developers that cannot service their debt conceals information the state itself needs to manage an unwinding it cannot afford to lose control of, beyond simply mispricing risk for investors who can absorb the loss. Inflation in this setting threatens the allocative function credit ratings are supposed to serve for the state’s own industrial and fiscal planning, in addition to the informational function they serve for private capital. That is a different order of stake than anything the SEC faced in 1974, and it offers one explanation, though not the only plausible one, for why an institutional response that never occurred in the United States is occurring in China now.
A third factor, the comparative cost of institutional reconstruction, deserves only brief mention alongside the first two. China’s domestic credit rating agencies entered the 2000s with comparatively little accumulated reputation to lose, a legacy of the state-engineered irrelevance Kennedy (2008) documents through the 1990s, which made building CBR from nothing in 2010 and suspending Dagong’s licence in 2018 administratively cheaper undertakings than dismantling an established American duopoly would have been. This helps explain why reconstruction was easier to attempt in Beijing than it would have been in Washington. It does not explain why either state wanted reconstruction in the first place, and the two explanations already developed, the memory of NCO’s collapse and the state’s direct stake in the outcomes being evaluated, carry considerably more of that weight.
Nothing about the proposition developed here is specific to bond markets, and it is worth pausing on that before closing. An auditor’s signature promises, at bottom, a willingness to withhold it. A financial regulator’s licence promises a willingness to revoke it. A court’s judgement carries force only if litigants believe an unfavourable ruling remains genuinely available to either side in the dispute, not merely to whichever side the institution privately expects to lose. Scientific peer review functions, to the extent it functions, because reviewers retain some willingness to reject a submission from a source whose future goodwill the journal might otherwise prefer to keep. Accreditation bodies certify because decertification remains a credible possibility and not a theoretical one. Central banks set policy credibly only for as long as markets believe a rate decision unwelcome to elected government remains available to the people making it. Each of these institutions is vulnerable to the same mechanism documented here in unusually measurable form: a party with the standing to punish a negative judgement, through withdrawn business, budget cuts, appointment power, or simple reputational retaliation, can narrow the range of judgements an evaluator is willing to make long before that narrowing appears as any formal change in the rules. Credit rating agencies are simply the case in which the evidence happens to be unusually well documented on both sides of a genuine comparison.
What remains unresolved, once these explanations have been set beside each other, is the question underneath them. Three evaluative institutions faced documented or demonstrable failures of the same underlying capacity within the space of half a century. NCO lost that capacity outright and did not survive losing it. Moody’s and S&P retained a version of it compromised only at the margin, inside a market still absorbing the memory of what a total loss had just cost it, and were embedded more deeply into the architecture of American finance as a result. China’s domestic agencies lost the capacity gradually and less visibly than NCO did, and are now being forced through the most sustained reconstruction of evaluative credibility either country in this comparison has attempted. Accuracy alone does not explain the divergence, since none of the three institutions was, in the ordinary sense, unusually inaccurate across the bulk of its ratings; most bonds and most commercial paper, in every case examined here, did not default. History suggests an evaluative institution can survive a demonstrated loss of refusal capacity, provided the loss is partial and the memory of a worse, total loss is recent enough to make the alternative look costlier. China’s crackdown suggests an institution can also be made, deliberately and at real cost, to demonstrate a capacity it had visibly stopped demonstrating, without any comparable memory of catastrophic collapse to force the state’s hand. What decides which path a given system takes, whether it is a recent history of total evaluative failure, the state’s own stake in the outcomes being evaluated, the accumulated reputation of the incumbent, or some fourth factor this comparison has not isolated, is a harder question than three cases can settle. It is the question NCO, Moody’s, and China’s domestic agencies leave behind for whoever wants to take it further.
References
Bush, C. (2022) ‘The Chinese credit rating industry: Internationalisation, challenges and reforms’, *Journal of Economics and Business*, 118, 106032.
Cash, D. (2024) ‘A Modern Credit Rating Agency: The Story of Moody’s’. Routledge.
Gjata, J. (2017) ‘The Private Roots and Public Branches of Regulation-by-Information: Understanding the Legal Incorporation of Rating in Finance and Accreditation in Healthcare (1900-1970s)’. Doctoral thesis, University of Virginia.
Hale, T. and Wang, X. (2025) ‘China’s corporate bonds win record share of top credit ratings’, Financial Times, 16 August.
Handal, K.V. (1972) ‘The Commercial Paper Market and the Securities Acts’, The University of Chicago Law Review, 39(2), pp. 362-402.
Jiang, X. and Packer, F. (2017) ‘Credit ratings of domestic and global agencies: What drives the differences in China and how are they priced?’, BIS Working Papers No. 648.
Jiang, J. (X.), Stanford, M.H. and Xie, Y. (2012) ‘Does it matter who pays for bond ratings? Historical evidence’, Journal of Financial Economics, 105(3), pp. 607-621.
Kennedy, S. (2008) ‘China’s Emerging Credit Rating Industry: the Official Foundations of Private Authority’, The China Quarterly, 193, pp. 65-83.
Lee, H. (2016) ‘Regulation of Gatekeepers after the Financial Crisis: An Example of Credit Rating Agencies in China’, Company Lawyer, 37(11), pp. 352-360.
Liu, S. and Wang, H. (2021) ‘Comprehending China’s Domestic Agency Ratings: A Perspective from Default Probability-Implied S&P Ratings’, SSRN Working Paper.
Liu, S. and Wang, H. (2022) ‘Blessing or Curse? The Effects of Credit Rating Standard Loosening in China’, SSRN Working Paper.
Lu, P., Yao, Y., Li, F. and Li, Y. (2026) ‘Investor-paid ratings and rating quality in China: Evidence from negative rating actions’, Finance Research Letters, 99, 109922.
Lu, P., Yao, Y., Li, Y. and Xiong, X. (2026) ‘Investor-pays ratings and credit risk information: Evidence from China bond rating’, International Review of Economics and Finance, 109, 105404.
Securities and Exchange Commission (1997) ‘Nationally Recognized Statistical Rating Organizations’, Federal Register, 62(249), 30 December.
Tian, W., Zhou, X., Tian, Y. and Meng, W. (2020) ‘Short-term competition and long-term convergence between domestic and global rating agencies: Evidence from China’, PLoS ONE, 15(5), e0232804.
U.S. Senate (1972) The Financial Collapse of the Penn Central Company: Staff Report of the SEC to the Special Subcommittee on Investigations. Government Printing Office.
Yang, C. (2026) ‘China targets panda bond reform, mandates global credit mapping to lure foreign capital’, SCMP (22 Jul).
Zelezetskii, I. and Rutledge, A. (2025) ‘Lost in Translation: How the Financial Times Misread Chinese Credit Ratings’, LinkedIn 18 August.
Zhou, X., Wang, H. and Zhang, L. (2024) ‘Research on the effect of multiple credit ratings from the perspective of financial regulatory systems in Chinese bond market’, PLoS ONE, 19(11), e0312533.