China's shadow banking in 2020-2022: an empirical study

Shadow banking remains central to financial stability debates despite post-GFC regulatory tightening (e.g., Basel III, Dodd-Frank). Events such as the 2023 failures of Silicon Valley Bank and Credit Suisse highlight vulnerabilities characteristic of shadow banking (asset–liability mismatches, risk-oriented investment, weak controls) even within regulated banks. This paper examines China’s shadow banking during 2020–2022, a period shaped by COVID-19 and intensified NBFI regulation. China’s shadow banking stock reached RMB 47.6 trillion by end-2022, after rapid pre-2017 expansion, a 2018–2019 contraction, and subsequent stabilization. The study uses detailed balance sheet data to assess balance sheet evolution, growth dynamics, and macro-financial impacts. Research questions: (1) How did the balance sheet of China’s shadow banking change in 2020–2022? (2) What factors drove these developments? (3) Does shadow banking still significantly influence financial regulation and monetary policy? Contributions include: documenting close parallels between China’s shadow banking and recent bank failures’ business models; introducing the concept of reintermediation (a reversal of pre-2017 disintermediation) and a "Pull-Push" growth framework; and identifying empirical breakdowns in pre-COVID macro-financial relationships in China, notably a weakened monetary policy transmission.

Recent work explains shadow banking growth via liquidity and credit transformation and investors’ demand for money-like claims (Gennaioli et al., 2013; Sunderam, 2015; Moreira and Savov, 2017). Regulatory drivers and consequences for financial stability are emphasized by Farhi and Tirole (2021), Irani et al. (2021), Gebauer and Mazelis (2023), and Lyonnet and Chretien (2023), with monetary policy also relevant (Xiao, 2020). For China, some studies note beneficial aspects (e.g., improved allocation efficiency and corporate bond market development; Zhu, 2021; Chen et al., 2020) but also stress the channeling of funds to local government and real estate sectors (Allen et al., 2019; Zhu, 2021; Li, 2021). Monetary policy’s quantity-based framework and tightening episodes are linked to shadow banking expansion and reduced policy effectiveness (Chen et al., 2018; Li, 2020b; Cheng and Wang, 2022). Shadow banking’s contribution to financial stability risks via real estate and regulatory indicator “twists” is documented in Li (2020a). Li (2019a) provides a functional definition and measurement framework for China’s shadow banking.

Data: Monthly data from January 2020 to December 2022 (36 observations). Variables are compiled following Li (2019a) and official sources (NBS, PBOC, Wind). Dependent variable: sshadowyoy, the YoY growth rate of the outstanding balance of China’s shadow banking. Explanatory variables: (1) infnoeyoyytd: YoY growth rate (year-to-date) of infrastructure investment excluding electricity; (2) reinvyoyytd: YoY growth rate (year-to-date) of real estate investment; (3) loanyoy: YoY growth rate of RMB loans; (4) repo7drate: monthly average 7-day repo rate. Model: sshadowyoy_t = α + β1 infnoeyoyytd_t + β2 reinvyoyytd_t + β3 loanyoy_t + β4 repo7drate_t + ε_t. Rationale: Investment variables proxy demand-side "pull" from underserved sectors; loanyoy captures banks’ supply-side substitution between on-balance-sheet lending and shadow banking; repo7drate captures interbank liquidity and the return or cost of financing extra liquidity, with sign depending on regime. Estimation: Baseline OLS with heteroskedasticity-consistent (White) standard errors and OLS with HAC (Newey–West) standard errors (lag order p = n^(1/4)). To address autocorrelation/heteroskedasticity and potential endogeneity, 2SLS, LIML, and GMM are also employed with HAC (Newey–West) covariance and program-selected lag length per Newey and West (1994). Time-series properties: Stationarity tested using ADF, Phillips–Perron, DF-GLS, and KPSS. ADF and DF-GLS suggest four variables except repo7drate are non-stationary; PP suggests all five are non-stationary, prompting co-integration analysis. Co-integration: Lag order chosen by likelihood ratio tests and information criteria (AIC, HQIC, SIC), with Johansen trace and maximum eigenvalue tests at 5% significance. Results indicate at least three co-integration relationships among the variables during 2020–2022. Summary statistics (selected): sshadowyoy mean −1.01% (SD 3.07; min −5.43; max 5.53); infnoeyoyytd mean 3.90%; reinvyoyytd mean 4.01%; loanyoy mean 12.06%; repo7drate mean 2.17%.

• Size and balance sheet: China’s shadow banking totaled RMB 47.6 trillion by end-2022. Bond investments accounted for 36.6% of total assets, with bond holdings expanding 48% during 2020–2022. Funding relies on uninsured interbank borrowing and wealth management products, increasing contagion risk through asset–liability mismatches. - Parallels to recent bank stresses: Business models resemble those of SVB and Credit Suisse—large bond portfolios funded by unstable liabilities—implying similar interest rate and liquidity risk exposures. - Growth dynamics and reintermediation: After a 10% contraction in 2018–2019, shadow banking stabilized in 2020–2022 with only 3% cumulative expansion. Reintermediation is evident: in 2020–2022, new RMB loans reached RMB 61 trillion while new shadow banking financing was RMB 1.5 trillion (only 2.5% of new RMB loans). As a share of nominal GDP, shadow banking fell by 7 percentage points from 2019 to 2022 and by 27 points versus 2016, while loans and banks’ bond investment rose by 22 and 7 points, respectively. - Regulatory environment: Continuous financial and real estate sector regulations constrained shadow banking, leaving it in liquidity surplus. In November 2022, authorities announced sixteen measures to support real estate, easing sector pressures. - Macro-financial relationships post-2020: (1) RMB loan growth became positively correlated with shadow banking growth (co-movement), replacing the pre-2019 competitive relationship. (2) The transmission from interbank rates (7-day repo) to RMB lending weakened. (3) Transmission from money supply to ultimate targets weakened; monetary policy ceased to lead economic activity. (4) The linkage between RMB loans and CPI inflation weakened. Despite these shifts, shadow banking continued—though to a lesser extent—to distort financial regulatory indicators. - Time-series evidence: Stationarity tests indicated widespread non-stationarity; Johansen tests supported at least three co-integration relationships among growth in shadow banking, investment activity, lending growth, and interbank rates over 2020–2022.

Findings indicate that China’s shadow banking balance sheet shifted toward standardized bond investments funded by volatile, uninsured liabilities, heightening maturity and interest rate mismatch risks akin to those observed in recent Western bank failures. The observed reintermediation—shadow activities moving back to banks’ balance sheets—reflects regulatory pressure on NBFIs and the real estate crackdown, alongside subdued demand under COVID-19. This reallocation changed growth dynamics: traditional bank lending and banks’ bond portfolios expanded relative to shadow channels. Empirically, several pre-pandemic macro-financial linkages broke down: interbank rate changes transmitted less effectively to bank lending; money-supply-based signals no longer led real activity; and credit–inflation correlations weakened. Nonetheless, shadow banking continues to influence financial regulation metrics, albeit with reduced intensity. Together, these results answer the research questions by documenting (i) a bond-heavy shadow banking balance sheet, (ii) regulation- and demand-driven reintermediation captured in a Pull–Push framework, and (iii) a diminished yet present impact on monetary policy transmission and regulatory indicators.

The study documents that, during 2020–2022, China’s shadow banking stabilized in size but reconfigured toward bond investments funded by uninsured liabilities, paralleling risk profiles seen in recent Western banking crises. It introduces the reintermediation concept as a counterpart to pre-2017 disintermediation and formalizes growth drivers within a Pull–Push framework. Empirical analysis reveals weakened monetary policy transmission and altered macro-financial relationships, while shadow banking’s distortion of regulatory indicators persists to a lesser degree. These results underscore the need for continued regulatory vigilance over both NBFIs and banks’ on-balance-sheet activities that mimic shadow banking, especially given exposure to interest rate and liquidity risks.