China CNY/CNH Spread Arbitrage with GARCH (code included)
The Chinese foreign exchange (FX) market, characterized by the existence of two Renminbi currencies, namely the onshore Renminbi (CNY) and the offshore Renminbi (CNH), presents a particularly interesting example of FX market. The dynamics surrounding the USD/CNY and USD/CNH pairs also hold significant importance in the global financial context. This article aims to delve into the historical aspects of the FX market, dissect the underlying dynamics of Chinese economic fundamentals, scrutinize the CNY/CNH spread and its causative factors, apply financial modeling (such as the GARCH model) to comprehend the behavior of the spread and show practical instances of spread arbitrage.
Note: The data and code are in the notebooks at the bottom of the article.
Chinese FX regimes
Within China, the Renminbi (RMB) operates in two distinct realms: CNY and CNH. The former, CNY, denotes the Renminbi traded within mainland China, controlled with tighter regulations under the eye of the PBOC. Conversely, CNH represents the Renminbi traded outside mainland China in international financial hubs like Hong Kong and Singapore, offering a more market-driven environment. Consequently, the investors can trade and hedge their currency exposures, through the CNH markets.
The disparities between CNY and CNH exchange rates arise from varying market conditions and regulatory constraints. CNY, subject to PBOC’s control, operates within a trading band, limiting fluctuations within a 2% margin from the daily reference rate. On the other hand, CNH in offshore markets, is influenced predominantly by global demand and supply dynamics, allowing for more reliable pricing.
The People’s Bank of China (PBOC) assumes a central role in managing the Yuan’s exchange rate regime. Through mechanisms such as the daily reference rate and trading bands, the PBOC shapes the onshore CNY market. The managed-floating exchange rate system, adopting elements from both free-floating and controlled regimes, signifies China’s gradual shift toward market-oriented policies.
Market forces, albeit impacting both CNY and CNH rates, exert a different impact due to China’s partially restricted financial markets. The PBOC’s interventions have a more significant influence on the onshore CNY market compared to the offshore CNH market. However, as China progressively opens up the convergence between CNY and CNH rates is poised to diminish discrepancies over time.
China’s central bank employs various policy tools to influence Yuan pricing. Notably, interventions in the CNH market during extreme volatility, like hiking borrowing costs, showcase the PBOC’s measures to stabilize the currency. This intervention impacts both onshore and offshore rates, albeit to varying degrees, owing to some restrictions that I’ll cover later.
As China strides toward greater financial market openness and globalizing the Yuan, the trajectory of the onshore CNY market evolves, getting closer to the freely floating currency.
Brief history of USD/CNY regimes
The Renminbi (CNY), also known as the Chinese Yuan, has seen various shifts in its exchange rate regimes over the years. Initially fixed at 8.28 to the dollar, it underwent revaluation in July 2005, moving into a managed appreciation regime. However, this regime ended with the global financial crisis in April 2008, pegging the Renminbi at around 6.8. Subsequently, amid a period where the Chinese economy was growing rapidly while other parts of the world faced economic crises, the Renminbi was under a managed floating regime.
In 2013–2014, Chinese growth started slowing down while the US began signaling interest rate hikes. This led to the Renminbi being devalued in August 2015 for the first time since January 1994, sparking capital flight until the end of 2016. By 2017, capital controls helped stabilize the Renminbi, and it began to appreciate.
The US-China trade war in June 2018 brought about a new exchange rate regime. Then, induced by the COVID-19 pandemic, the Renminbi faced depreciation. However, China managed to control the situation swiftly while experiencing increased global demand for its goods. This, coupled with China’s dominance in global exports, led to an appreciation of the Renminbi due to high-yielding CNY bonds and portfolio flows.
The Renminbi’s performance started to change at the beginning of 2022. The volatility significantly increased and the spread started to move more violently. Also, the interest rate spread between the US and China started to decrease, which is a reason for the USD/CNY appreciation. Below is the USD/CNY exchange rate and the spread of interest rates computed as a difference between USA and China interest rates.
The interest rate differentials pose a challenge for corporate clients, who have to hedge their currency exposure. On the graph we can see the plotted swap points, the today’s exchange rate in time t if we were to hedge today. The trade partners selling USD/CNH face more uncertainty as e.g. Polish importers of Chinese goods have to first buy USD/PLN to sell the USD/CNY. The volatility and interest rate differentials in different currencies don’t zero out, but usually compound — especially now, when the interest rate differential and volatility are high.
USD/CNY and USD/CNH
While intuitively the CNY:CNH ratio should be 1, this is not the case. There are multiple reasons why both exchange rates differ. While this might not be apparent at first glance, there are multiple potential causes, which intensify at different times. Here are a couple most important ones.
Regulatory differences
CNY is subject to stringent regulations and oversight by both the People’s Bank of China (PBOC) and the Chinese government. Capital controls are in place to manage the flow of funds into and out of the country. While certain channels for arbitrage exist — such as transactions settled between exporters and importers in either market — these opportunities are limited due to the efficacy of these capital controls. As a result, complete alignment of exchange rates between the onshore and offshore markets is hindered due to restrictions on arbitrage
The PBOC routinely intervenes in the onshore market (CNY) to ensure stability, employing tools like daily reference rates. However, these interventions might take time to influence the offshore market (CNH), leading to discrepancies in their values and subsequently widening the spread between them.
Conversely, the CNH is traded in offshore markets like Hong Kong, Singapore, and London, where it encounters fewer regulatory constraints and is primarily influenced by global market forces.
Investor Behavior and Speculation Impact
Moreover, differences in investor behavior and speculative activities contribute to the spread between CNH and CNY rates. The accessibility of CNH to global investors and its perception as a potentially more speculative investment option can intensify market volatility, resulting in wider spreads compared to the more regulated onshore CNY market.
Liquidity Variations and Global Factors
Furthermore, differences in liquidity conditions between the onshore and offshore markets significantly impact the pricing of the CNY and CNH. While the offshore CNH market operates without direct intervention from Chinese authorities, fluctuations in liquidity levels can influence pricing and contribute to differences in bid-ask spreads between the two markets. Additionally, global economic conditions, risk sentiment, and financial market dynamics introduce further complexities, affecting the pricing differential between CNH and CNY rates in response to factors like risk aversion, market volatility, and contagion effects.
The CNH is driven by international market dynamics, including supply and demand from global investors, traders, and corporations. Factors such as geopolitical events, trade relations, and global economic conditions impact the offshore market differently compared to the onshore market. Hence, differences in supply and demand between the two markets contribute to the spread.
Even though China tries to prevent arbitrage opportunities they do exist. Capital controls seem to work relatively well given the scale of the economy, albeit they are sometimes leaky. The exporters and importers can choose to settle in either CNY or CNH depending on which is a better rate. However, this creates also a self-correcting mechanism as the spread hovers around 0 and has the characteristic of a stationary process.
As we can see on the graph, the spread volatility seems to cluster. The stationary and clusters make it a good candidate for GARCH implementation and test whether the market participants could anticipate the spread and settle either in CNY or CNH. We will perform a walk-forward analysis with a moving window.
To capitalize on the CNH/CNY spread one has to first consider what rate is more advantageous. The simple framework could be as follows:
- USD/CNH > USD/CNY: Exporters may find it advantageous to convert USD to CNH, while importers might sell RMB and buy CNY.
- USD/CNH < USD/CNY: In this case, exporters may convert USD to CNY, while importers might sell RMB and buy CNH.
- USD/CNH = USD/CNY: Participants would be indifferent, but analyzing spread changes can aid in anticipating favorable moves.
Spread analysis
The disparities between USD/CNY and USD/CNH play a pivotal role in shaping the risk landscape for businesses involved in international trade. Understanding and analyzing these disparities is crucial for companies aiming to effectively manage currency exposure, optimize transactions, and mitigate risks stemming from fluctuations in exchange rates. For multinational corporations conducting operations in both onshore and offshore markets, a strategic approach to currency risk management becomes imperative. By delving into spread dynamics, businesses can fine-tune their currency conversion strategies and hedging techniques, ultimately minimizing costs and maximizing returns in international transactions.
Furthermore, these implications extend beyond individual businesses, significantly impacting global trade, investment decisions, and overall financial market stability. Changes in policies or shifts in exchange rate regimes can substantially influence market sentiments, alter trade competitiveness, and drive changes in capital flows, thereby shaping the broader global economic landscape.
GARCH Model Application
In navigating the complexities of financial data, the GARCH (Generalized Autoregressive Conditional Heteroskedasticity) model emerges as a valuable tool, commonly employed in finance to forecast volatility in asset returns. Specifically, employing a GARCH(1,1) model enables the estimation of key parameters such as volatility persistence, short-term volatility, and long-term volatility based on historical data.
For instance, within the scope of analyzing the USD/CNY and USD/CNH spread dynamics, the GARCH(1,1) model might yield a few essential parameters such as:
- Omega: Representing the short-term volatility component.
- Alpha: Reflecting the persistence of volatility.
- Beta: Indicating the contribution of past errors to current volatility.
Here, the estimated parameters are as follows:
- omega = 7.2737e-06
- Alpha = 0.2
- Beta = 0.78
The model aids in predicting spread changes and volatility, providing insights for market participants to make informed decisions on currency conversions based on anticipated market movements.
To further comprehend the forecasted volatility and potential market movements, the conditional volatility derived from the GARCH model is often plotted. This visualization, often scaled by ±2 standard deviations, provides bands or boundaries illustrating the predicted ranges within which the spread might fluctuate based on the model’s estimation.
A critical aspect of the GARCH model analysis involves computing errors as the variance between the conditional volatility and the actual spread values. Understanding and analyzing these errors provide valuable insights into the model’s accuracy and potential areas where improvements or adjustments might be required. The histogram of test data errors is plotted below:
Walk-forward GARCH Model Application
The walk-forward GARCH model presents a dynamic approach to predict spread values by iterative computing and fitting the model on a moving window of data, thus incorporating the most recent market changes. This method allows for a continuous adaptation of the model to evolving market conditions, enhancing its ability to capture the nuances of spread dynamics.
The testing window, typically comprising the last 20% of the dataset, is utilized to assess the model’s predictive capabilities. Plotting the predicted values against the actual spread values within this testing window enables an evaluation of the model’s accuracy and effectiveness in forecasting spread movements.
Moreover, in the process of employing the walk-forward GARCH model, it’s essential to analyze the errors generated by the model. Plotting these errors provides insights into the deviations between predicted and actual spread values. The identification of extreme errors within the plot indicates potential areas where the model might encounter challenges or inaccuracies in prediction.
The walk-forward GARCH model, leveraging the most recent data available at the time of collection, can for example provide forecasts for the upcoming period, such as for the next 5 days. These predictions offer insights into the expected trends and movements in the spread, empowering market participants to make proactive decisions based on anticipated market behavior.
The insights from the walk-forward GARCH model’s predictions and error analysis hold significant implications for decision-making in spread arbitrage analysis. Businesses and investors can utilize these insights to adjust their trading strategies, refine risk management approaches, and optimize currency conversion tactics, aiming to capitalize on favorable market movements and mitigate risks associated with potential spread fluctuations.
So how to use it? GARCH predicts the conditional volatility which shows us the most probable bands of the spread. Exporters/importers can leverage these predictions to choose between settling in USD/CNY or USD/CNH. For instance, during forecasted high volatility, exporters might opt for the currency offering a favorable rate at that moment to mitigate potential losses. Speculators/traders can use these forecasts to inform their trading strategies based on expected spread movements.
One could also buy and sell the USD/CNY and USD/CNH simultaneously depending on where the spread is, counting on the convergence to the mean (zero spread). However, this strategy carries risks due to the time it might take for the spread to converge to zero. Traders employing this approach must account for transaction costs, and implement risk management strategies to navigate potential fluctuations and unforeseen events.
It is crucial to acknowledge that while the walk-forward GARCH model offers valuable predictive capabilities, it is not without limitations. Challenges such as sudden market shifts, unforeseen events, or structural changes may pose difficulties for the model in accurately forecasting spread values. Hence, continuous refinement and adaptation of the model based on real-time market data and insights become imperative for enhancing its predictive accuracy.
It is also important to acknowledge the limitations of the GARCH family models. The fact is that are very vast and neither of them seems to be superior. It might mean that they are not as useful in practice, even though very often used in academia. The underlying problem is that the financial market is the place of extreme events. Most of the time, a small portion of values make up for the majority of the kurtosis (4th moment). And because the fact that we do not know much about the tail fatness of financial data (4th moment), entails that we do not know much about its 2nd moment (variance — the thing we tried to forecast). This is the argument that questions a majority of widely used models, however, it is probably the most important one. Quantitative finance is not easy.
Conclusion
In conclusion, the dynamics of the Chinese FX market, particularly the mainland Renminbi (CNY) and offshore Renminbi (CNH) rates, present a complex landscape shaped by regulatory differences, market forces, investor behavior, and global economic factors.
The existence of two distinct realms for the Renminbi, each subject to varying degrees of regulatory oversight and influenced by different market conditions, results in persistent disparities between USD/CNY and USD/CNH exchange rates. While efforts to minimize arbitrage opportunities exist, the efficacy of capital controls and interventions by the People’s Bank of China (PBOC) contributes to the spread between the two rates.
The application of financial modeling, notably the GARCH model, offers insights into forecasting volatility and spread movements. By estimating parameters and visualizing conditional volatility, market participants gain a better understanding of potential market fluctuations, aiding decision-making in currency conversions and risk management strategies.
However, it’s crucial to acknowledge the limitations of such models, especially in the context of financial markets characterized by extreme events and uncertainties. The unpredictability of tail events challenges the accuracy of forecasting volatility, questioning the reliability of models solely based on historical data.
Nonetheless, the analysis of spread dynamics through models like GARCH provides valuable insights for businesses engaged in international trade, investors navigating currency exposures, and policymakers observing global financial market stability. It underscores the significance of continuous adaptation, refinement, and real-time market insights to enhance the predictive accuracy of models and make informed decisions in the realm of global foreign exchange markets.
China data modeling notebook: https://github.com/antek0308/Volatility_notebooks/blob/main/Medium/China/china.ipynb
Spread modeling notebook: https://github.com/antek0308/Volatility_notebooks/blob/main/Medium/China/spread_modelling.ipynb
