Valuation Model of Chinese Convertible Bonds Based on Monte Carlo Simulation

arXiv:2409.06496 · q-fin.PR, math.PR · Submitted 2024-09-10 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Quantitative Finance. The day in quantitative finance: models of markets, risk and prices.

Ruth: We're looking at the paper: "Valuation Model of Chinese Convertible Bonds Based on Monte Carlo Simulation".

Warren: Here's the gist of the paper. Valuation models for Chinese convertible bonds (CCBs) can be tackled using Monte Carlo simulation and dynamic programming to determine a price based on underlying stock variables.

Ruth: The results show that taking long positions in the 10 most undervalued CCBs and rebalancing daily significantly outperforms the benchmark double-low strategy.

Warren: Valuation Approaches Valuation models for convertible bonds can be classified into two main categories: the separation approach and the aggregate approach.

Ruth: The separation approach divides the convertible bond into a combination of simpler assets, such as a bond and an option.

Paper summary: Ruth: To recap, this paper uses Monte Carlo simulation and dynamic programming to price Chinese convertible bonds <ref:2409.06496#pg1>. They are setting up a framework where at each exercise time, they use the state variables of the underlying stock to regress the continuation value and apply standard backward induction to get the coefficients from current time all the way back to zero <ref:2409.06496#pg1>.

Warren: It’s all about discounting those expected cash flows, which they define as V t equals sup tau in gamma t E

f t (tau): <ref:2409.06496#pg3>.

Iris: They also mention that well-known numerical methods like Monte Carlo simulation and the binomial tree method can accommodate more complex bond provisions, although they often suffer from low computational efficiency <ref:2409.06496#pg2>.

Ruth: The paper says Monte Carlo simulation is particularly suitable for modeling discrete coupon and dividend payments, capturing the realistic dynamics of underlying state variables, and incorporating path dependent call features <ref:2409.06496#pg2>.

Warren: It also notes that it’s efficient in handling multiple state variables and flexible enough to accommodate expansions <ref:2409.06496#pg2>.

Iris: They mention that the CCB market began to develop gradually after the year two thousand and previous studies used methods like a binary tree method with a sample of twenty-four CCBs, but none considered the impact of downward adjustment <ref:2409.06496#pg2>.

Ruth: The core idea they are pushing is how this simulation process ultimately determines the exact price of that convertible bond <ref:2409.06496#pg1>.

Warren: And they show that they can simplify the problem to focus on maximizing exercise value from the investor’s perspective, looking at four possible states at each time, like call, put, conversion, and continuation <ref:2409.06496#pg3>.

Conclusion: Ruth: So we've looked at how this paper uses Monte Carlo simulation and dynamic programming to price these Chinese convertible bonds <ref:2409.06496#pg1>. They are using this framework to handle those complicated things, like soft call or put options and those downward adjustments, which are hard to model with simpler methods <ref:2409.06496#pg2>.

Warren: It means we can price these bonds in a way that actually includes those specific bond features instead of just treating them as simple assets <ref:2409.06496#pg1>.

Iris: But they have to remember their limitation, which is that they model the downward adjustment clause as a probabilistic event, setting that probability of it happening at eighty percent when a certain stock price variable equals one <ref:2409.06496#pg1>.

Ruth: Exactly, so while the model integrates the downward adjustment clause in a straightforward way, it's still an approximation based on that eighty percent probability assumption <ref:2409.06496#pg1>. It’s a simplification they had to make to keep the method manageable.

Warren: So if you're looking at this paper, remember the power comes from combining Monte Carlo simulation with dynamic programming and using multiple regression for better identification <ref:2409.06496#pg1>. It’s a lot of heavy machinery for pricing these things <ref:2409.06496#pg1>.

Iris: It's a complex method, definitely computationally intensive, but the validation shows it works well when compared against market prices for those ten bonds <ref:2409.06496#pg1>. It's a solid piece of work for understanding how to price these specific financial instruments <ref:2409.06496#pg1>.

Ruth: The title itself, "Valuation Model of Chinese Convertible Bonds Based on Monte Carlo Simulation," tells you immediately what the paper is about <ref:2409.06496#pg1>.

Warren: It’s essentially a blueprint for getting a better price for these specific types of bonds by using simulation techniques instead of just looking at them in isolation <ref:2409.06496#pg1>.

Iris: The authors, Yu Liu and colleagues, have developed this framework to address the complexities inherent in Chinese convertible bond markets <ref:2409.06496#pg1>.

Ruth: And they’re showing that this isn't just theoretical work; it’s giving us a tool that can be applied to real trading decisions <ref:2409.06496#pg1>.

Warren: So the real question is, how much better does this model actually perform when you put it into a live trading environment where things are messy and unpredictable <ref:2409.06496#pg1>.

Yu Liu

The Chinese University of Hong Kong, Shenzhen

q-fin.PR, math.PR

Submitted: 2024-09-10

Updated: 2026-11-04

Comments: Contact email updated to yu130dfb4@gmail.com (previous email yuliu3@link.cuhk.edu.cn is deactivated and no longer accessible)

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Importance score: 82/100

The gist: Valuation models for Chinese convertible bonds (CCBs) can be tackled using Monte Carlo simulation and dynamic programming to determine a price based on underlying stock variables.

Key concepts

Separation Approach
This method breaks down a convertible bond into simpler components, like a standard bond and an option. It often fails to capture the intricate features of CCBs, such as specific conversion rights or adjustments, making it less accurate for complex instruments.
Aggregate Approach
This treats the CCB as a 'black box.' Instead of breaking it down, the model simulates its behavior by observing cash flows generated by all its clauses. This allows for a comprehensive view of the bond's value through simulation.
Downward Adjustment Modeling
This provision is treated as a probabilistic event linked to exercising the put option. When triggered, a high probability (like 0.8) is set for this adjustment to occur, and the conversion price is revised based on recent stock prices to reflect potential losses.
Multiple Regression
This technique improves pricing accuracy by using multiple regression in the simulation framework. Empirical validation demonstrated that applying this method reduced the average Root Mean Square Error (RMSE) when comparing model prices to actual market prices for CCBs.

Terminology

Summary

Valuation models for Chinese convertible bonds (CCBs) can be tackled using Monte Carlo simulation and dynamic programming to determine a price based on underlying stock variables. p. 0 The results show that taking long positions in the 10 most undervalued CCBs and rebalancing daily significantly outperforms the benchmark double-low strategy. p. 0

Valuation Approaches

Valuation models for convertible bonds can be classified into two main categories: the separation approach and the aggregate approach. The separation approach divides the convertible bond into a combination of simpler assets, such as a bond and an option. p. 0 This method may fail to capture the complex features of CCBs, which include various clauses such as conversion rights, call/put provisions, and downward adjustments. p. 0 The aggregate approach treats a CCB as a ”black box,” allowing for the observation of cash flows generated by the CCB’s clauses through simulation. p. 0

Pricing Methodology

The paper proposes an effective pricing method for Chinese convertible bonds based on Monte Carlo simulation of stock prices. This method can efficiently manage the various complex clauses of Chinese convertible bonds, including soft call/put options and downward adjustments. p. 1 The framework involves generating M paths of the underlying stock price over T days using geometric Brownian motion. p. 1 The value of the convertible bond at time t is defined as V (t) = sup τ∈Γt E[ft (τ)]. p. 2 The optimal solution for the regression coefficients is found by minimizing a convex optimization problem where ∇L(θt) = 0. p. 5 The exercise strategy at time t is determined by comparing the values of mS(i)t, yˆ(i)t, Kt(if Ft ≥ pF) and Pt(if Yt ≥ pY) and choosing the strategy with the maximum potential profit. p. 3

Empirical Validation

Empirical studies confirm the effectiveness of “Multiple Regression” in pricing algorithms. When comparing model prices with market prices for 10 CCBs, the average RMSE decreased from 3. p. 714 to 2.96 when “Multiple Regression” is applied. p. 7 Furthermore, backtesting showed that after implementing the “Multiple Regression”, the backtesting returns of Least Squares factor significantly improve compared to the Double Low strategy. p. 7

Downward Adjustment Modeling

The downward adjustment provision is treated as a probabilistic event associated with the activation of the put option. When Yt = 1, a probability p is set to a relatively high value such as 0. p. 48 for downward adjustment occurring. p. 4 The adjusted conversion price Ct is revised to be the maximum of the average underlying stock prices over previous 20 trading days and last trading day. p. 4

Conclusion on Strategy

The paper concludes that taking long positions in undervalued CCBs and rebalancing daily significantly outperforms the benchmark double-low strategy. This approach integrates the downward adjustment clause with the put provision in a straightforward manner. p. 0 The overall pricing method is scalable and can be adapted if additional clauses are introduced. p. 1

References

Ammann et al. (2008).

Beveridge and Joshi (2011). p. 6

Brennan and Schwartz (1978). p. 1

Longstaff and Schwartz (2001). p. 1

Lvov et al. p. 1 (204), Lvov et al. p. 1 (204). Pricing convertible bonds by simulation>. p. 0

Cr´epey and Rahal (2011). p. 1

Da-wei et al. p. 1 (204), Da-wei et al. p. 1 (204). Empirical research of modified valuation model in chinese convertible bond market>. p. 0

Ingersoll Jr (1977). p. 0

Li-hong et al. p. 0 (2010).

Ma et al. (2020).

Milanov and Kounchev (2012). p. 1

--- Page 1 ---

Valuation Model of Chinese Convertible Bonds Based on Monte Carlo Simulation Yu Liu School of Science and Engineering, The Chinese University of Hong Kong, Shenzhen, People’s Republic of China ARTICLE HISTORY Compiled August 11, 2026 ABSTRACT We tackle the problem of pricing Chinese convertible bonds(CCBs) using Monte Carlo simulation and dynamic programming. At each exercise time, we use the state variables of the underlying stock to regress the continuation value, and apply standard backward induction to get the coefficients from the current time to time zero. p. 0 This process ultimately determines the CCB price. p. 0 KEYWORDS Chinese convertible bonds; Monte Carlo simulation; Dynamic programming; Pricing; Downward adjustment. p. 0

--- Page 2 ---

by discounting the expected cash flows. Well-known numerical methods used in this approach include Monte Carlo simulation Ammann et al. p. 1 (2008), the binomial tree method Milanov and Kounchev (2409. p. 1064961,Well-known numerical methods used in this approach include Monte Carlo simulation Ammann et al. p. 1 (2008), the binomial tree method Milanov and Kounchev. p. 1

--- Page 3 ---

St = St−1e (r−q−σ 2) + σ(Wt−Wt-1)

--- Page 4 ---

V (t) = sup τ∈Γt E[ft (τ)]

--- Page 9 ---

St = St−1e (r−q−σ 2) + σ(Wt-Wt-1)

--- Page 9 ---

V (t) = sup τ∈Γt E[ft (τ)]

--- Page 10 ---

V (t) = sup τ∈Γt E[ft (τ)]

--- Page 13 ---

V (i),t = max(mSˆ (i)t, yˆ (i)t, Pt

--- Page 14 ---

V0 = 1/M X M i=1 V(i) 0

--- Page 5 ---

Table 2.: Exercise strategy at each exercise time t. p. 5

--- Page 7 ---

Table 1.: Regression at time t. p. 3

--- Page 8 ---

Figure 3.: Simulation of stock price for Daqin CB. p. 2 p. 2

--- Page 9 ---

Table 3.: Some key notations of CCBs. p. 8

--- Page 10 ---

Figure 1.: Simulation of stock price for Daqin CB.

--- Page 15 ---

Figure 3.: Model price vs market price for various CCBs with ”Multiple Regression”

--- Page 16 ---

Table 7.: The pricing errors for 10 different CCBs with “Multiple Regression”. p. 9

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Improved Pricing Accuracy via Least Squares Regression

The AI system can be improved by implementing a Least Squares framework to regress continuation value at each exercise time using bases like St, St squared, Ft, Ft squared, and their cross-products with Yt. This allows the model to capture complex state dependencies more effectively than simpler models.

  1. Enhanced Strategy Selection via Multi-Regression

The system can perform multi-regression by dividing the stock price into four distinct intervals (Pi1 to Pi4) and performing separate regressions within each interval. This refines the continuation value estimation based on whether the underlying stock price is in a high, balanced, depressed, or low range.

  1. Robust Downward Adjustment Modeling

The system can treat the downward adjustment clause as a probabilistic event triggered by the put provision and incorporate it simply into the pricing framework by setting a probability p to 0.8. Furthermore, it can adjust the conversion price Ct to be at least the maximum of average underlying stock prices over previous trading days and the last trading day.

  1. Superior Backtesting Performance

The AI system can use model prices as a determining factor for backtesting strategies, such as taking long positions in the most undervalued CCBs and rebalancing daily. This strategy significantly outperforms the benchmark double-low strategy, showing improved cumulative returns (29.17% vs 3.55%).

  1. Reduced Pricing Error Metrics

By implementing Multi-Regression and including Yt as a base in the regression, the system can reduce average RMSE from 3.42 to 2.96 for several CCBs, indicating a more accurate prediction of market prices compared to using only St, St squared, Ft, and Ft squared as bases.

Sources

Related papers