Abstract
This article adopts the partial least squares regression (PLSR) method to forecast the price of the domestic (China-made) large aircraft C919. Through analysis of the pricing background and the relationship between performance indices and prices of foreign aircraft, we build a preliminary model. Comparing this with other modeling methods by calculating average relative error, we find the preliminary model fits better. Considering the difference in environment between domestic and foreign markets, we propose the concept of a discount factor. This factor is determined to be 0.6529 by analyzing domestic and foreign small aircraft of the same type. Finally, through the modified model, we obtain a forecast price of 48.95 million dollars per aircraft. The forecasting model and discount factor have significant reference value for the price forecast of domestic large aircraft.
Introduction
Predicting the market price for a new commercial aircraft like the C919 is notoriously difficult. Manufacturers must determine if they can recover massive R&D investments while remaining competitive against established giants like Boeing and Airbus. This task is complicated by data scarcity, as there are few reference aircraft, and high multicollinearity, where performance metrics like wingspan, takeoff weight, and engine thrust are highly correlated. Furthermore, models based solely on foreign aircraft do not account for China's specific economic environment, including lower labor costs and different market entry strategies.
Methodology
To address these challenges, the researchers employed Partial Least Squares Regression (PLSR). Unlike standard Ordinary Least Squares (OLS), PLSR is robust for small sample sizes where variables are highly correlated. It extracts principal components that explain maximum variation in predictors while maintaining high correlation with the target variable (price).
The study follows a two-step modeling approach:
- Preliminary Model: A linear PLSR model built using data from 26 established foreign aircraft to establish a "global standard" price based on performance.
- Discount Factor ($C$): To bridge the gap between foreign and domestic pricing, the authors calculated a ratio using the ARJ-21 (a Chinese regional jet). By comparing the ARJ-21's actual price to the model's prediction for an equivalent foreign aircraft, they derived a discount factor of 0.6529.
Datasets and Feature Analysis
The study utilizes two primary datasets sourced from industry technical data. The first includes performance indices and average prices for 26 commercial aircraft from Boeing and Airbus (e.g., 737 series, 747, 777, A320, A380).
Metrics include capacity, dimensions, fuel load, takeoff weight, range, and thrust. By calculating the Variable Importance of Projection (VIP), the authors found that all eight performance metrics had VIP values near 1.0. This suggests that no single feature dominates aircraft pricing; rather, the "total package" of performance determines the value.
Key Results
Based strictly on foreign performance-to-price ratios, the preliminary forecast valued the C919 at $74.97 million**. However, after applying the 0.6529 discount factor to account for the "China Discount," the predicted price dropped to **$48.95 million.
This result aligns closely with statements from China’s Ministry of Commerce, which suggested the C919 would be priced under $50 million. This external validation suggests the model's discount factor approach is grounded in market reality, reflecting lower domestic production costs and the need for competitive pricing in a market dominated by incumbents.
Conclusion and Limitations
The PLSR model proved superior to log-linear and standard least squares methods, achieving the lowest average relative error (3.84%). While the model provides a robust framework, it has limitations: it does not explicitly account for inflation over time and assumes the discount factor derived from regional jets applies linearly to large-scale aircraft. Future work could incorporate time-value adjustments and more subjective "equalization-engineering" factors to further refine the forecast.
Summary updated