Integrated Yield and Demand Forecasting for Air Cargo in MEASA Emerging Markets: A Multi-Method Statistical Framework

Technical Summary: Integrated Yield and Demand Forecasting for Air Cargo in MEASA Emerging Markets

Problem Statement
The air cargo industry, particularly within the Middle East, Africa, and South Asia (MEASA) region, relies heavily on backward-looking operational metrics—specifically Month-over-Month (MoM) and Year-over-Year (YoY) comparisons—to manage capacity and pricing. However, these metrics are structurally inadequate for high-growth, disruption-prone markets like the South Asia–Europe corridors via the Istanbul hub. Existing forecasting models suffer from three primary deficiencies:

1. Structural Inadequacy: Standard single-variable models developed for stable North Atlantic or transpacific markets fail to capture the rapid growth, commodity diversity, and pronounced seasonality of MEASA lanes.
2. Lack of Disruption Variables: Existing literature lacks a quantifiable mechanism to integrate geopolitical disruptions (e.g., airspace closures, conflicts) into forecasting, often relying on binary dummy variables or post-hoc manual recalibration.
3. Separation of Yield and Demand: Yield and demand are typically modeled as separate problems, whereas commercial decisions require a unified view. Furthermore, a translation barrier exists between academic time-series modeling and the practical language of commercial reporting.

Methodology
The paper proposes a dual-component statistical framework operationalized within Microsoft Power BI, designed specifically for the South Asia–Europe corridors (India, Pakistan, Bangladesh, and Sri Lanka) routed through Istanbul.

• Yield Forecasting Framework:

  • Core Models: Utilizes Ordinary Least Squares (OLS), Seemingly Unrelated Regression (SUR), and ARIMAX.
  • Variables: The model incorporates GDP growth, aviation fuel price index, lane load factor, competitive capacity (available tonne-kilometers), and a novel Geopolitical Disruption Index.
  • SUR Application: SUR is employed to address correlated error terms across multiple lanes (India, Pakistan, etc.) when they share common macroeconomic and disruption shocks.
  • Non-Linearity Handling: While the regression models capture linear relationships, the framework acknowledges that yield pricing power exhibits threshold behavior (e.g., emerging sharply above 85% load factor).

• Demand Forecasting Framework:

  • Core Model: A SARIMAX (Seasonal AutoRegressive Integrated Moving Average with Exogenous variables) model.
  • Theoretical Bridge: The paper formally demonstrates that standard MoM and YoY metrics are limiting cases of the AutoRegressive (AR) and seasonal differencing components of SARIMAX, respectively.
  • Exogenous Variables: Includes GDP, fuel prices, load factor, competitive capacity, the Geopolitical Disruption Index, an E-Commerce Penetration Index, and a European Buying Season Index (critical for textile lanes).
  • Ensemble Augmentation: A Random Forest layer is trained on the residuals of the SARIMAX model. This machine learning component captures non-linear interactions (specifically load factor–yield non-linearities above 85% utilization and fuel price–demand threshold effects) that parametric models systematically underestimate.
  • Final Output: Final Forecast = SARIMAX Forecast + RF Residual Correction.

Key Contributions
The paper identifies three original contributions to the air transport management literature:

1. Quantifiable Geopolitical Disruption Index: The introduction of a scaled, continuous variable to quantify the impact of geopolitical events (specifically applied to the 2025–2026 India–Pakistan airspace closure) on both yield and demand, moving beyond binary dummy variables.
2. Integrated Yield–Demand Framework: A unified analytical structure that simultaneously models yield and demand, addressing the gap where these are traditionally treated as separate problems.
3. Mathematical Equivalence Derivation: A formal derivation proving that standard commercial MoM and YoY metrics are mathematically equivalent to specific components of the SARIMAX framework. This bridges the gap between practitioner analytics and academic modeling, facilitating deployment in operational reporting environments.

Results and Findings

• Disruption Accuracy: The application of the Geopolitical Disruption Index to the 2025–2026 India–Pakistan airspace closure scenario demonstrably improved forecast accuracy during the disruption period compared to models lacking this variable.
• Non-Linear Capture: The Random Forest augmentation layer successfully captured load factor–yield non-linearities above 85% utilization, a threshold behavior that SARIMAX alone missed.
• Commodity Mix as Yield Protection: Analysis across four South Asian sub-markets identified the commodity mix—specifically the share of pharmaceutical and perishable cargo—as the strongest mechanism for yield protection. These segments command premiums that buffer against yield compression seen in general cargo.
• Operational Deployment: The framework was successfully operationalized in Power BI, providing commercial teams with point forecasts, confidence intervals, seasonal decompositions, and scenario analysis tools (e.g., modeling fuel shocks or airspace closures).

Significance and Claims
The paper asserts that the MEASA region presents a distinctive market environment characterized by structural growth, diverse commodity mixes, and recurring geopolitical risks, rendering standard forecasting approaches insufficient. The significance of this work lies in its ability to:

• Provide a statistically rigorous yet operationally deployable tool for managing supply chain analytics in one of the world's most dynamic air cargo geographies.
• Enable proactive commercial decision-making (allotments and pricing) by shifting from backward-looking metrics to forward-looking, integrated forecasts.
• Offer a methodology for rapid recalibration of forecasts in response to geopolitical events, sanctions, or routing changes.
• Facilitate the adoption of quantitative methods in commercial cargo teams by aligning academic models with the existing language of commercial reporting (MoM/YoY).

The authors note that the framework relies on accurate input data and sufficient historical data for the Random Forest layer to function effectively. It is designed to augment, not replace, domain expertise, particularly when dealing with events outside the model's training distribution.