Joint Treatment Effects of Land Use Variables on Transportation FFCO2 Emissions

A Counterfactual Framework for Urban Decarbonization Policy

Hawkins-TECH-Lab

University of Calgary

2026-07-08

Introduction & Motivation

The Urban Transportation Decarbonization Challenge

  • The Imperative: Reducing Fossil Fuel CO₂ (FFCO₂) emissions in urban transportation is critical for meeting municipal net-zero targets.
  • The Complexity: Built form and land use variables (density, diversity, design) do not act in isolation. They exert joint treatment effects on travel behavior.
  • The Research Gap: Traditional elasticities often model land use attributes linearly or in isolation, failing to capture non-linear interactions, self-selection bias, and spatial dependencies.

Phase 1: Conceptual Framework

Beyond Isolated Elasticities

Traditional Approaches: * Separate treatment of density (e.g., population/employment density). * Ignores compounding feedback loops. * Prone to omitted variable bias from unobserved sorting.

Proposed Framework: * Treats land use configurations as a joint multi-valued treatment. * Explicitly models the structural interactions between built form features. * Acknowledges the Lifeworld-System dynamics governing household location and travel choices.

Phase 2: Econometric Methodology

Causal Inference & Counterfactuals

To evaluate the true impact of structural interventions, we define the causal effect using potential outcomes. Let \(Y_i(t)\) be the potential FFCO₂ emissions of zone \(i\) under joint land use treatment vector \(t \in \mathcal{T}\).

\[\tau(t, t') = \mathbb{E}[Y_i(t) - Y_i(t')]\]

Overcoming Confounding via Propensity Scores

We utilize Generalized Propensity Scores (GPS) to adjust for multi-valued joint treatments, ensuring covariate balance across disparate urban fabrics:

\[R = r(T, X) = f_{T|X}(T|X)\]

Core Assumption (Weak Ignorability): Conditional on observable spatial and demographic covariates \(X\), the assignment to a specific land use intensity \(T\) is orthogonal to the potential emissions outcomes.

Phase 3: Empirical Model Specification

Joint Treatment Identification

We specify a multi-equation structural system where spatial parameters are evaluated simultaneously.

\[Y_{\text{FFCO2}} = \alpha + \mathbf{\beta \cdot T_{\text{Joint}}} + \mathbf{\gamma \cdot X} + \epsilon\]

  • \(\mathbf{T_{\text{Joint}}}\): Vector representing cross-densities of employment, residential concentration, and intersection density.
  • \(\mathbf{X}\): Matrix of socio-economic control variables (household income, vehicle ownership rates).
  • Non-linearities: Captured via non-parametric extensions and machine learning-driven propensity weighting (e.g., WeightIt framework).

Phase 4: Model Estimation & Key Results

Average Treatment Effects (ATE) on Zone-Level FFCO2

Counterfactual Model Estimates (N = 1,420 Urban Zones)
Treatment Mix (\(T_{\text{Joint}}\)) Density Metric ATE (\(\Delta\) Tonnes CO₂/yr) Std. Error
Baseline (Sprawl Suburban) Low Reference —
Transit-Oriented Core High \(-42.3\) 3.12
Balanced Mixed-Use Med-High \(-28.7\) 2.45
Jobs-Rich Compact High \(-34.1\) 2.89

Key Insight: Non-Linear Thresholds

  • The Synergy Effect: Increasing density without a corresponding increase in land use diversity yields diminishing returns in FFCO₂ mitigation.
  • Thresholds: A minimum intersection density is required before transit-oriented development triggers significant vehicle-kilometers traveled (VMT) reduction.

Policy Implication:

Single-dimension zoning amendments (e.g., merely increasing allowable floor-area ratios) are structurally insufficient. Interventions must be bundled.

Phase 5: Policy Implications & Next Steps

Systems-Level Planning Interventions

  1. Integrated Spatial Targeting: Allocate infrastructure funds preferentially to zones where the joint probability of behavioral shifts is maximized.
  2. Active Transportation Infrastructure: Couple compact zoning with pedestrian overpass priorities and multi-modal network connectivity to lock in low-emission baselines.
  3. Future Scalability: Extending the framework to incorporate Bayesian Additive Regression Trees (BART) to relax parametric assumptions regarding the joint treatment surface.

Discussion & Conclusion

Strategic Takeaway

Urban form is a highly coupled system. Only by evaluating the joint treatments of density, diversity, and network design can we reliably predict—and realize—the deep decarbonization of urban transportation networks.

Thank you! Questions & Collaborative Discussion