📈 economics

Constrained Public Portfolio Allocation Decisions Across Transport and Non-Transport Sectors

This paper utilizes a latent class multiple discrete-continuous extreme value (LC-MDCEV) model to analyze how individuals allocate constrained public budgets across competing transport and non-transport sectors in New South Wales, revealing that distinct behavioral segments with varying levels of expenditure diversification drive funding priorities more significantly than moderate changes in project benefit framing.

Original authors: John M. Rose, Andrea Pellegrini, Matthew J. Beck

Published 2026-07-04

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: John M. Rose, Andrea Pellegrini, Matthew J. Beck

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are the mayor of a town with a very specific, unchangeable rule: you have exactly $100 to spend, and you must decide how to split it among ten different neighbors. One neighbor needs a new park, another needs a hospital wing, a third needs better roads, and a fourth wants to switch all the town buses to electric power.

If you give $50 to the hospital, you only have $50 left for everyone else. This is the core problem the researchers in this paper are trying to solve: How do regular people actually decide to split a limited public budget when everything they want costs money?

Here is a simple breakdown of what the researchers did, what they found, and why it matters, using everyday analogies.

The Experiment: The "Public Piggy Bank" Game

The researchers from the University of Sydney didn't just ask people, "Do you like roads?" (which is easy to say yes to). Instead, they played a game with over 1,200 people in New South Wales, Australia.

They gave each person a hypothetical public budget (ranging from $5 million to $50 million) and a menu of projects. The menu included:

Non-Transport: Healthcare, schools, police, parks, arts.
Transport: New roads, public trains/buses, and electrification (switching cars and trucks to electric).

The Catch: You couldn't just pick your favorite. You had to write down exactly how much money went to each project. If you gave $10 million to schools, that money was gone and couldn't go to roads. This forced people to make real trade-offs, just like a government does.

The Discovery: Two Types of Spenders

The researchers used a sophisticated computer model (a "Latent Class MDCEV") to look at the answers. They found that people weren't all thinking the same way. They fell into two distinct "personality types" regarding how they spent the money:

The "Concentrated Spender" (Class 1):
- The Analogy: Imagine a person who says, "I'm going to put almost all my money into the things I care about most, and ignore the rest."
- Behavior: This group picked a few favorites (like healthcare, education, and police) and dumped a huge chunk of the budget into them. They were willing to give zero money to the things they didn't prioritize. They were decisive and focused.
The "Diffuse Spender" (Class 2):
- The Analogy: Imagine a person who says, "I don't want to leave anyone out. I'll give a little bit to everyone so everyone gets something."
- Behavior: This group spread the money out thinly across almost all categories. They were less likely to give a project a massive boost, but they were also less likely to give a project zero dollars. They preferred a balanced, "safety-first" portfolio.

The Big Surprises

The study revealed some interesting truths about what people actually value when they have to pay the price:

The "Electric" Hurdle: Even when the researchers told the participants that electric vehicles and electric buses were great for the environment, people were still hesitant to fund them. Conventional roads and regular public transport got much more money than "green" electrification projects. It seems that when people have to choose between a new road and a new electric bus, they often pick the road, even if they care about the environment.
The "Local" Effect: People generally liked projects in their own backyard more than projects far away. If a project was in their specific region, they were more likely to fund it.
The "Core Services" Rule: No matter which group you were in, Healthcare, Education, and Police were almost always the winners. These "core" services got the most money, while things like "Arts and Museums" or "Sports Facilities" often got the short end of the stick.

Why the "Two Types" Matter

The researchers found that if you just look at the "average" person (ignoring the two types), you get a blurry, inaccurate picture. It's like saying the average temperature of a room is 70°F, but in reality, one corner is freezing and the other is boiling.

The Concentrated Spenders showed that if you want to build a specific type of infrastructure, you need to know that a large chunk of the population will either fully support it or ignore it completely.
The Diffuse Spenders showed that another chunk of the population wants to spread the wealth evenly, even if it means no single project gets fully funded.

The Takeaway

This paper teaches us that asking people "Do you support green transport?" is too simple. When you force them to look at a limited budget and choose between a new hospital and an electric bus, the answer changes.

The study concludes that public support isn't a single, solid block. It's a mix of different spending styles. Some people want to go "all in" on a few things, while others want to "spread it thin." For governments, this means that a "one-size-fits-all" plan for spending money might fail because it doesn't account for these different ways people think about value and trade-offs.

In short: When the money is real and limited, people's priorities get very specific, and they split into two very different camps on how to spend it.

Technical Summary: Constrained Public Portfolio Allocation Decisions Across Transport and Non-Transport Sectors

Problem Statement
Public investment decisions are inherently characterized by scarcity, requiring governments to allocate finite fiscal resources across competing sectors such as transportation, healthcare, education, and utilities. While existing transport research extensively utilizes stated preference methods like Discrete Choice Experiments (DCEs) to evaluate infrastructure preferences, these approaches often rely on single-choice frameworks where respondents select one preferred alternative from a mutually exclusive set. Such methods fail to capture the complex portfolio trade-offs inherent in real-world public budgeting, where citizens must simultaneously support multiple projects, exclude others, and navigate explicit opportunity costs. Furthermore, existing literature often isolates transport decisions from broader cross-sectoral expenditure contexts, limiting the understanding of how transport investments compete with non-transport priorities under constrained budgets.

Methodology
This study employs a repeated constrained public budget allocation experiment administered to residents of New South Wales (NSW), Australia. Unlike conventional single-choice tasks, respondents were required to distribute a fixed hypothetical public budget across a portfolio of competing transport and non-transport expenditure categories (e.g., healthcare, education, roads, public transport, electrification initiatives). The experimental design featured multiple tasks with varying budget levels and project compositions, generating repeated compositional data characterized by interdependent allocation shares, non-negativity constraints, and frequent corner solutions (zero allocations).

To analyze this data, the authors developed a Latent Class Multiple Discrete-Continuous Extreme Value (LC-MDCEV) model. This framework integrates three critical features:

Multiple Discrete-Continuous Structure: It accommodates simultaneous positive allocations to multiple alternatives and zero allocations to others within a single task.
Availability Constraints: The model explicitly accounts for the fact that not all project types are available in every task, treating unshown alternatives as structurally unavailable rather than rejected.
Latent Class Heterogeneity: The model assumes respondents belong to unobserved behavioral segments (classes) that differ in their underlying utility structures and satiation parameters.

The estimation utilized a custom MATLAB code to maximize the likelihood function, incorporating respondent-level socio-demographic characteristics into the class assignment model. Following estimation, a forecasting framework was applied to a synthetic population representing NSW households to simulate allocation behaviors under six alternative policy scenarios, including restricted availability sets (transport-only vs. non-transport-only) and varying benefit framings.

Key Results
The analysis yielded several significant findings regarding public expenditure prioritization:

Model Performance: The LC-MDCEV specification substantially outperformed the aggregate MDCEV model, with the log-likelihood improving from -25,789.10 to -19,566.68 and the $\rho^2$ increasing from 0.4447 to 0.5803. This indicates that unobserved heterogeneity in allocation behavior is substantial and cannot be captured by homogeneous aggregate models.
Behavioral Segmentation: The model identified two distinct latent classes:
- Class 1 (Concentrated Allocators): Comprising approximately 29% of the sample, this group exhibits weaker satiation effects, leading to concentrated expenditure portfolios where large portions of the budget are allocated to a smaller number of preferred categories (e.g., healthcare, education, policing).
- Class 2 (Diffuse Allocators): Comprising approximately 71% of the sample, this group displays strong satiation effects, distributing expenditures more diffusely across available alternatives and exhibiting higher zero-allocation shares for less preferred categories.
Transport Investment Preferences: Support for transport expenditure is highly conditional on the project type and the broader expenditure environment. Conventional infrastructure (roads and public transport) receives strong support across both classes. In contrast, transport electrification initiatives (private vehicle, freight, and public transport electrification) receive significantly weaker support, particularly from Class 2. Even when environmental benefits for electrification were substantially increased in forecasting scenarios, support remained moderate compared to conventional infrastructure.
Regional and Environmental Sensitivity: While regional differences in allocation priorities were systematic (e.g., Greater Sydney households prioritizing healthcare and electrified public transport more than regional households), these spatial variations were less influential than the latent behavioral segmentation. Furthermore, the composition of the feasible expenditure set (e.g., removing non-transport options) materially altered allocation intensity and participation, demonstrating that public support is relative to the available opportunity set.

Significance and Claims
The paper claims to make substantive contributions to both transport economics and participatory budgeting literature by demonstrating the value of combining repeated constrained portfolio allocation experiments with latent class MDCEV models.

Methodological Contribution: The study addresses a gap in transport research by applying a rigorous behavioral modeling framework to public expenditure decisions that span both transport and non-transport sectors. It successfully captures complex behavioral features—such as corner solutions, repeated decisions, and allocation intensity—that are difficult to recover within conventional single-choice stated preference frameworks.
Policy Implications: The findings suggest that public support for transport decarbonization is more segmented and conditional than aggregate opinion measures imply. The authors argue that uniform statewide expenditure strategies may fail to reflect the diversity of underlying public allocation preferences. Consequently, policymakers should recognize that citizens differ not only in the projects they support but also in their preference for concentrated versus diversified expenditure portfolios.
Forecasting Utility: The developed forecasting framework allows for the evaluation of how allocation preferences vary across population profiles and policy scenarios, providing a tool to assess the impact of changing project availability and benefit framing on public expenditure priorities.

The authors maintain a modest tone regarding their claims, acknowledging that the study relies on stated preference data from a hypothetical environment and that the findings are specific to the NSW context. They note that while the model captures significant heterogeneity, the two-class structure is a parsimonious representation of potentially more complex behavioral segmentation.