Yongjian Ke

Research: Contract, relationship and consent

· 7 min

My doctoral work asked a narrow question: how should risk be shared between a government and a private partner in a public-private partnership, and what happens when it is shared badly. I am still working on a version of that question, but the object has widened. A contract allocates risk between the parties who sign it. It does not reach the relationships that decide whether those parties cooperate when the contract is silent, it does not reach the wider set of firms that actually delivers the work, and it does not reach the people who live beside the project and never signed anything. Most of what I research now sits in that gap.

Risk allocation and what it cannot reach

The early work was about allocation itself. Delphi studies and surveys of Chinese PPP practice produced preferred allocations for specific risks (Ke, Wang, & Chan, 2010). The result that mattered was not the preferred list. It was the size and direction of the gap: risks were regularly assigned to the party least able to price or manage them, and the reasons were political and commercial rather than analytical (Ke, Wang, & Chan, 2013). Where only two or three bidders are realistic, allocation is not an optimisation. It is what the parties can agree given who is at the table.

Two things followed. The first is that the definition of the model is not a housekeeping matter. PPP is used for arrangements that have very little in common, which makes comparative claims unstable, and I have argued that the term is better treated as a family resemblance than as a category with necessary conditions (Ke, Cheng, Zhang, & Liu, 2024). The second is that allocation on paper and allocation in practice are different objects. That moved the work towards how projects behave rather than how they are structured.

What relationships do that contracts cannot

Relational contracting was the next step. Working with Florence Ling at the National University of Singapore, we used a common instrument across Singapore, Beijing, Hong Kong and Sydney to ask which relational practices are actually used in public projects and which of them are associated with better outcomes (Ling, Ong, Ke, Wang, & Zou, 2014; Ke, Ling, & Zou, 2015). Comparison was the point. Relational practices travel unevenly between jurisdictions, and the obstacles are institutional rather than attitudinal. Procurement rules, audit expectations and the career incentives of public officials determine how much discretion anyone has to build a relationship in the first place.

Psychological contracts took the same idea further, to the expectations each party holds about the other that are never written down and cannot be enforced (Davis, Jefferies, & Ke, 2017). This matters for a class of failure that contract analysis cannot explain, where no clause was breached and cooperation collapsed anyway.

More than two parties

The contract framing carries a second simplification, which is that there are two parties. A large project is delivered by a set of firms that assemble for the job, separate when it ends and reassemble in a different combination for the next one. A study led by Ralf Müller at BI Norwegian Business School, on which I was the co-investigator for Australia, treats those networks as the thing being governed rather than as the means of governing a project (Unterhitzenberger et al., 2023; Müller et al., 2024). It separates the governance of one network delivering one project, the governance of the several networks an owner runs at the same time, and the policies set by governments and investors that decide what kinds of network are possible at all.

Two findings connect this to the rest of the page. Whether a network ends up hierarchical or collaborative is largely settled by how it was formed, which is a procurement decision taken before the project exists. And policy made at the top does not reach the delivery end on its own; its influence ran through whether accountabilities between organisations had been clearly defined at the intermediate level.

When the asset is a service

Social infrastructure puts the problem in its sharpest form. A motorway either carries traffic to a standard or it does not. A hospital or an aged care home is harder, because what is being bought is care given to a person, and most of what makes that care good or bad is difficult to write into a schedule. A body of work with Kun Wang and Shankar Sankaran looks at hospital and aged care partnerships in China and Australia, and at the Northern Beaches Hospital inquiry in Sydney, to ask what happens when the contract reaches into a service it cannot specify. That work is set out at length in a separate brief.

The related question is how a whole care system performs on social criteria rather than financial ones, which is the subject of an assessment of the Australian aged care system (Wang & Ke, 2024).

Consent as a governance problem

Social licence entered the field from mining, where the granting community is usually local, identifiable and directly affected. Infrastructure is not like that. The people who host a motorway, a rail line or a waste-to-energy plant are often not the people who benefit from it, the asset is public, and consent is granted and withdrawn by a shifting population rather than by a defined group.

A project can satisfy every clause in its contract and still lose the consent of the people it was built among.

With colleagues in Australia and China I am developing social licence for infrastructure as a framework specific to that setting, drawing on stakeholder theory and social contract theory, and asking who grants legitimacy, who receives it, and how it is sustained across a project lifecycle rather than secured once at approval. The empirical work sits mostly around facilities that attract concentrated local opposition and disperse their benefits widely. Procedural fairness carries more weight than compensation across that literature (Jin et al., 2024), the analytical frameworks for public acceptance and social licence are not interchangeable (Xu et al., 2023), and legitimacy is better treated as something with resilience than as a binary state (Jiang et al., 2025). This stream is supported by three National Natural Science Foundation of China grants and by partnership funding at UTS.

Making these things measurable

A fair criticism of work on relationships, sustainability and legitimacy is that it describes things which cannot be measured and therefore cannot be managed. Much of my method work is an answer to that. It uses indicator sets built from an existing framework and applied to project documents, content analysis of inquiry records and media coverage, sentiment analysis of platform posts as a continuous proxy for public attitude (Mathur, Ninan, Vuorinen, Ke, & Sankaran, 2021), survey experiments, and evaluation models that accept imprecise underlying judgements.

The reason is practical rather than technical. If something cannot be stated as an indicator, it does not enter the output specification, the payment mechanism or the performance assessment, and it will not be managed.

Digital transformation and AI readiness

The most recent stream asks a related question about capability. Readiness is the more useful frame than capability, because the constraint on most project organisations is not the availability of tools but whether their data, governance and skills allow them to use what already exists (Dani, Ke, & Alkilani, 2026). Two current projects follow from this: an AI readiness framework for construction project management, and decision making with imperfect data in major projects. A parallel line of joint work with colleagues in China examines digital transformation policy in Chinese construction.

How I work

I am a qualitative and mixed methods researcher. My own studies are mostly comparative case designs, document analysis and structured expert judgement, usually across Australia and China. Much of the work is collaborative, and where quantitative modelling appears in joint papers it is normally led by co-authors with that expertise. I set this out because it defines what I can contribute to a project and what I should not be asked to lead.

What this work does not settle

Almost none of the case studies has a counterfactual. Projects are compared with the intentions in their own business cases rather than with a comparable project delivered another way, so the model and the circumstances remain hard to separate.

Cases are often selected on outcome, either because a public inquiry has generated a usable record or because practitioners nominated a project as successful. Neither describes the middle of the distribution.

The evidence base for PPPs is geographically narrow. The countries with the deepest experience produce most of the research, which limits how far any of it generalises.

Sentiment analysis measures the people who post, who are not the affected population. It gives continuity and scale, and gives up representativeness to get them.

Social licence is at risk of conceptual stretch. If every form of public reaction is described as licence, the term stops discriminating between cases, which is the main thing I am trying to avoid in developing it.

The layered account of network governance orders its levels by logic and sequence rather than by anything observed, so the direction of influence between them is an inference from cross-sectional data.

Concessions run for decades and studies do not. Most of what has been written about performance, including my own, observes an early period and infers the rest.

Papers referred to on this page