Yongjian Ke

Public-private partnerships and the paradox of where they go

· 8 min

A public-private partnership is meant to send private money and private expertise to places that are short of both. Six papers written over the past decade, using four quite different methods, point the other way. PPPs collect where capital, competence and prior experience already sit, and the appetite for them is strongest where all three are thinnest. Most of this work was done with Zhe Cheng, first at the Chinese Academy of Sciences and later at Xi'an University of Architecture and Technology, with two further papers led by Liang Ma at Shanghai University and by Menglei Jiang and Yong Liu at Zhejiang Sci-Tech University. The funding came mainly from the National Natural Science Foundation of China.

Counting them first

China had no official PPP project database until 2015, and the first one, published by the National Development and Reform Commission, listed projects in the market and in prospect but none that had finished. The only publicly available alternative was the World Bank's Private Participation in Infrastructure database, which excludes projects invested in by state-owned enterprises. In China those are the main players, so the exclusion removes most of the market.

The first paper therefore built a database by hand (Cheng, Ke, Lin, Yang, & Cai, 2016). Search across provinces and PPP terms, manual cross-validation against the project records published by large investors, then classification by financial closure year, location, sector, investor and contract type. Build-transfer projects were excluded on the grounds that they added to local government debt without improving service. The result was 1,221 projects with contracts awarded by the end of 2013, worth a little over 1,000 billion RMB.

Set that against China's infrastructure investment demand across the same period, which was above 70 trillion RMB. Whatever else PPPs were doing in China, they were not closing the funding gap.

The composition was narrow. Build-operate-transfer accounted for 82.5 per cent of contracts. Water supply and sewage treatment took 54.6 per cent of projects, followed by municipal facilities and transport. State-owned enterprises held 43.7 per cent and private firms 42.7 per cent, while foreign investors fell from the majority position they held before 2000 to 3 projects out of the 174 awarded in 2013. Concession periods were usually set at 30 years, the regulatory maximum, even though the research literature offers several methods for calculating an appropriate one. The distance between available method and actual practice recurs across the whole set.

The paradox of application

Eastern China took 48.8 per cent of the projects and the northeast took 4.9 per cent. By province the range ran from 2 in Tibet to 122 in Fujian, with a median of 31. Project numbers tracked fiscal revenue, which is to say local governments chose PPPs more often in good revenue years rather than lean ones. The paper called this the paradox of application and read it as a Matthew effect. The regions with the strongest case for an alternative financing route were the least able to attract one.

The pattern is not a fixed law, and the later data show it loosening. In the provincial counts to 2016 used in the comparative paper, Xinjiang held 168 projects, fourth in the country, while ranking 22nd in GDP, and Sichuan sat just behind it. The policy push from 2014 moved projects inland. What did not change is the association between local fiscal and administrative strength and PPP activity, which the later papers pick up in a more careful form.

The same shape in three countries

The second paper compares the UK, India and China using their official project databases rather than the perception surveys that most comparative PPP research relies on (Cheng, Ke, Yang, Cai, & Wang, 2020). The three were selected for comparable market size, a spread from developed to developing, and the existence of usable data. The United States was excluded for want of a national database, which is itself a result worth noting.

Timing differs. UK activity peaked between 1998 and 2004, India between 2007 and 2011, and China later and in repeated booms. Spatial imbalance does not differ. London recorded 244 projects and Wales 72. Maharashtra recorded 215 and Puducherry 1. Shandong recorded 333 and Tibet 2. In all three countries the projects gather in the wealthier regions.

Almost everything else diverges. The private finance initiative accounts for 91 per cent of UK projects, while build-operate-transfer accounts for 47 per cent of Indian projects and about 77 per cent of Chinese ones. UK PPPs are concentrated in public services, with education and healthcare at 55.2 per cent of all projects. India's are 59 per cent transport. China's are mainly municipal facilities and transport. The paper's reading is that the concept diffuses while the practice localises: countries copy the idea and then rebuild it around their own institutions.

The UK exported the model and then, in November 2018, withdrew PF2, its own successor to the private finance initiative, after six uses in six years. I would treat the paper's closing proposal of a global PPP knowledge community as an aspiration rather than a finding. Nothing in the data tests whether shared guidance changes national practice, and the PF2 episode points the other way.

What are we counting?

Every count above rests on a boundary decision that the field has never settled. The fourth paper takes that on directly (Ke, Cheng, Zhang, & Liu, 2024). It borrows Ludwig Wittgenstein's family-resemblance idea, already used in construction research to define partnering, alliancing and relational contracting, on the argument that a concept this varied cannot be pinned down by necessary and sufficient conditions. Instead it is described by a set of overlapping features, some of which most cases share.

The method is a content analysis of the 50 most cited journal articles with public-private partnership in the title, coded against 26 elements drawn from the literature. Elements mentioned by at least 15 of the 50 papers form the core. Six qualify: appropriate risk allocation and sharing (34 mentions), clarity of roles and responsibilities (27), injection of expertise and resources (22), cooperation and teamwork (19), long-term contract (17), and a bundle of services (16).

What sits outside the core is the more interesting result. Value for money is mentioned by 12 of the 50 papers, so it is a secondary feature rather than a defining one, even though it is the argument most often used in public to justify a PPP. Government affordability is not mentioned at all, and neither is government support and guarantees. Affordability is the issue that eventually broke the private finance initiative in the UK, and it plays no part in how the field describes what a PPP is.

The model is then applied to two American cases. The iROX road expansion in Florida, procured as design-build-finance, bundled nine smaller projects into one contract, ran for three years of design and construction, and left operation and maintenance out. It fails on expertise, cooperation, contract length and bundling, so it is not a PPP on this test, although Florida counts it as one. The Hudson-Bergen light rail contract, covering design, construction, operation and maintenance for 15 years, satisfies almost everything, and the one thing that cannot be confirmed from public documents is whether risk was materially transferred.

That is a usable test, and there is a practical need for it. China's PPP database has been periodically cleared of projects that turned out to be financing devices wearing PPP labels. The obvious caution, which the authors state, is that counting mentions in 50 articles measures what researchers write about rather than what matters on a project.

The sectors where the case is hardest

Social infrastructure is where PPPs are most difficult to justify: low profitability, complex operations, and often no natural revenue stream. The review paper asks whether they are still an answer there, working from 234 journal articles published between 1990 and 2021 (Ma, Hu, Zhu, & Ke, 2023).

The literature is lopsided. Of the publications that name a country, 72 study the UK. Australia follows with 21, then China and Italy with 13 each and the United States with 11. Six research themes emerge, led by financing and economic viability with 44 papers, then governance and regulation, performance management, contract and relationship management, risk management, and the factors that help or hinder.

Practice is lopsided in a different direction. Social infrastructure PPPs were the majority of Australian PPPs from 2000 to 2013, and almost all of them are funded by government payments made each quarter rather than by user charges. In China the same category is marginal: education, healthcare and sports account for 5.0, 2.4 and 1.2 per cent of PPP projects respectively.

The review's stated answer is that PPPs retain real potential for social infrastructure. The evidence underneath it is mixed, and the paper says so. Comparisons of PPP and conventionally procured schools and hospitals come out in different directions depending on the country and the measure, and the New Karolinska Solna hospital in Sweden became a public scandal over cost and operational failure. More importantly, the question cannot be closed, because the financial information needed to settle it is not disclosed. The review's own conclusion is that PPPs may be a workable option for delivering infrastructure that is urgently needed, and a much less convincing one across a full concession, given the pattern of cost overruns and financing costs. That is more guarded than the title suggests, and it is the honest position.

Which cities are ready

The fifth paper turns the question around. Instead of asking whether a project suits the PPP model, it asks whether a city does (Cheng, Zhu, Wang, & Ke, 2024). Eleven indicators, grouped under development needs, fiscal capacity, government credit, regulatory capacity, government efficiency and urban development characteristics, are weighted by the entropy method and ranked using TOPSIS across all 287 prefecture-level cities in China.

Two hundred and nineteen of the 287, just over 76 per cent, fall into the low suitability or unsuitable bands. Eighteen cities are highly suitable. Beijing, Shanghai, Shenzhen and Guangzhou lead, as expected, but the list also includes Heze, Jining and Luoyang, which sit in the middle of the national economic ranking and first, second and fifth for demonstration projects. Experience substitutes for wealth, at least partly.

The weights deserve attention before the rankings are used for anything. GDP carries 23.2 per cent. Infrastructure investment carries 15.0 per cent. The two indicators counting how many PPP projects and demonstration projects a city already has carry 27.5 per cent between them. Roughly two thirds of a city's suitability score therefore comes from being large, being wealthy, or having done PPPs before. Entropy weighting derives weights from the dispersion in the data rather than from a judgement about what matters, so this is a property of the dataset rather than a choice. It does mean a city is judged ready for PPPs largely because it is already running them, which makes the model a good description of where the market has settled and a weaker guide to where it could go.

Who wants to use them

The last paper asks the officials (Jiang, Cui, Xia, Skitmore, Ke, & Liu, 2025). It surveys 353 civil servants with PPP experience across the eastern, central, western and northeast regions, and models their responses using the technology acceptance model extended with the advocacy coalition framework. All twelve hypothesised paths were supported.

The chain runs from complexity to policy, then from policy to relative advantage, belief in the benefits and facilitating conditions, then to perceived usefulness and perceived ease of use, then to attitude and intention. Policy is the strongest lever in the structure, with paths to relative advantage (0.738), belief in the benefits (0.672) and facilitating conditions (0.650) that are exceeded only by the link from attitude to intention (0.732). Whatever else shapes local uptake, policy moves most of it.

The item scores matter more than the paths. Every perceived ease of use item sits at about 3.2 on a five-point scale, the lowest scoring construct in the study, while usefulness items run from 3.44 to 3.90 and complexity items from 3.58 to 3.90. Officials are persuaded that PPPs are useful. They are not persuaded that they are workable.

Then the regional comparison. Intention to use is significantly higher in the central, western and northeast regions than in the east. Set that beside the suitability rankings, where the eastern cities hold most of the high scores, and the two results cross.

Willingness is not the constraint. Capability is, and it sits in the wrong cities.

One result runs against the grain and should be reported as such. Perceived complexity was rated significantly higher in the east and centre than in the west and northeast, which is not what a capability account would predict. The paper attributes this to the eastern and central markets attracting more private interest and therefore requiring more information before a decision, and it attributes the higher ease of use scores in the northeast to the revitalisation program and the experience it built. Both explanations were arrived at after the results and neither was tested.

What follows for practice

Four things follow, and none of them is an argument for or against PPPs in general.

Build capability before pipeline. Where suitability is low, the answer is not more projects but more capacity: people who can specify outputs, monitor performance and hold a counterparty to a 30-year contract. The survey says the same thing in different language. Usefulness is already accepted and ease of use is not, so ease of use is what policy has to work on.

Test the project, not the label. The six core elements give a serviceable check. A project that bundles nothing, transfers no material risk and runs for three years is a procurement method with a financing arrangement attached, whatever it is called, and treating it as a partnership will not make it behave like one.

Publish the numbers. The same obstacle blocks the financing question and the performance question: project financial information is not public. Disclosure would not by itself show whether PPPs deliver value, but nothing else will.

Treat PPPs as a complement. On the suitability evidence, something under a quarter of Chinese cities are in a position to run them well. That is an argument for a controlled volume in the right places rather than a delivery default, which is where the papers collectively land.

What the papers do not settle

None of the six produces outcome data. They count projects, define the concept, map a literature, score readiness and measure perceptions. Whether these PPPs delivered a road or a hospital better or worse than a conventional contract would have is not tested anywhere in the set. The review reports other people's answers to that question, and those answers disagree.

The databases are not strictly comparable. UK figures run to 2013 and Indian and Chinese figures to 2016, compiled under different national definitions and different disclosure rules. Cross-country counts carry more uncertainty than the comparison can quantify. The 2016 Chinese database was assembled from public sources by hand with inclusion rules chosen by the authors, and a different set of rules would produce a different count.

The suitability model is a single cross-section to 2021, with the circularity described above, and it needs a volume of data that most countries do not publish. The survey pooled responses collected between March and October 2020 with a second round gathered in December 2024, recruited by snowball sampling through the research team's networks. Four years is a long time in Chinese PPP policy, and the two waves are not separated in the analysis.

Finally, the set is anchored in China. The comparative paper reaches to the UK and India and the review is global, but the primary data are Chinese, and the firmer conclusions should be read as claims about China first and as hypotheses everywhere else.

Journal papers discussed on this page