2.1 Definition of an SOE ‘Ideal-Type’

A positive or empirical definition of an SOE would require collecting information on different types of economic organizations (Thynne Reference Thynne2021) and eventually propose a classification that describes such observations. The OECD (2015a, 2024), in its Guidelines on SOE Corporate Governance, uses this definition:

This definition may be empirically helpful as a generic description, but it is somewhat of a truism: an SOE is simply an enterprise owned by the state. In this Element, for the purposes of normative theory, we prefer the definition provided by Bernier et al. (Reference Bernier, Florio, Bance, Bernier, Florio and Bance2020a, p. 2), in The Routledge Handbook of State-Owned Enterprises: an SOE is an organization that (a) directly produces public services or goods either through liberalized market arrangements or under a franchised monopoly, (b) is ultimately owned or de facto controlled by public sector entities, (c) has a public mission, (d) whose ownership in principle can be shifted to the private sector, and (e) has budgetary autonomy and managerial discretion.

We focus now on the normative dimensions implied by this definition, and their relations with earlier economic literature. Firstly, however, let us consider the meaning of ‘enterprise’. In English, there are two possible meanings: ‘an organization, especially a business, or a difficult and important plan, especially one that will earn money’. Examples given in the Cambridge Dictionary include a state-owned enterprise, a commercial enterprise, the spirit of competition and free enterprise, a government-sponsored enterprise, and a charitable enterprise. Some of these examples are quite distinct from something ‘that will earn money’. There is a dual meaning: on one side, there is an organization, while on the other, a plan. The Oxford Dictionary provides this definition of ‘enterprise’: ‘undertaking, esp. bold or difficult one (FREE enterprise, PRIVATE enterprise); business firm; courage; readiness, to engage in enterprises’.Footnote ¹

Interestingly, this dualism has a common root in the Latin imprehendere, where ‘im’ stands for ‘in’ and ‘prehendere’ means ‘to take,” together meaning ‘to take inside’.Footnote ²

Economic theory, at its highest level of abstraction, ignores this semantic duality between an organization with a purpose and the effort to implement a plan or project. As is well-known, but worth reiterating, in the Arrow‒Debreu general equilibrium model of the economy, a firm is simply an atomistic unit acquiring certain inputs and transforming them into outputs. This abstract firm cannot be anything other than ‘private’, as there is no role for government in an Arrow‒Debreu economy (with lump-sum ‘taxation’ in some versions of the second theorem not really implying a government, being simply a mathematical adjustment of initial conditions). Thus, whilst a given firm in such an economy can be owned by a single individual or jointly by all individuals, this would be entirely immaterial: the solutions of the systems of equations in terms of quantities and prices would remain unaffected. This firm has no mission attached to ownership, except profit maximization. The algorithm for cost minimization would lead to solutions in terms of quantity and price that are indistinguishable for a company owned by an individual or jointly by all of them. Thus, at this level of abstraction, a firm is neither a proper organization nor an enterprise (with a semantic reminder that managerial effort, beyond its use of inputs, is needed to design and implement a project). See Box 1 for a clarification of this point.

Let us now consider SOEs from the perspective of this preliminary discussion. From a positive standpoint, we observe a large variability of forms of SOEs, ranging from small municipal firms in charge of water distribution in a small jurisdiction, to giant multinational corporations extracting and processing oil globally. It is certainly possible to build a taxonomy of existing SOEs, but what differentiates an SOE from any private enterprise given their wide range of forms?

We shall argue that the answer to this question lies in a normative model, in other words, a model explaining why anything different from private enterprises should exist in the first place. To see this, the core underlying concept is: ownership by whom to do what.

There is a wide literature on the ownership rights of firms (e.g., Hansmann Reference Hansmann2000), particularly in the law and economics tradition, hence we will not discuss in detail the specific content of these rights, including the right for the owners to be the residual claimants of profits, to appoint and fire managers, or to shut down the firm if needed, and so on. Thus, let us say that a private enterprise is an organization where the ownership rights can be effectively claimed, directly or by delegation, by individuals, and is characterized by its own objectives, usually assumed to be profit maximization.

Conversely, at this abstract level, an SOE is defined as an organization where all (or a controlling part) of the related ownership rights can be effectively claimed by a government (or a public sector) entity on behalf of all citizens, to do something that a private organization would not do or would do differently. ‘Effectively’ here means that the legal provisions on ownership rights should not be blocked or so diluted as to become just a misnomer for de facto something else, such as when an individual is not free or a government agent is captured by an individual interest (e.g., an oil and gas company formally ‘public’ but actually controlled by some oligarchs). This is our starting point for a normative model of public enterprises, inspired by the aforementioned definition by Bernier et al. (Reference Bernier, Florio, Bance, Bernier, Florio and Bance2020a), further developed and qualified as follows:

1. 1. It is not the production of a tangible or intangible good per se that counts in the ideal-type of SOE, but rather the fact that whatever good is produced should in principle provide a ‘service’ to the public. Thus, the focus should be on identifying this service, understanding its rationale, and evaluating productive performance. Hence, from a normative perspective, the issue is exploring why a government should want to own or co-own an entity providing such a service. Moreover, while earlier economic literature assumed that the typical market environment for an SOE is a monopoly, the aforementioned definition extends the perspective to oligopoly and other market forms. As we shall discuss later, this dramatically changes the way a normative theory of SOEs can be designed, for example concerning pricing rules, which were a typical object of the traditional view of economic textbooks such as Atkinson and Stiglitz (Reference Atkinson and Stiglitz1980).

2. 2. Ownership without control would render the SOE definition meaningless for normative analysis. From a positive perspective, ownership rights, and actual control may not always be fully aligned, but we can assume some extent of alignment. However, control may vary in meaning and intensity. From a normative perspective, the issue is defining the optimal or desirable degree of control, which relates to point (5) below.

3. 3. Is an organization without a public mission (Sorrentino Reference Sorrentino, Bernier, Florio and Bance2020) considered an SOE? Empirically, according to the definition provided by the OECD (2015a), the answer is yes. However, from our normative perspective, it is not. The default rule for an economic organization without a public mission would be to revert to cost minimization, and equivalently to profit maximization. This point has wide implications for the concept and role of a public mission as defined and attributed by a government to an economic organization.

4. 4. The explicit possibility of privatization as part of the definition excludes from the SOE ideal-type those specific government activities that cannot be meaningfully provided by private companies, such as law-making and most judicial and cabinet core functions. Conversely, prisons, police services, hospitals, space exploration, and certain defence services, just to provide some concrete examples, have been partly privatized in some countries.

5. 5. Any organization without its own budget is formally part of another organization. The same applies if an organization is so tightly controlled by another entity that its managers have no practical budget autonomy.

With this ideal-type in mind, we shall now discuss some relevant theoretical issues for SOEs which meet the aforementioned criteria.

2.2 Looking Back: Some Prescriptions From Earlier Public Economics Literature

Atkinson and Stiglitz (Reference Atkinson and Stiglitz1980), unlike more recent textbooks (e.g., Hindriks and Myles Reference Hindriks and Myles2013), allocated two chapters to public sector production. Their ‘Lecture 15’ discusses pricing policies by public enterprises (which are not precisely defined in the text) and justifications for deviations from the benchmark case, such as marginal cost pricing in the tradition of Hotelling. The argument is based on the analogy between optimal taxation and public enterprise pricing rules. This involves a two-step process:

This two-step decision process is analogous to the Ramsey pricing mechanism but with notable differences. In the Ramsey setting, the constraint (i.e., the amount of tax revenue to be collected) is exogenous, while Atkinson and Stiglitz consider endogenous constraints. For instance, a government setting the required rate of return for one nationalized industry must simultaneously consider the requirements of another industry.

After presenting results on optimal public pricing under constraints, the authors discuss input choice by public enterprises and production efficiency, referencing Boiteux (Reference Boiteux1956) and the seminal work of Diamond and Mirrlees (Reference Diamond and Mirrlees1971a, Reference Diamond and Mirrlees1971b). This theory concludes that in an economy without pure profits, and for an individualistic social welfare function where at least one good increases utility for all individuals, production efficiency is desirable. This implies transfer prices within the public sector should be marginal costs, but this is not necessarily true for sales to the private sector.

The discussion extends to rules for the shadow price of capital (i.e., the social discount rate) and the use of international prices for open economies as shadow prices by the private sector, viewing these as issues arising in second-best economies. In ‘Lecture 16’, the authors raise several pertinent questions:

Current mainstream views on SOEs still address these questions but often overlook the theoretical assumptions framing them. For example, Putniņš (Reference Putniņš, Bernier, Florio and Bance2020) summarizes earlier literature suggesting a screening approach using five criteria:

1. 1. Is there a substantial market failure (monopolies or lack of competition, externalities, public goods)? If the answer is ‘no’, production should be left to the free market. If the answer is ‘yes’ there is a potential role for an SOE, but conditional on the answer to other questions.

2. 2. Can the market failure be resolved by regulation or targeted taxes/subsidies? If the answer is ‘yes’, there is no need to establish an SOE.

3. 3. If regulation or tax policies are not feasible, is it possible to specify quality and quantity of the good to be provided through an appropriate contractual arrangement? If the answer is ‘yes’, public provision through public procurement and private supply is better than public production. Thus, there is no need for an SOE.

4. 4. Under direct public provision through production, SOEs are ‘better suited to situations in which there exists an economic market for the outputs, the goals of the government intervention are relatively simple, a relatively large emphasis is placed on financial objectives relative to non-financial ones, and there is the need to be innovative’.

5. 5. Finally, SOEs should not be implemented if the ‘welfare losses from government failure, rent seeking, and government intervention induced technical/allocative inefficiency exceed the welfare losses due to the market failure’.

Putniņš (Reference Putniņš, Bernier, Florio and Bance2020, p. 217) frames his approach in terms of Kaldor‒Hicks efficiency, where an action that increases social welfare but makes some people worse off can be turned into a Pareto improvement by compensating the losers.

However, comparing social welfare changes ‘with’ and ‘without’ an SOE cannot rely on market signals to calculate social benefits and costs, as such markets either lead to undesirable equilibria or do not exist. Thus, framing the reasoning in terms of ‘local’ market failures can be misleading.

The notion of social efficiency in public economics, as seen in Atkinson and Stiglitz (Reference Atkinson and Stiglitz1980), involves defining social marginal costs, shadow prices, and a social discount rate as policy evaluation criteria. However, most empirical literature on SOEs (and privatizations), including performance analyses reviewed in Section 3, often uses market prices as evaluation signals, referencing ‘local’ market failures. Since any legal entity under government control is defined as an SOE according to the OECD (2024) definition, there is a risk of evaluating diverse organizations using a benchmark suited only for private or profit-oriented entities. Therefore, we present a benchmark analytical framework more suitable for SOEs under our normative definition.

2.3 A Cost‒Benefit Analysis Perspective: Observed vs Shadow Prices

The creation and operation of an SOE can be considered a public investment, thus evaluated in terms of societal welfare changes. Here, we present a simple social cost‒benefit analysis (CBA) framework in a second-best environment (Lipsey Reference Lipsey2007), as a benchmark model for public policy or project evaluation. We gradually relax some assumptions and discuss issues such as non-benevolent governments, principal‒agent problems, state capture, and administrative inefficiency or corruption (Hellman et al. Reference Hellman, Jones and Kaufmann2003).

Starting with a simple second-best environment: a benevolent, fully informed government aims to maximize social welfare but cannot apply optimal personalized lump-sum taxes. It controls the economy through prices, distortive taxes, quantity constraints on private production and consumption, and property rights laws. Some goods, unprofitable at market prices, may be socially desirable, warranting public production. Public production plans can be executed by SOEs, but they would be costly to delegate to private enterprises, such as through compensatory subsidies or rents.

Government decisions and SOE plans should use shadow prices for inputs and outputs. Shadow prices, the marginal social value of goods, can differ significantly from market prices (Johansson and Kriström Reference Johansson and Kriström2018; Florio and Pancotti Reference Florio and Pancotti2023). Establishing an SOE is worthwhile if social benefits exceed social costs at shadow prices compared to alternatives, creating a shadow profit from production.

Unlike socialist planning, the government uses signals like taxes and production plans in a mixed economy. A policy links changes in public sector production plans to signals (Drèze and Stern Reference Drèze and Stern1990). For example, providing more carbon-free electricity requires excess demand, net of private provision at tax/subsidy-inclusive market prices. The shadow price of carbon-free electricity relates to the policy enabling this production. Only with the ‘right’ policy are the shadow prices guiding SOE decisions ‘right’ (Box 2).

An SOE without a valuable public production plan and supporting right policy is meaningless in social welfare terms. A benevolent and fully informed government needs a supporting policy for an SOE to be necessary but not sufficient. Del Bo and Florio (Reference Del Bo and Florio2012) suggest these conditions for SOEs: (1) private enterprises are free to exit unprofitable markets; (2) the government is benevolent, fully informed, and able to select optimal policies, but cannot use lump-sum taxes or optimal quantity constraints on all goods; (3) public projects pass a social cost‒benefit test at shadow prices; (4) public procurement or subsidies to private enterprises are costlier than establishing an SOE.

These conditions justify SOEs as public sector production units filling demand gaps and SOEs are needed when public provision is more efficient than subsidizing private firms, such as in regulated electricity markets (Florio Reference Florio2013) or drug markets (Florio et al. Reference Florio, Gamba and Pancotti2023).

Privatization literature often assumes government inefficiency in managing SOEs while regulators can manage private enterprises efficiently. This dichotomy is flawed. Evaluating economic decisions in welfare terms requires optimal supporting policies across government branches. Suboptimal policies disrupt welfare calculations for public provision/production benefits.

However, these conditions are restrictive. What if the government is not benevolent? This concern will be addressed next.

2.4 Departures From Second Best: The Quality of Government

The previous discussion should be considered as a benchmark case. Governments consist of individuals with varying degrees of public ethos, motivation, skills, and access to information. To what extent would the previous conditions justifying an SOE withstand the criticism that government failures may be as important as, or even greater than, market failures?

Ignoring for a while the issue of imperfect or asymmetric information, it is useful to distinguish (Hellman et al. Reference Hellman, Jones and Kaufmann2003; Alesina and Tabellini Reference Alesina and Tabellini2004) between two concepts of non-benevolent government: ‘state capture’ and ‘administrative corruption’. State capture refers to corruption aimed at changing the laws, while administrative corruption refers to corruption aimed at influencing the implementation of existing laws. To simplify, let us assume that any government consists of three types of agents: policy makers, who select the laws (policy options); civil servants or public administrators, who select production plans; and managers of SOEs, who implement the production plans with a degree of discretion. Besley (Reference Besley2006) focuses on good versus bad politicians in terms of political agency in a democracy, while Niskanen (Reference Niskanen1971) and others focus on a self-interested bureaucracy with a private agenda. Managers of SOEs may be considered similar to bureaucrats according to this public choice literature, as agents ultimately responding to politicians.

We shall now remove the assumptions of the previous second-best world one at a time, and still assume that politicians as principals are benevolent and fully informed, while the agents in the departments and in the public enterprises are self-interested and possibly corrupted. A simple thought experiment in this context suggests that the marginal social value of goods is not affected: civil servants and SOE managers, even if able to distort the production plans, should still be instructed to evaluate projects, that is, marginal changes of the plan, as explained earlier. Shadow prices depend upon optimal policies, not upon optimal plans, because shadow prices by definition concern marginal changes and are not influenced by inframarginal values. Optimal plans are desirable, but if suboptimal plans are adopted, the evaluation of benefits and costs is still possible. The production plan must be feasible, and resources must be available. The plan should respect exogenous constraints, such as the quantity and quality of the labour force to be hired. However, if the social benefits of a (small) project exceed the costs at the ‘right’ shadow prices, the project would, in principle, increase social welfare (shadow profits), even if the production plan was initially exaggerated or otherwise distorted. In a sense, this is a ‘local’ government failure, and remedies can be found after monitoring and evaluating the bad implementation of the project in social welfare terms.

The situation is different if self-interested politicians adopt suboptimal policies favourable to themselves or their cronies. This will undermine project selection because shadow prices will not lead decisions in the right (social welfare-enhancing) direction anymore. If voters cannot discriminate between good and bad politicians, according to the political agency theory by Besley (Reference Besley2006), there are three main wide consequences: (1) government fails when policies result in a society being inside its Pareto frontier (Besley Reference Besley2006, p. 49), meaning there are feasible and more socially desirable policies not adopted; (2b) politicians adopt policies favouring political elites or special interests, with adverse distributional effects, an issue prominent in Pareto’s views in his Treatise of Sociology (1916), but not in his Cours d’Économie politique (1896); (3) politicians adopt policies not leading to a Pareto improvement (in the economist’s meaning) compared to a ‘do nothing option’.

Let us reverse the previous setting: civil servants and public enterprise managers are benevolent in designing and implementing production plans, while politicians are corrupt. For example, if elected law-makers adopt a tax structure falling under one of the three aforementioned cases, it would always be possible to improve social welfare by not implementing the project and simply improving the tax structure. From the administrator’s or manager’s perspective, the existing suboptimal policy cannot be ignored because any new SOE’s expenditure will create unnecessary welfare losses.

This has the unavoidable consequence that government corruption distorts both decisions to create (e.g., nationalizations) or privatize SOEs. Laffont (Reference Laffont2005) suggested that benevolent or very corrupt governments may both privatize (or nationalize) SOEs, for very different reasons. Governments with an intermediate level of corruption may behave by combining two objectives: maximization of social welfare and maximization of private benefits of the politicians, and the outcome in social welfare terms should be evaluated on a case-by-case basis.

In his own words:

See Box 3 for illustrative examples of cases in which an SOE should or should not be privatized, or vice versa.

Now, consider an intermediate case: managers of public enterprises, public administrators, and politicians are, to a certain extent, self-interested. However, constitutional arrangements, such as the rule of law and an independent judiciary, establish a system of checks and balances. In such an environment, extracting private rents through policy design and/or mismanagement is relatively costly for both principals and agents. In this context, shadow prices remain sufficient statistics for the value of changes of the public plan. Therefore, investment decisions by SOEs can still be based on welfare-improving reasoning.

An undistorted policy adoption process does not require that all politicians are fully altruistic. This would be unrealistic and an overly strict condition, implying that SOE managers should be instructed to achieve the maximum possible social welfare. Reasonably good policies can still be adopted if politicians, even if not fully benevolent, are able to extract rents from their office through non-law-making channels. Politicians, especially those in majority rather than opposition parties, can enjoy rents in the form of remuneration, free flights, and accommodation, secretarial staff, and other perks. They benefit from their prestige and high social status and may even enjoy the stamina of political competition per se. However, such forms of selfish motivations do not necessarily imply adopting the ‘wrong’ policies, where ‘wrong’ implies deliberately deviating from what can be achieved in social welfare terms, given the available resources, information, and legal, and cultural constraints. Politicians may still adopt welfare-improving policies while enjoying their privileges.

Political agency theory discusses issues relevant to our topic. An example is logrolling activities, described as Coasian political coalitions by Besley (Reference Besley2006). Such coalitions, where votes are exchanged to obtain electoral majority, can be either beneficial or highly damaging in terms of social welfare in an elected parliament. Autocratic systems are prone to government capture, but democracies are not entirely protected from opportunism. Early public choice economists have often focused on rules (‘constitutions’, ‘budgetary rules’, etc.) that constrain policy adoption. The welfare effect of such rules cannot be established once and for all. More generally, transparency, freedom of information, tenure duration, public ethos, etc., can be seen as mechanisms to protect policymaking from rent seeking.

2.5 Soft-Budget Constraints

Earlier literature on SOEs was particularly concerned with Kornai’s (1979) argument that loss-making firms in the public sector are not disciplined by the same budget constraints that private firms must face. According to Maskin (Reference Maskin1996), there are three possible origins of a situation where ‘a funding source finds it impossible to keep an enterprise to a fixed budget, i.e., … the enterprise can extract ex-post a bigger subsidy or loan than would have been considered efficient ex-ante’:

1. 1. Centralized credit: Suppose there are two types of projects, ex-ante ‘slow’ and ‘fast’ in terms of payback. Credit centralization is inefficient because a single bank will fund both, as it cannot discriminate ex-ante, and slow projects anticipate that the bank will not stop funding them ex-post.

2. 2. Monopoly: Government bargains investment with a monopolist and offers a subsidy to unprofitable projects. This can be used by the monopolist to extract rents from the government, creating allocative inefficiency, and the social cost of distortionary taxation to fund the subsidy.

3. 3. Ownership: Under ‘socialism’, the banks and firms are jointly owned, leading to inefficiency even under decentralization, as financing decisions are influenced by the firms.

While these models often explained the inefficiency of Soviet-type firms, they have also been used to criticize SOEs in market economies receiving government subsidies or excessive loans. This inspired EU legislation forbidding ‘state aid’ to firms, with exceptions granted by the European Commission in specific cases as a commitment mechanism.

Empirical evidence does not strongly support the application of soft-budget constraints to contemporary SOEs in developed market economies. First, they are often subject to similar legal constraints as private firms, and second, in regulated industries, regulators often oppose government budgets supporting incumbents. This mostly empirical issue now includes theoretical arguments that depart from Kornai’s views. Dewatripont and Maskin (Reference Dewatripont and Maskin1995), and Bai and Wang (Reference Bai and Wang2022) suggest that constraining an agent to immediately terminate a bad project is efficient even with agency issues in project selection. They assume the existence of a project screening agent hired by the principal. The agent’s effort is not observable by the principal. Only after the agent implements a selected project does the principal observe its quality and decide to discontinue it. Restricting project termination can decrease the principal’s cost by incentivizing the agent to put more effort into initial selection. In the authors’ words: ‘Although continuing a bad project has a cost to the principal, it is worthwhile as long as the marginal benefit of higher effort is greater than the cost’.

Whether this argument is relevant for SOEs in developed market economies, the cost‒benefit analysis discussed previously needs modification when there is information asymmetry and a principal‒agent problem involving government departments and SOE managers.

2.6 Information Asymmetry, SOEs, and Public Procurement

Information asymmetry and/or incompleteness have been core topics in public economics for decades, impacting SOE theory. Public provision through public enterprises and public procurement through privately owned enterprises cannot be assumed to bring the same benefits, as outlined by Laffont and Tirole (Reference Laffont and Tirole1993). Under information asymmetry, an SOE may be a socially desirable mechanism to counteract socially costly rents of more efficient agent types. The difficulty in specifying ex-ante contracts is solved by ownership rights, applicable to both private and public enterprises under contract. For example, ownership of a firm includes the right of the shareholder to be continuously informed and may be embodied in the right to appoint directors.

Under asymmetric and incomplete information, public production by SOEs and public procurement by private enterprises will not be welfare equivalent. Public ownership may be superior when the cost of incentivizing the private firm under a procurement contract is too high. Two inefficiencies need to be compared: the socially costly rents earned under public procurement by private owners and the socially costly rents of the not fully benevolent regulator (see Willner Reference Willner2010 and Ceriani and Florio Reference Ceriani and Florio2011). Ceriani and Florio (Reference Ceriani and Florio2011) consider a network industry. Initially, the service is provided by a public monopolist with some x-inefficiency compared to cost minimization. Consumer surplus is then computed in a comparative static framework, and indifference conditions are found under an unregulated private monopoly, a regulated private monopoly, a vertically disintegrated monopoly, a duopoly, and a liberalized market. The welfare results depend on the relative size of the x-inefficiencies of the public monopolist, allocative inefficiencies of the private monopoly, the cost of unbundling, and the costs related to establishing a competitive market.

While consumer surplus may be informative and easy to estimate, the comparison of the welfare impact of these inefficiencies cannot be done at market prices because such prices are distorted. This has implications when interpreting empirical analyses comparing profitability between SOEs and private companies, a topic discussed in Section 3.

2.7 Mixed Oligopoly

Private enterprises and SOEs increasingly operate in the same market environment. An important strand of literature related to our topic deals with mixed oligopoly, where a social welfare-maximizing SOE competes in a market with profit-maximizing firms. The message from these models is that the resulting equilibria are often (but not always) welfare-superior to an oligopoly with only private competitors (see De Fraja and Del Bono Reference De Fraja and Delbono1990). Strategic interactions are studied in seminal papers such as those by De Fraja and Del Bono (Reference De Fraja and Delbono1990) and Matsumura (Reference Matsumura1998). More applied perspectives investigate how universal service obligations in certain industries have redefined the role of public enterprises in markets open to competition (see, for example, Bel and Calzada Reference Bel and Calzada2009).

Other studies in this context depart from the assumption that public managers have no private agendas and combine mixed oligopoly with public choice issues. A theoretical contribution that combines profit and size or scale objectives is Sappington and Sidak (Reference Sappington and Sidak2003), who opened the way to literature on predatory pricing by SOEs below marginal costs to acquire larger market shares and bigger volumes of operations. In some models, the SOE has an objective function that combines (with a set of weights) social welfare and certain firms’ profits or other benefits (Matsumura Reference Matsumura1998, Lee et al. Reference Lee and Hwang2003, Hindriks and Claude Reference Hindriks and Claude2006). In others, the scale of operations takes the place of social welfare.

For a review of the topic, readers are referred to another Cambridge Element in this series: Poyago-Theotoky (Reference Poyago-Theotoky2024). In her words:

The author identifies three topics that are particularly important in recent developments of the literature:

1. 1. Whether the presence of public ownership in an oligopolistic market may have a positive impact on service quality in sectors such as health services, education, transport, and postal services.

2. 2. The potential role of SOEs in these markets in sustaining environmental policy objectives and corporate social responsibility.

3. 3. The different motivations of management and staff when SOEs compete with private firms.

She also reports advancements in the literature on R&D investment under mixed oligopoly, a topic we discuss later from a CBA perspective. An interesting topic within an industrial organization framework is optimal privatization when there are knowledge spillovers, following Gil-Moltó et al. (Reference Gil-Moltó, Poyago-Theotoky and Zikos2011, Reference Gil-Moltó, Poyago-Theotoky, Rodrigues-Neto and Zikos2020). The market structures considered in these models are both duopoly and oligopoly, with the usual assumption that the private firm maximizes profits while the SOE maximizes social welfare, both being recipients of R&D subsidies. It has been found that the greater the spillovers, the greater the socially optimal R&D; thus, privatization may decrease aggregate R&D and social welfare. However, if in a mixed market the number of competitors is high, the result can be reversed, and privatization can be beneficial. This suggests that R&D subsidies (including tax expenditures) should be designed by carefully considering market structures.

Readers are referred to Poyago-Theotoky (Reference Poyago-Theotoky2024) for details on these models and policy implications. We shall briefly discuss these issues below and in the review of the empirical literature in Section 3.

2.8 Prosocial Behaviour

A substantial stream of more recent literature has explored the role of intrinsic motivation and prosocial behaviour in the public sector, both with theoretical models and some empirical studies (Polidori and Teobaldelli Reference Polidori and Teobaldelli2013). While this literature has not been systematically linked to SOEs, we believe that this is potentially an interesting research area for our subject given our broader definition (which would include, for example, providers of health and education services). How should SOE theory be redefined when managers and staff are hired from the more altruistic segment of the population, characterized by a distribution of different types with respect to the degree of altruism?

According to Francois and Vlassopoulos (Reference Francois and Vlassopoulos2008), staff with intrinsic motivation may represent an advantage for public organizations, relative to their private counterparts, in terms of level of effort, loyalty to the public mission, and internalization of norms and objectives. The existence of a specific public service motivation is well-known among scholars in public administration. Economists have been slower to consider that some people are happy to contribute to the public good (in a broader meaning, such as public health) and what this implies for the organizations hiring these workers. The empirical evidence suggests that there is a sorting mechanism, with individuals with higher intrinsic motivation for specific jobs being more often hired in public sector organizations. The evidence is not without some controversies, possibly due to confounding factors, but also to the fuzzy boundaries of the concept. Bénabou and Tirole (Reference Bénabou and Tirole2006), for example, argue that prosocial behaviour may be driven by three different mechanisms: intrinsic (psychological); extrinsic (material rewards outside the organization); and reputational (which may grant higher rewards elsewhere). This discussion is also related to different concepts of altruism, as illustrated by the idea of ‘warm glow’ (a form of utility), see Andreoni (Reference Andreoni1990), Tonin and Vlassopoulos (Reference Tonin and Vlassopoulos2010).

Theoretical models, according to Polidori and Teobaldelli (Reference Polidori and Teobaldelli2013), often boil down to agents’ utility functions where the ingredients are the satisfaction arising from income-based consumption, less disutility of effort, plus the utility from intrinsic motivation. Impure altruism (Besley and Gathak Reference Besley and Gathak2005) would link the latter to effort, while pure altruism (Francois Reference Francois2000) to the actual level of the public good provided. These models are often framed in a principal‒agent relation with different types and moral hazard.

Some of this literature points to the displacing effect on intrinsic motivation of offering monetary reward for effort in the public sector, and ultimately on the inefficiency of public organizations trying to mimic the incentives offered to employees by private companies. Differently from some versions of the ‘new public management’ approach (Hood Reference Hood1995), paying more money to do more may have an adverse effect in an organization where doing more (e.g., for patients) or achieving more (e.g., for public health) is seen by the altruistic types as a reward in itself. A theoretical example of why this may happen is Grönblom and Willner (Reference Grönblom and Willner2014), in a model where motivation crowding out can happen in SOEs with performance-related pay, but also in privatized firms. The authors combine agency theory and potential intrinsic motivation and show how ‘fat cat’ salaries may arise from privatization. They conclude that ‘no form of ownership or market structure is always superior if intrinsic motivation can exist or be crowded out in all organizations’. Willner and Grönblom (Reference Willner, Grönblom, Bernier, Florio and Bance2020) extend the theoretical analysis to a comparison between public monopoly and oligopoly with potential intrinsic motivation. Their model is inspired by earlier work by James (Reference James2005), who explores the impact of extrinsic remuneration on both performance and intrinsic motivation.

In our welfare framework, the benefit‒cost ratio of establishing an SOE would be enhanced if it can be predicted that intrinsically motivated personnel would be attracted to work (with higher effort and lower salary than elsewhere) especially in organizations where profit maximization is not the sole or dominant objective (such as health, education, environmental services, research). Of course, empirical analysis is necessary to confirm whether this hypothesis holds true.

2.9 Looking Forward: New Public Missions and New Types of Public Enterprises

The aforementioned social cost‒benefit conceptual framework is potentially helpful when considering new public missions for SOEs, such as their role in knowledge creation. There is empirical evidence that private enterprises, including privatized companies, tend to underinvest in R&D for reasons well explained in the seminal theoretical paper by Romer (Reference Romer1990) and in several other contributions of the endogenous growth literature, both theoretical and empirical (see also Section 2.7 for R&D in mixed oligopoly). For an empirical analysis of the decline of corporate R&D, see Arora et al. (Reference Arora, Belenzon and Patacconi2017, p. 3), who study trends of scientific publications co-signed by researchers affiliated with a company since 1980. They conclude that: ‘large firms still value the golden eggs of science (as reflected in patents) but seem to be increasingly unwilling to invest in the golden goose itself (the internal scientific capabilities)’.

The core reason for this decline is related to the existence of a positive externality, such that knowledge acquired by a firm spills over to other firms through different transmission channels, including product and process imitation, mobility of researchers across firms, and information provided by publications and patents. The latter, while on one hand offering a temporary legal monopoly to innovators, on the other hand discloses (or should disclose) information that can be instrumental to the creation of further knowledge. Hence, because private firms cannot entirely protect their investments, they will underinvest compared to what would be socially desirable.

While a popular policy course by governments is to offer subsidies or tax rebates to firms to push their R&D activities and spending (the pharmaceutical industry is a notable example), an alternative is the direct provision of knowledge and knowledge-based products and services by government-sponsored organizations. This may happen in different ways, for example through government funding of research universities or institutes. Moreover, some SOEs may be instructed by the government to sustain their R&D expenditures beyond what a short-term cost minimization objective would justify.

The social cost‒benefit analysis of such policies, particularly regarding the potential role of SOEs, presents intriguing challenges in the realm of public investment. This is because, unlike the provision of already existing goods, such as transport or electricity, where marginal social value can theoretically be computed based on existing information, evaluating a project that produces a good of unknown value poses a significant question.

If scientific knowledge is viewed as a public good with initially unknown value, its production could be considered a novel (risky) public service, addressing investment challenges that are unappealing to private investors. Organizations specializing in knowledge creation could thus be regarded as new types of public enterprises. For example, the National Institutes of Health (NIH) is the most important organization for biomedical research in the world under the health department of the federal government of the US. While most of the NIH budget (over 50 billion USD in 2023) is devoted to disbursing research grants (ultimately an indirect subsidy to the pharmaceutical industry), its Intramural Research Program is among the largest biomedical R&D public organizations in the world, with around USD 5 billion per year and about 1,200 principal investigators and 5,000 Postdoctoral Fellows. (See details at https://irp.nih.gov/about-us). In a sense, NIH actually or potentially competes with R&D activities of privately owned drug companies. NASA, the European Space Agency, US National Laboratories, and many government-funded bodies and research infrastructures with autonomous budgets and management increasingly resemble a distinct form of SOE delivering knowledge in specific fields as a novel type of public service (Florio Reference Florio2019).

Lazzarini et al. (Reference Lazzarini, Mesquita, Monteiro and Musacchio2021) suggest that SOEs are given, by governments, the task to perform risky investments along new technological paths. They build an extended agency model where ‘given relatively higher levels of managerial autonomy, SOEs may outperform private-owned enterprises (POEs) in some types of inventive output’, particularly in countries with a good quality of government. The empirical results confirm this intuition and are discussed in the next section.

More generally, governments have identified a new set of priorities, including energy transition and climate change, sustainable mobility, science and innovation policy, the digital divide, and needs arising from ageing and migration. Additionally, there is a focus on addressing ‘capitalism’s good jobs problem’ (Rodrik and Stantcheva Reference Rodrik and Stantcheva2021). Public missions of SOEs and their roles may be reshaped by these policy priorities in countries where governments support a new sustainable development agenda for the twenty-first century. This field is still developing, but government science and innovation policies increasingly invest in large-scale research and technology infrastructures, exemplified by initiatives like the European Strategy Forum on Research Infrastructures (ESFRI). A cost‒benefit analysis of public investment in research infrastructure emphasizes forecasting benefits primarily deriving from R&D, including human capital formation, technological spillovers, direct cultural effects, and a public good value as a form of indirect cultural effect, even when the future value of discovery is entirely unknown, see Box 4.

Box 4 CBA of research infrastructures

For research infrastructures (RIs) the net present value of social benefits can be estimated by a simple equation:

$NP{V}_{RI}=\left[SC+HC+TE+AR+CU\right]+B_n-\left[K+L_s+L_o+OP+EXT\right]$

Benefits include: SC = knowledge output value; HC = human capital formation; TE = technological externalities to firms supplying technologies; AR = benefits to downstream users of innovations; CU = cultural goods arising from outreach activities; Bn = non-use value based on the willingness to pay of taxpayers for scientific knowledge, a public good. Costs include: K = economic value of capital; Ls = labour cost of scientists; Lo = other staff costs; OP = other operating costs; EXT = negative externalities arising from the construction and operation of the RI. Labour costs are estimated by shadow wages. In general, marginal costs or willingness to pay are proxies of shadow prices. Each component is taken at its present value, the discounting process is s_t = 1/(1 + r)^t where r is the social discount rate. SC is measured by the opportunity cost of scientists to produce a publication and to read and cite another publication.  HC is valued as the increased earnings gained by former RI’s students and former employees, since the time they leave the RI project, against a suitable counterfactual scenario. TE is given by the discounted incremental shadow profits S_ts_tΠ_jt by companies (j) of the RI’s supply chain or by other economic agents who have benefit from a learning externality. AR is project specific (health for example), and each of them ultimately is related to the willingness to pay for them by users. CU carried out by the RI produce different cultural effects on the general public, which can be valued by estimating the willingness to pay W_gt for such activities. Finally, the term B_n captures the social value of a pure public good of discovery of nature by taxpayers and can be proxied by stated or revealed willingness to pay for discovery, similar to non-use values in environmental economics.

Source: Florio (Reference Florio2019).

3.1 Performance: What is Measured and What Should be?

A significant portion of empirical literature on the state-ownership effect on performance refers to a rather standardized regression model, at least in its baseline form (see Box 5). What distinguishes the existing literature, however, is how firm-level performance is measured. A first set of measures broadly considers a firm as successful by looking at its ability to generate profits. Profitability can then be measured by looking at simple measures of profits or profit growth, or by resorting to a series of accounting measures, usually in the form of ratios. Empirical studies comparing private and public enterprises along these lines are discussed in Section 3.1.1.

Performance can be measured from another angle by considering a firm’s ability to combine inputs in an efficient manner in order to produce the maximum amount of output, given the technological possibilities available to the firm, summarized by the production function. This type of reasoning leads to the use of productivity measures, the simplest being input productivity, most notably labour, or total factor productivity when all inputs are considered (Section 3.1.2).

Typically, these two families of indicators are usually deemed sufficient to analyze and compare privately owned firms. However, when considering SOEs, additional aspects should be taken into account. When comparing private and public enterprises, one should bear in mind that SOEs might have a broader mandate than their private counterparts and be assigned, by their government principals, multiple objectives, other than simple profit maximization or productive efficiency. Therefore, other measures of performance should be considered and more research is still needed to reach a consensus on how to translate the multiple goals that characterize SOEs into testable hypotheses.

Box 5 Comparing the performance of SOEs and POEs

Most of the analyses proposed in the empirical literature to compare the performance of SOEs and POEs are based on a model, more or less explicitly formalized, of the following type:

$Y_i=\alpha +\beta SO{E}_i+{\sum }_{j=1}^J{\gamma }_j{Z}_{ij}+{\sum }_{c=1}^C{\delta }_c{W}_{ic}+{\sum }_{s=1}^S{\eta }_s{V}_{is}+{\epsilon }_i$ (1)

where:

• - the observation unit i is an individual firm;

• - the dependent variable $Y_i$ is the performance variable of interest;

• - $SO{E}_i$ is the explanatory variable of interest, identifying SOEs alternatively through dichotomous or continuous measures;

• - $Z_{ij}$ are firm-level control variables, where j = 1, … , J;

• - $W_{ic}$ are country level control variables, where c = 1, … , C;

• - $V_{is}$ are industry level control variables, where s = 1, … , S;

• - ${\epsilon }_i$ represents the error term.

In this model, the coefficient $\beta$ represents the state-ownership effect on firm-level performance. When panel data are available, the temporal dimension t can be exploited by including time fixed effects, $f_t$ , and time-variant controls ( $Z_{ijt}$ , $W_{ict}$ , $V_{ist}$ ).

By including control variables at each level, the baseline model allows for specific influences of firm, country, and industry characteristics on the dependent variable. However, many papers circumscribe the scope of analysis by narrowing the sample of observation to a single country (or geographic area) or to a specific industry (or group of industries). This has the benefit of comparing SOEs and POEs within a more homogenous firms’ domain, albeit at the inevitable cost of reduced generalizability.

The broader and more heterogeneous the sample of considered firms, the more relevant the endogeneity concerns associated with the fact that the state does not randomly select firms in which to hold ownership stakes. In this case, to better identify the treatment effect of state ownership, it can be useful to employ matching methods that make the treated group (SOEs) and the control group of POEs as comparable as possible, except for ownership differences. This involves adopting matching techniques suitable for selecting a subset of POEs that resemble SOEs based on a specific set of observables. See, for example, Abadie et al. (Reference Abadie, Drukker, Herr and Imbens2004) and Abadie and Imbens (Reference Abadie and Imbens2011) for details on nearest-neighbour matching estimators. At the researcher’s discretion, the search for similar firms can be conducted by country or by industry to impose matching within a sufficiently homogeneous environment.

Source: The authors, adapted from the referenced empirical literature in this section.

3.1.1 Profitability Indicators

Firms are usually thought of as profit-maximizing entities (see Section 2.1). Thus, in this perspective, a natural benchmark for comparing private firms and SOEs is examining whether ownership has an effect on profitability indicators, with the caveat that a firm can be highly profitable for the ‘wrong reasons’ and in fact part of the reported profits can be considered as rents in an economic perspective. Unfortunately, accounting indicators do not convey the difference, and are taken at face value by most of the empirical literature that we discuss later.

A few of the most well-known and used indicators include Return on Sales (ROS), Return on Assets (ROA) and Return on Equity (ROE) and Tobin’s q, the latter two for listed firms. ROS is typically used to evaluate a firm’s operational efficiency and provides information on how much of total sales are actually converted into profits. It is computed as profits over sales. ROA summarizes how much profit a firm generates by means of its assets, and thus hints towards the efficiency of a firm’s management structure in generating value for investors from the economic resources it has invested in the past. It is computed as profits over total assets. ROE is an indicator of a firm’s ability to generate profits from shareholders’ investments. It is computed as profits over shareholder equity (or market value). Tobin’s q sheds light on a firm’s equilibrium between market and intrinsic value. It is computed as a firm’s market value divided by its assets’ replacement cost.

Studies based on the analysis of samples of large and in some cases listed firms, from different non-financial sectors and with an international coverage are the main focus of this subsection. Presented in a chronological order, the first contributions, such as Dewenter and Malatesta (Reference Dewenter and Malatesta2001), find that state ownership is associated with lower profitability, while more recent papers find that SOEs are similar, or even better performing, than POEs. Single country studies, of which we present a selection based on geographic coverage, present a similar time trend, with earlier studies suggesting that SOEs are not as performing as POEs; however, results might be dependent on the specific country analyzed.

Dewenter and Malatesta (Reference Dewenter and Malatesta2001) offer a seminal and thorough analysis of the relative performance of private versus public enterprises in financial terms. With data for 1975, 1985, and 1995 for the 500 Fortune companies, the authors consider the impact of ownership on profitability (measured by accounting ratios, ROS, ROA, and ROE), leverage (total liabilities over assets), and labour intensity. Focusing on profitability, regression results clearly point towards lower and statistically distinguishable from zero estimated coefficients for SOEs. This result suggests that, controlling for relevant firm-level variables (size and sector) along with national growth and regional and time dummies, public ownership is detrimental for firms’ profitability. The results also suggest that SOEs tend to hire more labour than their private counterparts. This paper, from our perspective, has an important limitation in terms of interpretation for welfare analysis, as it does not control for price levels (and for mark-ups correlated with market concentration) when SOEs and POEs operate in a mixed oligopoly. Hence, if an SOE sells the service at a lower price than the POE, it will look as less profitable even if it covers its costs and is not loss making. It is an empirical fact that in the same market different prices are possible, particularly in network/regulated industries, for example because of incomplete information on the consumers’ side. Florio (Reference Florio2013) for example shows that in the EU for electricity, gas, and telecommunications SOEs or partially government-invested companies over several decades tended to charge lower prices than POEs.

Wolf (Reference Wolf2009), in a cross section of countries between 1987 and 2006, examines the differential in several output indicators, including profitability (along with production, output efficiency and revenue generation), between private and public firms in the oil industry. Empirical results clearly support the view that SOEs underperform POEs along the examined dimensions. However, using more recent and comparable data, Kowalski et al. (Reference Kowalski, Büge, Sztajerowska and Egeland2013), in a paper analyzing the behaviour and role of SOEs in an international trade perspective, provide a descriptive statistical comparison of profitability of public versus private firms in the sample of Forbes Global 2000 list. Large state-owned firms are similar to their private counterparts when profitability is measured by means of ROE; they outperform private firms in terms of ROS and underperform if ROA is considered instead. Similar results were obtained by Bernier et al. (Reference Bernier, Florio and Bance2020b), who found that SOEs belonging to the Forbes 2000 list have higher ROS than all the Forbes 2000 firms, suggesting that they are outperforming private firms, while ROA is lower. Comparison based on ROA and cost share, computed as operational costs as share of sales revenue, is at the basis of an econometric analysis of Norwegian firms in 1999 (Christiansen and Kim Reference Christiansen and Kim2014). The authors, while providing evidence of the increasing role of SOEs in the marketplace over the previous decade, compare the performance of private and public enterprises in five sectors (air transportation, electricity, mining, oil & gas, and telecommunication) using data for 2013 from the Forbes Global 2000 list of companies, showing that SOEs have generally exhibited higher rates of return with respect to POEs.

This reverses the message in the previous literature in the wake of Dewenter and Malatesta (Reference Dewenter and Malatesta2001) and the same question should be asked: are now SOEs leveraging on higher market power to earn rents? Or are they able to earn good returns and still charge lower prices than their private counterparts?

Lazzarini and Musacchio (Reference Lazzarini and Musacchio2018) show that the performance of listed SOEs, compared by matching techniques to their relevant private counterparts, is similar, except in the case of severe recessions or similar shocks (see Box 6). Their analyzed sample, made up of 477 SOEs that are listed on the stock markets in sixty-six countries, between 1997 and 2012, exhibits similar ROA and Tobin’s q with respect to privately owned firms. These findings are obtained by means of a difference-in-difference estimation and static and dynamic panel methods. This finding is robust to different definitions of ownership, including majority and minority state-ownership, and across different industries, hence it is more convincing than other studies which are more descriptive.

Box 6 Comparing SOEs and POEs performance amid exogenous environmental change

Matching methods discussed in Box 5 are useful when the “selection on observables” assumption holds, that is, when the treated group and the control group differ only by a set of observable characteristics. Any bias due to unobservable factors need alternative approaches. If state ownership was a time-variant variable, this would motivate the use of panel data and fixed effect models. Unfortunately, state ownership is typically (quasi-)time invariant so that firm fixed effects would absorb (most of) the ownership effect of our interest. This is equivalent to saying that it is impossible to apply a difference-in-differences approach within this framework, since it is not possible to observe most of the treated (SOEs) before and after the treatment (change in ownership). In addition, any changes in ownership are likely to be endogenous, adding concerns of biased estimates.

However, some empirical analyses (see, e.g., Lazzarini and Musacchio Reference Lazzarini and Musacchio2018) have managed to extract useful information on the state-ownership effect by assessing the performance of SOEs in the context of an exogenous environmental change. In this case, it is possible to resort to a model of the following type:

$Y_{it}=\alpha E{C}_{cst}+\beta SO{E}_i\cdot E{C}_{cst}+{\sum }_{j=1}^J{\gamma }_j{Z}_{ijt}+{\sum }_{c=1}^C{\delta }_c{W}_{ict}+{\sum }_{s=1}^S{\eta }_s{V}_{ist}+u_i+f_t+{\epsilon }_{it}$(2)

where $E{C}_{cst}$ is the variable measuring a specific event at the country-industry-year level, that is, the environmental change of interest. As the interaction term with SOE_i is time-variant, it is now possible to include firm fixed effects, $u_i$ , to control for unobserved heterogeneity. Time fixed effects, $f_t$ , can also be included to capture time-specific effects. All other time-variant controls can be retained in the equation. In this framework, the coefficient $\alpha$ represents the effect of a particular event on the performance of POEs, while the coefficient $\beta$ captures the differential effect of state-ownership as a consequence of the event itself.

While this approach does not allow for the assessment of the state-ownership effect before the event of interest (since the non-interacted term $SO{E}_i$ is absorbed by firm fixed effects), it enables the measurement of the differential effect of state ownership during the event. For instance, Lazzarini and Musacchio (Reference Lazzarini and Musacchio2018) employ this strategy to demonstrate that severe recessions decrease the profitability of SOEs compared to similar POEs. This highlights how the profitability of SOEs needs to be balanced against the social and political objectives that governments seek to pursue through their ownership. Frigerio and Vandone (Reference Frigerio and Vandone2020) employ a similar approach to analyze the lending activities of state-owned development banks during electoral cycles and assess whether political opportunism exists in their operations.

Source: The authors, adapted from Lazzarini and Musacchio (Reference Lazzarini and Musacchio2018).

Goldeng et al. (Reference Goldeng, Grünfeld and Benito2008) examine the effect of ownership on ROA and on costs relative to sales revenue, along with other relevant controls at the firm level (size, age, sector, location) and at the more aggregate level (market structure) using data for registered firms in Norway during the 1990s. Taking these metrics into consideration, state-ownership, when measured as an indicator variable, is associated, all else being equal, with inferior performance compared to private ownership. Chan et al. (Reference Chan, Chen and Wong2018) consider New Zealand’s SOEs and a comparable sample of POEs between 1990 and 2010 in different sectors, and examine within a cross-sectional and panel setting, ROA, ROE, and Return on Revenue (ROR, computed as earnings before interest and tax divided by net revenue), along with asset and labour turnover, and labour intensity. Strong and unequivocal results are reported for ROA and ROR, consistently lower for SOEs. Kabaciński et al. (Reference Kabaciński, Kubiak and Szarzec2020) consider large Polish SOEs between 2013 and 2015. Estimation of their empirical model reveals that SOEs outperform private firms when considering ROA and are indistinguishable from them when considering the EBIT (earnings before interest and taxes) margin. Szarzec and Nowara (Reference Szarzec and Nowara2017) analyze the largest firms in Central and Eastern Europe between 2007 and 2013 across different non-financial sectors and find that SOEs are substantially comparable to POEs when looking at several measures of performance. De Lange and Merlevede (Reference De Lange and Merlevede2020) create a dataset of European firms across different sectors between 2002 and 2012 with detailed information on ownership and find, by looking at a series of real and financial performance indicators, that SOEs are outperformed systematically by POEs. However, they also show how the negative relationship between performance and public ownership can be mitigated in countries with good institutional quality, which is the subject of Section 3.2. Ben-Nasr (Reference Ben-Nasr2016) indicates that in government-controlled firms, shareholders attribute diminished importance to augmenting net working capital (NWC). This is evidenced by the observation that an increase in excess-NWC is correlated with a less substantial rise in firm value, particularly for government-controlled firms with a low level of NWC investment. Using a sample of European firms Jaslowitzer et al. (Reference Jaslowitzer, Megginson and Rapp2018) empirically test the relationship between state ownership and corporate investments opportunities measured by Tobin’s Q and find that state-owned firms invest considerably less and in a way that is significantly less responsive to changes in investment opportunities.

The effects of state ownership can vary depending on the phases of the economic cycle. For instance, Borisova et al. (Reference Borisova, Fotak, Holland and Megginson2015) observed that government ownership is typically linked to a higher cost of debt in normal economic conditions. However, during financial crises, when implicit government guarantees take precedence, state ownership becomes associated with a lower cost of debt. In a similar vein, Beuselinck et al. (Reference Beuselinck, Cao, Deloof and Xia2017) demonstrated that SOEs experienced a comparatively smaller decline in stock value compared to POEs in the aftermath of the Global Financial Crisis.

Bozec et al. (Reference Bozec, Dia and Breton2006), however, shed serious doubts on the validity of comparisons between POEs and SOEs based on profitability measures due to the fact that results might depend on the objectives of the firms that are being compared. They suggest performing this type of analysis by considering productivity measures, which is the object of the following subsection. We tend to agree with them based on the theoretical arguments presented in Section 2, although some caveats are in order since productivity may also be a biased measure from a welfare analysis perspective.

3.1.2 Productivity

Economists often argue that the most relevant empirical measure of firm performance is productivity which reflects a firm’s ability to use inputs in production efficiently and intensely (Comin Reference Comin, Durlauf and Blume2010). The natural empirical indicator is total factor productivity (TFP), that is, the residual of a production function (Chambers Reference Chambers1988), or the portion of output not explained by the quantity of inputs used in production. In what follows, selected studies analyzing the effect of public ownership in different countries, sectors, and time periods will be surveyed, with the aim of providing an overview of contributions studying the impact of state-ownership on TFP.

There are however some caveats to be considered. Firstly, most empirical methods estimating TFP are constrained by the fact that data on input and output in physical terms are not available, hence proxies are used starting from accounting data on revenues and costs. Unfortunately, these may be affected, on the revenue side, by the issue of price distortion, which may also be relevant on the cost side: to repeat our argument, as observed prices are not shadow prices, TFP measures at market prices share the problems highlighted when discussing profitability indicators, even if their impact is somewhat mitigated if the econometrician is aware of the problem and controls for price levels. On this issue, see Box 7.

Box 7 TFP measurement methodologies and limitations in the SOEs literature

TFP measures at the firm level can be derived from firm-level production functions. Addressing various econometric challenges associated with TFP estimation (for a recent overview, see Van Biesebroek, Reference Van Biesebroek2008, Van Beveren, Reference Van Beveren2012), three distinct methodologies can be employed. In the first estimation approach, TFP is derived from a Cobb–Douglas and translog production function: $y_{it}={\beta }_0+{\beta }_k{k}_{it}+{\beta }_l{l}_{it}+{\beta }_m{z}_{it}+v_i+u_{it}$ . Typically, the output ( $y$ ) is proxied by operating revenues, capital input ( $k$ ) by tangible fixed assets, labour input ( $l$ ) by employees, and intermediate inputs ( $z$ ) by material costs. The lowercase letters represent natural logarithms. After estimating the production function by OLS with fixed effects, a measure of TFP is derived as the residual of the estimated production function: $TFP=\exp ({\hat{\beta }}_0+{\hat{v}}_i+{\hat{u}}_{it})=\exp \left({\hat{\omega }}_i+{\hat{u}}_{it}\right)$ , where ${\omega }_i$ is firm-level unobserved productivity, and $u_{it}$ denotes unexpected deviations from the mean.

An expanded variation of this methodology (Berndt and Christensen, Reference Berndt and Christensen1973) involves adopting a translog production function that accommodates complementarities between labour and capital, as well as substitutability between capital and materials:

$y_{it}={\beta }_0+{\beta }_k{k}_{it}+{\beta }_l{l}_{it}+{\beta }_m{z}_{it}+{\beta }_{ll}{l}_{it}^2+{\beta }_{lk}{l}_{it}{k}_{it}+{\beta }_{lz}{l}_{it}{z}_{it}+{\beta }_{kk}{k}_{it}^2+{\beta }_{kz}{k}_{it}{z}_{it}+{\beta }_{zz}{z}_{it}^2+v_i+u_{it}$

A third measurement approach is derived by employing the Levinsohn and Petrin (Reference Levinsohn and Petrin2003) methodology. In this context, intermediate inputs serve as a proxy for firm-level productivity shocks. This strategy is designed to address potential simultaneity issues that may arise when unobserved productivity shocks influence firms’ choices of inputs. Finally, Ackerberg et al. (Reference Ackerberg, Caves and Frazer2015) highlight how the Olley and Pakes (Reference Olley and Pakes1996) and Levinsohn and Petrin (Reference Levinsohn and Petrin2003) TFP estimation techniques may suffer from functional dependence problems that lead to a non-identification of the labour coefficient in the first stages and propose an alternative approach.

Estimation of TFP from revenue-based data, rather than quantity data, may also suffer from input-price bias, that is, the bias that arises from unobserved (by the analyst) input prices by firms. Another potential bias can arise in the case of multi-product firms from the unobserved allocation of inputs across different product. De Loecker et al. (Reference De Loecker, Goldberg, Khandelwal and Pavcnik2016), propose a method to overcome these biases. Regardless of the methodology employed, the TFP measures discussed are susceptible to fluctuations in input and output prices as they are based on balance-sheet data (revenue and cost values). This characteristic diminishes their appropriateness for discerning and contrasting the inherent efficiency or productivity distinctions between SOEs and POEs, since the pricing mechanisms, cost structures, subsidies, and regulatory constraints encountered by SOEs typically differ from those faced by their private counterparts (see Section 2). Consequently, the relevant balance-sheet data are defined through possibly distorted prices. It is essential to acknowledge that econometric analysis comparing the productivity of SOEs and POEs encounter this limitation. To overcome this challenge, alternative measures or adjustments to TFP calculations should be considered to effectively account for contextual variations, but the empirical literature in this direction is still scant.

Source: The authors, adapted from Borghi et al. (2016) and Castelnovo et al. (2019).

Taking advantage of the widespread wave of privatization since the 1980s, Megginson and Netter (Reference Megginson and Netter2001) and Megginson (Reference Megginson2005), document a productivity advantage of POEs with respect to SOEs. A possible explanation is set forward in Bai et al. (Reference Bai, Li, Tao and Wang2000) who stress how often SOEs are providers of social welfare. Given the double task of reaching productive efficiency and social welfare provision, SOEs are given a low profit incentive, leading to lower performance with respect to their private counterparts. Boardman et al. (Reference Boardman, Vining and Weimer2016) studies the enduring impacts of privatization on productivity, focusing on privatizations in Canada. The key finding reveals a rising trend in the productivity of privatized state-owned enterprises relative to SOEs, although the productivity of privatized firms continues to mimic that of consistently private entities. A different conclusion is set forth by Willner (Reference Willner2001) who, considering ownership per se, not related to privatization, finds no significant impact of public ownership on firms’ productivity. A comprehensive review of the empirical evidence on private and public enterprises’ efficiency reinforces this finding, stating that ‘research does not support the conclusion that privately owned firms are more efficient than otherwise-comparable state-owned firms’ (Mühlenkamp Reference Mühlenkamp2015, p. 553).

A first, tentative answer to the initial question of whether ownership matters for firm-level performance, measured now by TFP, is that there is no compelling evidence in favour of the widely held view that POEs are more productive than SOEs, but there is no evidence to the contrary either. Does this line of reasoning hold also when considering firms belonging to one sector at a time?

Moving away from cross-sectoral studies, several authors have analyzed POEs and SOEs within specific sectors, mainly within network industries, where state-ownership is still widespread. Kwoka (Reference Kwoka2005) starts from the premise that previous studies comparing the effect of ownership on firms’ performance and profitability had provided mixed evidence. By focusing on the electricity utilities sector in the United States, the author demonstrates how sectoral issues matter. SOEs in the distribution segment of the industry have an advantage over POEs due to the quality issues related to the product, while POEs exhibit higher cost-efficiency in the generation segment. Focusing on the electricity retail sector in Germany, Stiel et al. (Reference Stiel, Cullmann and Nieswand2018) do not find any evidence that SOEs or POEs are different in terms of firm-level productivity. Moving on to the transport sector, Backx et al. (Reference Backx, Carney and Gedajlovic2002) introduces the role of the percentage of state-ownership in shaping the effect on productivity. By analyzing the airline industry, the authors find, using panel data between 1993 and 1997 for a sample of international medium-large passenger airlines, that public ownership is detrimental to firm-level efficiency, and that this effect increases as the percentage of public ownership increases. Dalen and Gómez-Lobo (Reference Dalen and Gómez-Lobo2003) consider the Norwegian bus industry with a sample of 142 companies over the 1987–1997 period. Singling out the results related to public versus private ownership, the authors document that SOEs are more cost-efficient with respect to their private counterparts. A set of potential explanations are put forward, including economies of scale and scope (evidence of the latter provided, among others, by Growitsch and Wetzel (Reference Growitsch and Wetzel2009) for the railway sector), specifically related to the characteristics of the sector under consideration, and the fact that SOEs could be intrinsically more efficient than POEs. Roy and Yvrande-Billon (Reference Roy and Yvrande-Billon2007), using data from 1995 to 2002 for firms in the French urban transport networks, show that public administrations and semi-public enterprises exhibit lower levels of technical efficiency than fully private ones, with mixed ownership entities being the most inefficient. Boitani et al. (Reference Boitani, Nicolini and Scarpa2013), using data on urban local public transport in European cities between 1997 and 2006 find that majority and minority SOEs exhibit lower productivity than POEs, while also suggesting the importance of competitive tendering (see Section 3.2). Díaz and Charles (Reference Díaz and Charles2016) examine the provision of local public transport in France and find that private operators are more efficient than public ones, but less efficient than mixed ownership operators. Explicit consideration of the type of contractual agreements governing the type of contract makes ownership less relevant in explaining differences in efficiencies, as these are almost fully accounted for by contractual issues. Eller et al. (Reference Eller, Hartley and Medlock2011) compare the productivity of national and private international oil companies (NOCs and IOCs, respectively), finding that NOCs are less efficient than IOCs, this result primarily driven by internal institutional features, mainly related to differing objectives, within the firms.

Abbot and Cohen (Reference Abbot and Cohen2009) perform a literature review on the determinants of productivity in the water sector, focusing on studies up to the end of the 2000s. Focusing on their analysis of the impact of ownership, what emerges from existing literature is a blurred picture, with some studies documenting that SOEs are more efficient than POEs in this sector, while others find the opposite result. These mixed findings might be related, according to the authors, to two possible explanations. First, ownership matters less for firm-level productivity in the water sector than corporatization, which could be the main driver of improvements. Second, the water supply sector is characterized by monopoly features that allow productivity improvements only through regulation. Le Lannier and Porcher (Reference Le Lannier and Porcher2014) consider a cross section of the largest water services in France in 2009 and analyze the determinants of TFP for 117 firms. With respect to ownership, results point to a productivity advantage of SOEs, which are at least in part explained by the different approach to water conservation and differences in the water budget debt between private and public firms. Suárez-Varela et al. (Reference Suárez-Varela, García-Valiñas, González-Gómez and Picazo-Tadeo2017) use data for 2013 of firms providing Spanish municipalities with water services. Ownership is not neutral with respect to productivity and the analysis provides interesting results. Private providers tend to be technically more efficient that public ones when the management of the labour input is considered, a result that can be explained with the well-known labour hoarding hypothesis in state-owned firms in general, and when considering price efficiency. On the contrary, public ownership is associated with higher cost-efficiency. A different result is suggested by Walter et al. (Reference Walter, Cullmann, von Hirshhausen, Wand and Zschille2009) in their interesting meta-analysis of the benchmarking of the water distribution sector. Their analyses suggest that there is no evidence of a clear impact of public or private ownership, while the main driver of efficiency is the institutional setting.

A tentative conclusion from sectoral studies is that, in general, SOEs in network industries seem to be performing similarly or even outperforming POEs. One explanation for this result lies in the specificities of these sectors, which might favour vertical integration and where the role of competition is overall limited, pointing to mixed oligopoly. In turn, the theoretical models on mixed oligopoly mentioned in Section 2.7 overall do not suggest that productivity should systematically differ between public and private market players. The assumption in those models that SOE have a social welfare objective has implications for setting prices (and quality) but not for cost minimization by managerial teams. Hence the empirical results do not contradict most mixed oligopoly models.

However, taken as a whole, the surveyed literature provides varying evidence to the ownership-productivity nexus, which might go beyond the varying sectoral, geographic, and temporal coverage of the different studies. A growing body of research has been focusing on the relevance of a set of context conditions, external to the firm, the inclusion of which in the empirical analyses might in fact help disentangle the conflicting evidence surveyed here (see also Section 3.2).

3.2 The Role of Contextual Factors: Institutions, Regulation, and Politics

The most recent empirical literature has focused on the role of external factors in shaping the effect on firms’ activities and behaviour of ownership, including market structure and related reforms, regulation, political factors and, increasingly, the institutional setting in which firms operate (see Box 8). According to North (Reference North1991), “institutions are the humanly devised constraints that structure political, economic, and social interaction”. Indeed, all the human interactions take place within the choice set defined by the current institutional framework.

Box 8 Institutional quality as a mediating factor

Economic theory posits that effective institutions can play a crucial role in alleviating distortions linked to opportunistic behaviour and rent seeking, thereby tempering adverse government interference in SOEs. Accordingly, scholarly focus within the empirical literature has progressively pivoted towards investigating the impact of the institutional framework on the operations and outcomes of SOEs. A prevalent research hypothesis suggests that SOEs located in countries characterized by robust institutional quality are likely to exhibit improved managerial efficiency, leading to a diminished probability of markedly inferior firm performance, irrespective of the performance metric employed.

To test this proposition, we augment the baseline model (1) presented in Box 4 with two additional variables: the level of institutional quality ( $I{Q}_c$ ) and the interaction term between state ownership and institutional quality ( $SO{E}_i\cdot I{Q}_c$ ). The extended model is expressed as follows:

$Y_i=\alpha +\beta SO{E}_i+\theta I{Q}_c+\phi SO{E}_i\cdot I{Q}_c+{\sum }_{j=1}^J{\gamma }_j{Z}_{ij}+{\sum }_{c=1}^C{\delta }_c{W}_{ic}+{\sum }_{s=1}^S{\eta }_s{V}_{is}+{\epsilon }_i$(3)

Here, $\theta$ captures the influence of institutional quality on firm performance, irrespective of ownership structure, while $\phi$ measures the mediating effect of institutional quality on the relationship between state ownership and firm performance. This analytical approach enables the estimation of the marginal effect of state ownership, represented as:

$\frac{\partial Y_i}{\partial SO{E}_i}=\beta +\phi \cdot I{Q}_c$(4)

If $\beta <0$ and $\phi >0$ , it would signify that a positive effect of state ownership is contingent upon institutional quality ( $I{Q}_c$ ) exceeding a certain threshold, specifically, $I{Q}_c>-\frac{\beta }{\phi }$ .

The same rationale prompts the introduction of a triple interaction term in the model [2] presented in Box 5:

$Y_{it}=\alpha E{C}_{cst}+\theta I{Q}_{ct}+\lambda E{C}_{cst}\cdot I{Q}_{ct}+\beta SO{E}_i\cdot E{C}_{cst}+\phi SO{E}_i\cdot I{Q}_{ct}+\rho SO{E}_i\cdot E{C}_{cst}\cdot I{Q}_{ct}+{\sum }_{j=1}^J{\gamma }_j{Z}_{ijt}+{\sum }_{c=1}^C{\delta }_c{W}_{ict}+{\sum }_{s=1}^S{\eta }_s{V}_{ist}+u_i+f_t+{\epsilon }_{it}$(5)

In this context, the coefficient $\rho$ provides insights into whether heightened institutional development serves to mitigate potential performance differentials between SOEs and POEs, contingent upon specific environmental changes (events).

Commonly employed metrics for assessing institutional quality in empirical economic literature include, among many others: the Quality of Government index sourced from the International Country Risk Guide; regulatory quality (Kaufmann et al. Reference Kaufmann, Kraay and Mastruzzi2010); government effectiveness (Kaufmann et al. Reference Kaufmann, Kraay and Mastruzzi2010); political constraints (Henisz Reference Henisz2000, Reference Henisz2002).

Source: The authors, adapted from Borghi et al. (Reference Borghi, Del Bo and Florio2016) and Lazzarini and Musacchio (Reference Lazzarini and Musacchio2018).

Focusing on market structure, Asaftei et al. (Reference Asaftei, Kumbhakar and Mantescu2008) consider Romanian firms between 1995 and 2003 and verify whether state ownership is detrimental, or not, to productivity. Their results hint towards the importance of the environment in which firms operate since private firms indeed appear to exhibit higher efficiency levels than state-owned ones in highly competitive sectors but not in sectors with lower degrees of competition. Examining the role of regulatory reforms, Scheffler et al. (Reference Scheffler, Hartwig and Malina2013) examine the role of market reforms and regulation along with ownership on German bus transport companies. Their empirical results suggest that firms operating in regions where competitive tendering takes place are more efficient than those in regions that have not implemented these reforms, while ownership per se does not matter.

Another relevant strand of literature has focused on the impact of institutional quality and specific dimensions, such as for example corruption, on firm-level performance and how these factors influence the role of state versus private ownership. Driffield et al. (Reference Driffield, Mickiewicz and Temouri2013) introduce ownership as a factor influencing the impact of external institutional quality on firms’ performance and find that institutional reforms leading to increased country level competition influences mainly state-owned firms. Belloc (Reference Belloc2014) suggests that, contrary to the conventional wisdom, SOEs’ inefficiency is not due to state ownership per se but is likely related to other conditions, such as institutions, culture, legislation, and the degree of political competition. Dal Bo and Rossi (Reference Dal Bo and Rossi2007) present a theoretical model examining the link between corruption and firm-level efficiency between 1994 and 2001 and test its predictions with data on the efficiency of electricity distribution firms in Latin America. Their main result is that indeed national-level corruption is associated with firm-level inefficiency; specifically, firms in more corrupt countries tend to use more labour to produce a certain level of output. Additionally, the authors report that SOEs are, on average, less efficient than POEs, but do not test the mediating effect corruption could have on this relationship. This latter aspect is specifically tackled in Borghi et al. (Reference Borghi, Del Bo and Florio2016) and Castelnovo et al. (Reference Castelnovo, Del Bo and Florio2019), who examine the interplay between internal (ownership) and external factors (institutional quality) on firm-level TFP, respectively in the European electricity (2002–2009 period) and telecommunications sector (2007–2015 period). By focusing on the interaction between public ownership and different measures of institutional quality, the theoretical and empirical results suggest that SOEs in good institutional contexts may outperform POEs in terms of TFP. Underlying this empirical finding is the fact that a low-quality institutional environment is especially detrimental for SOEs. Recognizing the importance of different mandates and objectives for SOEs, other than efficiency, Estrin et al. (Reference Estrin, Meyer, Nielsen and Nielsen2016) consider the influence of external institutions at the national level on public enterprises’ internationalization strategies. Their empirical analysis, based on a sample of 306 matched POEs and SOEs from forty countries from the Worldscope database in 2010, suggests that ownership does not affect internationalization strategies. In fact, when institutional quality is similar, the behaviour of POEs and SOEs converges. National institutions, thought of as mechanisms shaping the ability of various shareholders of influencing SOEs internationalization decisions, exert an important effect on firm behaviour. Finally, Beuselinck et al. (Reference Beuselinck, Cao, Deloof and Xia2017) show that in the aftermath of the global financial crisis, SOEs underwent a comparatively modest decline in stock value contrasted with POEs but only in countries with sufficiently high institutional quality.

Political factors, including political connections that SOE managers might enjoy, are the subject matter of another set of articles. In an interesting meta-analysis on the interplay of ownership and political connection on firms’ strategies and performance, Tihanyi et al. (Reference Tihanyi, Aguilera, Heugens, van Essen, Sauerwald, Duran and Turturea2019) examine 210 studies from 139 countries. Focusing primarily on the management literature, the authors find that political connections have no consequences for firm-level performance, while public ownership has a small negative effect. The most interesting result, however, is the fact that both state ownership and political connections influence firms’ strategies in areas such as financial leverage, the level of R&D intensity, and internationalization. Further, the authors suggest that these strategies have a mediating effect on the ownership‒performance relation, thus suggesting relevant indirect channels through which being public or private and the interplay with political factors influence firm-level performance. Inoue (Reference Inoue2020) develops a conceptual model on the link between performance of SOEs and political incentives of managers. Using data from the water sector in Brazil with data on 2282 private and public enterprises between 2004 and 2014, the author shows how employment is higher and financial performance (measured by ROS) is worse, respectively, in election years, suggesting that SOE performance is contingent on political cycles and thus varies over time. Adua and Clark (Reference Adua and Clark2021) consider the impact of political partisanship on state-level utilities’ investment in energy efficiency in the United States in 2015. Examining the price setting behaviour of local district heating utilities in Sweden between 1998 and 2007, Biggar and Söderberg (Reference Biggar and Söderberg2020) show that political ideology at the municipal level has an impact on pricing, with left-wing local governments smoothing prices over time, while there is no clear systematic difference between private and public firms. Coulomb and Sangnier (Reference Coulomb and Sangnier2014) analyze the effect of being linked to France’s two presidential candidates in 2007 (in the sense of being part of their respective personal networks or firms being expected to benefit from their election) on firms’ abnormal returns. While the authors find strong evidence of an impact of political ties on returns, their results also show how political majorities does not operate primarily via state ownership of firms, since there is no detectable difference between private and state-owned firms. Qingyuan et al. (Reference Qingyuan, Lin and Xu2020) investigates the investment behaviours of European SOEs around national elections between 2001 and 2015. The authors find that SOEs increase their corporate investment during national election years and the effect is also stronger in countries with low institutional quality, more centralized political systems, and state-controlled banking systems.

Aguilera et al. (Reference Aguilera, Duran, Heugens, Sauerwald, Turturea and VanEssen2021) add another dimension and consider the interplay of political ideology and institutional quality on the nexus between ownership and productivity. By means of a meta-analysis of an international sample of 193 studies on the ownership‒performance relation, over the 1961–2003 period from 131 countries, the authors reach interesting results. First, the survey of the literature documents a small negative effect of state ownership on financial performance, with high heterogeneity both in terms of geography and size of the estimated effect. Second, both the government’s political ideology and institutions at the country level have a relevant effect on the ability of state owners of achieving both business and social goals. In countries with right-wing governments, SOEs are characterized by higher performance than in left-wing governed countries. Additionally, SOEs tend to perform better if the institutional quality is high, leading to low levels of political constraints for SOEs’ managers.

Soft-budget constraints, which have been frequently cited as characterizing the governance of SOEs (Kornai Reference Kornai1992), can also be seen as external institutional factors which might influence the performance of SOEs, as highlighted in Section 2.5. By examining initially post-Soviet countries in Eastern Europe, soft-budget constraints can be thought of in terms of, for example, state subsidies, soft taxation, non-performing loans, the accumulation of trade arrears between firms, and the build-up of wage arrears (Kornai Reference Kornai2001). The presence of various degrees of soft-budget constraints would thus help explain the estimated lower performance of SOEs with respect to their private counterparts, which was incorrectly attributed to ownership. Bertero and Rondi (Reference Bertero and Rondi2002) empirically address this issue by considering a regime change in Italian law, which brought a hardening of the budget constraint for SOEs in 1987. The empirical analysis on a panel of 150 Italian SOEs between 1977 to 1993 shows that under the soft-budget regime, investment is positively and significantly related to cash flow and that the regime shift leads to the disappearance of this relationship, suggesting that indeed a soft-budget constraint regime could lead to over investment and, consequently, to lower performance levels.

Taken together, these contributions on external factors highlight the importance these exert on firm-level productivity and show how institutional, political, and regulatory aspects influence the ownership‒productivity nexus. What emerges is that some of the differences in performance, productivity, and other measures of firm-level activity that have been reported in previous studies between private and public enterprises might, at least in part, be due to the impact of external contextual conditions, most notably related to institutional quality and political factors. When taking these aspects into account, the effect of ownership per se is weaker and, if present, is better understood in relation to the external institutional factors. By implication, institutions are probably the key omitted variable in most empirical analyses reported in Section 3.1.

3.3 Methodological Issues in the Empirical Analysis of SOE Performance

As shown in Table 1, earlier empirical literature on the economic performance of SOEs has predominantly relied upon a cross-sectional comparison of SOEs and POEs. Conventional ordinary least squares (OLS) approaches have been employed to estimate the regression models, typically incorporating additional regressors to account for potential confounding factors at both the firm and country levels. A notable limitation of these analyses arises from the frequent absence of comparable POEs within the same industry, country, and year for SOEs. In fact, Lazzarini and Musacchio (Reference Lazzarini and Musacchio2018) highlight that ‘only 15.6% of majority SOEs have private firms in the same country and sector’. Consequently, to address this issue, they resort to matching techniques to enhance the comparability of SOEs and POEs within their dataset, a methodological approach that could be the basis for future research on the subject.

Further analysis examine the influence of state ownership on performance by leveraging the experiences of SOEs that undergo privatization. However, in these instances, identifying the role of the state before and after privatizations poses a challenge. As highlighted by Dewenter and Malatesta (Reference Dewenter and Malatesta2001), alterations in firms’ performance post privatization may partly stem from changes induced by governments prior to privatization. This is due to the fact that governments often initiate restructuring efforts in firms before placing them on the private market.

Panel analyses and fixed effects models have been introduced to control for unobserved heterogeneity and the resulting issues of endogeneity, wherein some time-invariant component is possibly correlated with both the dependent variable (firm performance) and the explanatory variable (likelihood of being state-owned or selected for privatization over time). A limitation of these analyses is once again the reliance on ownership changes over time (e.g., privatizations), bringing forth selection bias challenges, as mentioned earlier. Ideally, a two-stage Heckman model could be employed to address selection bias by identifying, in the first stage, the probability of a firm transitioning from SOE to POE. Nevertheless, as observed by Goldeng et al. (Reference Goldeng, Grünfeld and Benito2008), this approach is not always feasible, depending on the coverage of available samples. Furthermore, as noted by Wolf (Reference Wolf2009), another limitation of fixed effects regressions in this context is their ability to quantify the impact of state ownership based only on firms that have changed ownership over time. Additionally, fixed effect models do not resolve potential endogeneity due to omitted time-variant variables. In such cases, an instrumental variable (IV) approach is necessary, as seen in Meyer and Pac (Reference Meyer and Pac2013), who employ a country political variable as an instrument, considering that the ideology of governing parties is expected to influence the likelihood of privatizations taking place.

Some analyses, such as Benito et al. (Reference Benito, Rygh and Lunnan2016), also resort to internal instruments using lagged variables (i.e., generalized method of moments or GMM) when suitable external instruments are not available (see also Roodman Reference Roodman2009). Through a meta-analytic regression analysis, Aguilera et al. (Reference Aguilera, Duran, Heugens, Sauerwald, Turturea and VanEssen2021) demonstrated that empirical studies controlling for endogeneity tend to reveal a positive relationship between state ownership and performance, underscoring the potential impact of reverse causality and selection effects on results of opposing signs.

Another potential limitation of the several empirical analyses is their focus on individual countries or markets. While, on the one hand, this focus is justified by the need to conduct the analysis within a sufficiently homogeneous context for comparing the performance of public and private enterprises, on the other hand, it complicates the generalization and external validity of empirical evidence.

Although more recent articles have attempted to introduce progressive methodological improvements to address the aforementioned limitations and enhance their evidence through robustness tests, some constraints are inherent to the subject matter and, consequently, to the type of available data. These limitations have contributed to fostering further developments in the empirical literature concerning SOEs, developments that we believe may yield additional insights in the near future.

Firstly, owing to the recent evolution of SOEs, an increasing number of analyses have begun to focus on continuous measures of state ownership rather than dichotomous variables. This shift has allowed for greater emphasis on the reality of hybrid organizations and public‒private partnerships (see Section 3.4). From a methodological perspective, this type of analysis opens up avenues for employing various econometric tools to account for the possibility of nonlinear effects of state ownership on firm performance. For instance, regression models can incorporate quadratic terms (to capture non-monotonic effects) or interaction terms (to capture moderation effects). Alternatively, researchers can employ threshold analyses to explore the presence of critical points or breakpoints in the relationship between state ownership and firm performance.

Secondly, rather than fixating on the comparison between SOEs and POEs, the research question is evolving, particularly considering that SOEs constitute an integral part of our economic systems. Consequently, the focus is increasingly shifting towards asking, ‘What are the conditions that promote the effectiveness and efficiency of SOEs?’ As detailed in Section 3.5, this shift prompts attention towards comparing SOEs with different corporate governance characteristics rather than emphasizing the distinctions between SOEs and POEs. It is evident that corporate governance stands out as one of the prime candidates among the potential channels through which variations in the institutional quality at the country level affect the performance of SOEs. Methodologically, this type of analysis broadens the spectrum of identification strategies because a single enterprise, while remaining state-owned since its establishment, may still undergo significant changes in terms of corporate governance.

Finally, there is a noticeable growth in the empirical literature that, instead of dwelling on the performance of individual firms, emphasizes the characteristics of firms targeted by SOEs through M&A deals (see, for example: Clò et al. Reference Clò, Fiorio and Florio2017b, Karolyi and Liao Reference Karolyi and Liao2017) or the foreign direct investments made by SOEs (Grøgaard et al. Reference Grøgaard, Rygh and Benito2019). Other research focuses on the spillover effects that SOEs may have on the aggregate economy. The final part (Section 3.6) provides a brief overview of some recent developments regarding the aggregate effects of state ownership on the surrounding socio-economic environment. Once again, the new focus of these analyses provides interesting methodological insights, as identification strategies can concentrate on observed differences at the aggregate level before and after the entry of SOEs into the target market or as a consequence of the intensification (or relaxation) of their activities. This type of analysis is well-suited for difference-in-differences methodologies, which in recent years have seen significant contributions and advancements. For instance, it is our opinion that difference-in-differences analyses with heterogeneous treatment (de Chaisemartin and D’Haultfoeuille Reference de Chaisemartin and D’Haultfoeuille2020, Reference de Chaisemartin and D’Haultfoeuille2023) can have important developments in this context since they provide robust estimators when the treatment effect is not constant between groups or over time. The review of this section indeed highlights how state-ownership can have profoundly differentiated effects depending on the reference cohort.

3.4 The Performance Effect of State Ownership: A Summary

From a bird’s eye overview of Table 1, when surveying a selection of papers comparing private and public enterprises on the basis of financial and accounting measures of performance and profitability, and then total factor productivity, no crystal-clear conclusion stands out. At closer inspection, however, the following key points emerge.

A first element is related to the choice of the dependent variable, with earlier works focusing more on performance, both in terms of profitability and productivity, while an increase in the use of ‘other’ variables has emerged over time. The impact of ownership is increasingly examined by considering environmental outcomes, the behaviour of firms on the M&A market, and on innovation, to mention the most frequent choices.

A second consideration is that, on average, papers that consider profitability indicators report more frequently a statistically distinguishable from zero negative estimated coefficient related to state ownership, while papers focusing on ‘other’ variables as described above tend to report on average a positive coefficient. This latter result might suggest which objectives and areas could be targets of further research and eventually policy intervention.

In general, however, the variety of results could be explained by the different estimation methodologies employed (see Section 3.3) and by the selection of the sample in terms of sectoral or geographic composition. In fact, conclusions vary greatly by sector, suggesting that sector-specific characteristics might have a relevant effect on the relationship between ownership and firm-level performance. What should be noted is that, over time, the management and environment in which SOEs operate has changed, with the introduction of improved management practices in SOEs, in the adoption of stringent budget constraints in the public sector in general, in the effect of aggregate reforms (e.g., liberalization, corporatizations, regulation) that might benefit SOEs comparatively more than private-owned enterprises, and these reforms have been adopted at different paces and degrees in different countries. Finally, conventional wisdom viewing private firms as naturally more profitable and more productive than publicly owned entities is challenged, and in some instances reversed, when aggregate contextual conditions are considered, most notably concerning institutional quality and aggregate political factors.

Ownership matters for performance and aggregate economic activity, but is it simply the distinction between private and state ownership that we should focus on? The answer is probably negative, as hinted above, when a distinction between the behaviour of majority and minority state ownership was made and by considering that, over time (see Table 1), research has moved from a dichotomous distinction between fully private or full public enterprises to considering the continuum between the two. Several scholars (Koppell Reference Koppell2003, Oum et al. Reference Oum, Adler and Yu2006, Christensen and Lægreid Reference Christensen and Lægreid2011, Doherty et al. Reference Doherty, Haugh and Lyon2014), in fact, have pointed towards the importance of the hybrid nature of several SOEs for understanding their behaviour, performance, and definition of their mission (Sorrentino Reference Sorrentino, Bernier, Florio and Bance2020). Ownership is increasingly seen not as a dichotomous state, but as a continuum, calling for the analysis of hybrid organizations. Bruton et al. (Reference Bruton, Peng, Ahlstrom, Stan and Xu2015) analyze four sectors across twenty-three states and provide a case-study analysis of thirty-six firms, suggesting that the emergence of hybrid organizations suggests caution when classifying firms as private or state-owned and basing performance comparisons on this dichotomy. Chen et al. (Reference Chen, Musacchio and Li2018) consider how the hybrid nature of many SOEs affects their behaviour on international markets, by adopting a principal-principal perspective. The authors argue that the private‒public agency problem present in many state-owned (at different levels) enterprises, is different from the same problem when all involved parties are private, and document the emergence of conflicts between large government and private block holders. These conflicts, in turn, affect managerial decisions, especially when these organizations enter the foreign M&A market. The definition itself of a hybrid organization is difficult to define, as this umbrella term is used for a variety of configurations, including truly mixed enterprises, public‒private partnerships, social entrepreneurship organizations, government-sponsored enterprises, and several other hybrid configurations (Vining and Weimer Reference Vining and Weimer2016). Vining and Weimer (Reference Vining and Weimer2017) show how the institutional arrangements for managing these hybrid organizations vary greatly across countries, sectors, and levels of government, suggesting that these enterprises are intrinsically complex in nature. Their research suggests that these types of organizations are not new, as Adam Smith had analyzed the East India Company and highlighted the issues related to its hybrid ownership nature. Vining et al. (Reference Vining, Boardman and Moore2014) analyze hybrids at the local government level, focusing on enterprises providing local public goods, and stress how they should be driven by the goal of maximizing local social welfare.

3.5 Corporate Governance

A relatively recent body of literature focuses on corporate governance and analyzes the role of the board of directors in managing SOEs and addressing their weaknesses. Indeed, while board composition has been a highly studied topic as far as private firms are concerned, it has received little attention in the context of SOEs until the beginning of the century, when many contemporary SOEs have been deeply reformed in order to ensure greater transparency and accountability, and better compliance with ethical and deontological requirements (Picot et al. Reference Picot, Florio, Grove and Kranz2015, Bacchiocchi et al. Reference Bacchiocchi, Ferraris, Florio and Vandone2019, Bernier et al. Reference Bernier, Florio and Bance2020b).

The theoretical framework is agency theory and the effectiveness of the board as an internal monitoring control to avoid technical and allocative inefficiencies and ultimately ensure wealth maximization to the shareholder (Jensen and Meckling, Reference Jensen and Meckling1976). This framework is fully compatible with the one for private companies, with a similar emphasis on the role of the board and issues related to disclosure, transparency, and accountability. What is different is that SOEs may face two other distinct governance challenges: (1) they may suffer from undue hands-on and politically motivated ownership interference, which may prevent the governance of the SOE to autonomously define the goals and the strategy of the institutions without being influenced by the private agenda of politicians that may use SOEs to extract rents (Putniņš Reference Putniņš, Bernier, Florio and Bance2020, Parker Reference Parker2021); (2) they may suffer from a lack of any oversight due to totally passive or distant ownership by the state, which may weaken the incentives of SOEs and their staff to perform in the best interest of the enterprise and the general public, and raise the likelihood of self-serving behaviour by corporate insiders (Grossi et al. Reference Grossi, Papenfuß and Tremblay2015).

As outlined in Table 2, our selection of empirical papers identified four analyses on the corporate governance of SOEs.

Bozec and Dia (Reference Bozec and Dia2007) investigate the board‒performance relationship for a sample of cross-industry Canadian SOEs. The authors scrutinize the impact of independence on the firms’ technical efficiency measured using non-parametric data envelopment analysis and find a positive relation, meaning that SOEs with a larger proportion of outsiders on their board and fewer public servants are more likely to be associated with a higher performance. Similarly, Menozzi et al. (Reference Menozzi, Gutiérrez Urtiaga and Vannoni2012) focus on board composition, defined in terms of outsiders as opposed to insiders and, among the outsiders, distinguishing between politically connected directors versus independent directors. They find that politically connected directors (i.e., those who are currently holding a seat in the parliament or in the municipal, provincial, or regional government or have held one in the past) exert a significant inflationary effect on employment while the opposite holds for the independent directors, who exercise a disciplinary role on labour demand and prevent SOEs to be over-employed. The composition of the board is also analyzed in Grosman et al. (Reference Grosman, Aguilera and Wright2019), who investigate the role of independent directors in mitigating blockholders’ mismanagement of funds and appropriation of firm wealth at the expenses of minority shareholders and creditors (so-called ‘blockholder appropriation’). The authors focus on Russia, which only recently adopted corporate governance practice such as the presence of independent directors, and find that the presence of independent directors is associated with a reduction of blockholder appropriation, and that this is more accentuated in SOEs than in private firms. The authors attribute the difference in the influence of independent directors across ownership types to SOEs going through a much more thorough selection process in the nomination of their independent directors, and to the declared effort to curtail corruption.

Sidki et al. (Reference Sidki, Boerger and Boll2024) move away from the internal versus outsider/independent trade-off and focus on skills and competences of politically connected board directors. The authors analyze the biographical information (i.e., education, management experience, industry experience) of board members of a sample of state-owned utility German companies and find statistically significant differences between politically connected and non-political members, especially with regards to management and industry expertise. However, despite the lower level of business competence among politically connected board members, the estimation results indicate no relevant influence of these factors on the financial performance of the analyzed utility companies. According to the authors, it may be the case that the influence of supervisory boards among SOEs is relatively weak in the German framework, where decision-making power lies in bodies outside the typical construct of corporate governance such as municipal councils.

By highlighting that board composition is positively related to firm efficiency, results from this recent body of literature are quite consistent with regulatory reforms adopted by several countries in line with the Principles of Corporate Governance for State-Owned Enterprises set by the OECD in 2015 (OECD 2015a,b)Footnote ⁹, whose main message is that the board is chiefly responsible for setting the corporate strategy and major plans of action, monitoring the implementation of the strategy and the achievement of the goals, ensuring the integrity of the corporation’s accounting and financial reporting system, safeguarding independent audit, control, and compliance (Iwu-Egwuonwu Reference Iwu-Egwuonwu2010, Pargendler et al. Reference Pargendler, Musacchio and Lazzarini2013, Tremml Reference Tremml2019, Vining and Laurin Reference Vining, Laurin, Bernier, Florio and Bance2020).

This issue is strictly related to the problem of time consistency, which arises whenever the effectiveness of a policy today depends on the credibility of the commitment to implement that policy in the future, and which requires a credible obligation of the state to guarantee the autonomy of SOEs, that in turn can be improved by taking all the precautions and providing all the incentives that are needed to ensure that the SOEs’ mission is sufficiently internalized by the managers (OECD 2020). In turn, a strong and well-defined public mission can attract staff with a specific intrinsic public service motivation and altruism. As highlighted in Section 2.8, this stream of literature is mainly theoretical (Francois and Vlassopoulos Reference Francois and Vlassopoulos2008, Polidori and Teobaldelli Reference Polidori and Teobaldelli2013, Willner and Grönblom Reference Willner, Grönblom, Bernier, Florio and Bance2020), while empirical research is still scant and not specifically related to SOEs, and offers few and sometimes contradictory evidence regarding these effects (Papenfuß and Keppeler Reference Papenfuß and Keppeler2020, Jacobsen Reference Jacobsen2021).

Alternatively, empirical analyses may also treat corporate governance as the dependent variable and assess how it is influenced by state ownership. This is the focus of analysis in Borisova et al. (Reference Borisova, Brockman, Salas and Zagorchev2012), which concludes that government ownership is generally detrimental to the corporate governance of European firms, particularly in the context of civil law systems. However, more empirical research is needed to test to what extent the guidelines on corporate governance are actually followed and their effects. It would be crucial to assess whether SOEs are progressively aligning with international best standards and to investigate whether these developments, if any, have the intended effects on the effectiveness of SOEs in achieving their objectives. Conducting such empirical studies will not only contribute to a deeper understanding of the evolving landscape of SOEs but also provide insights into the impact of their adherence to international standards, their public involvement, and the awareness about its mission and management from citizens, who ultimately ‘own’ SOEs.

3.6 The Aggregate Effect of State Ownership

As documented above, SOEs are still active players in the economy, producing goods and services, going abroad, and performing deals in the M&A market. They also play a role in innovation and some societal goals. Ownership matters for performance and productivity, although the actual net effect is different across sectors, countries, and institutional settings, and highly dependent upon empirical methods. Against this backdrop, what is the overall effect of public ownership at the aggregate level?

Table 3 presents a selection of empirical papers on the impact of SOEs on the aggregate economy.

Earlier literature focused on employment effects, highlighting the role of SOEs as employers, especially in declining industries or in times of crises (see e.g., Svejnar (Reference Svejnar1996) for Eastern Europe and Bognetti (Reference Bognetti, Bernier, Florio and Bance2020) for a historical perspective). However, Feldmann (Reference Feldmann2006), in an analysis of the relation between government size and unemployment, finds that the larger the share of public investment in SOEs, the greater the level of unemployment, especially for some categories of workers, casting doubts on this channel of aggregate impact of SOEs. Szarzec et al. (Reference Szarzec, Dombi and Matuszak2021) analyze the link between the economic weight of SOEs, measured by the share of state-owned total assets, operating revenues, and employment of large non-financial enterprises in each country, and growth in Europe between 2007 and 2016. Their main result is that the presence of SOEs is not positive or negative per se on economic growth, but is mediated by institutional quality, with a high percentage of SOEs having beneficial effects on growth only when the institutional setting is good. A possible mechanism through which SOEs may positively impact aggregate economic performance and growth is related to their role in knowledge creation and as vehicles of social objectives. Evidence on the former mechanism is provided, for example, by Antonelli et al. (Reference Antonelli, Amidei and Fassio2014) who consider the case of economic growth in Italy between 1950 and 1994 and show that R&D that was conducted in SOEs had a greater impact on country level TFP growth than private R&D. With respect to the latter, instead, Avsar et al. (Reference Avsar, Karayalcin and Ulubasoglu2013) document the existence of a U-shaped relationship between the size of the SOE sector, measured by the share of the SOE sector production in GDP, and national inequality (Gini coefficient) in the 1970–2004 period worldwide. For low levels of inequality, the correlation between inequality and the size of SOEs is positive, while for high levels of inequality the reverse is true. The authors also analyze the role of political ideology in shaping this relationship, providing additional evidence of the importance of context conditions when considering the impact of ownership on economic activities in general.