Supplier Participation in Latvian Knowledge-Services Procurement

Public procurement is one of several instruments through which the state allocates resources to external actors, alongside grants, subsidies and in-house service delivery. What distinguishes procurement from alternatives is that the state is purchasing a specific deliverable under a contract: the desired output must be specifiable, performance must be measurable, and contractual enforcement must be feasible. When these conditions hold, competitive tendering is used to elicit price and quality offers under rules designed to ensure transparency, equal treatment and accountability.

1. NRNVZV

   European Commission. (n.d.). Public procurement. https://single-market-economy.ec.europa.eu/single-market/public-procurement_en

The theoretical justification for why the state procures rather than produces in-house rests on transaction cost reasoning. Coase (1937) established that the boundary between internal organisation and market exchange is determined by the relative costs of coordinating activity within a hierarchy versus contracting for it externally. Williamson (1979, 1985) developed this insight into a governance framework in which the choice depends on three properties of the transaction: the specificity of the assets involved, the degree of uncertainty surrounding performance, and the frequency with which the transaction recurs. When asset specificity is low and outputs are reasonably measurable, market contracting, procurement offers efficiency advantages over hierarchical production because competitive selection disciplines price and incentivises quality. When asset specificity is high, uncertainty is substantial, or performance is difficult to verify, the transaction costs of market contracting rise and hierarchical or hybrid governance becomes comparatively attractive.

1. 5F5HP2

   Coase, R. H. (1937). The nature of the firm. Economica, 4(16), 386–405. https://doi.org/10.1111/j.1468-0335.1937.tb00002.x
2. ATATRE

   Williamson, O. E. (1985). The economic institutions of capitalism. Free Press.

Bajari and Tadelis (2001) apply this logic directly to public procurement, demonstrating that the choice between competitive fixed-price tendering and negotiated cost-plus contracting reflects the same trade-off: competitive procedures are efficient when the scope of work can be well specified ex ante, but become costly when projects are complex and require adaptation during execution, because rigid contracts cannot accommodate changes that were unforeseeable at the tender stage. This distinction is consequential for the present thesis. Knowledge-intensive services, such as consulting, IT development, analytical work frequently combine moderate asset specificity with high uncertainty about both the optimal solution and the criteria by which quality should be judged. They sit, in Williamson's terms, in a zone where market contracting is institutionally mandated by procurement law but where the transaction properties push toward governance forms that procurement procedures do not easily accommodate. The costs that suppliers perceive when deciding whether to bid are, in part, a reflection of this mismatch between the governance form imposed and the governance form the transaction's properties would suggest. This governance logic applies to any organisation facing a make-or-buy decision, whether private or public. A private firm choosing between hiring an in-house team and contracting a consultant faces the same transaction cost calculus. What distinguishes public procurement is not the economic logic of the choice but the institutional constraints imposed on how the contracting is conducted.

1. B357FL

   Bajari, P., & Tadelis, S. (2001). Incentives versus transaction costs: A theory of procurement contracts. The RAND Journal of Economics, 32(3), 387–407. https://doi.org/10.2307/2696361

Unlike private purchasing public procurement must simultaneously pursue economic efficiency and public legitimacy. Transparency, equal treatment, accountability and safeguards against favouritism are not optional features but constitutive rules of the market itself. In the European Union, procurement law has additionally served internal market objectives by opening contracts to cross-border competition, while contemporary policy debates increasingly emphasise multi-dimensional values, such as quality, sustainability, innovation and social benefit, alongside cost and compliance. This dual mandate creates a characteristic tension: the rules that protect legitimacy also impose costs on participants, and the balance between protection and accessibility is itself a design variable that affects who enters the market and on what terms.

From an institutional economics perspective, markets are not natural phenomena but structures produced and sustained by rules, monitoring and enforcement mechanisms. Procurement is a particularly clear case of this principle. A contracting authority that designs a tender is simultaneously designing a market: procedure choice, documentation requirements, submission deadlines, evaluation criteria structure, award weightings and remedies together constitute the institutional architecture within which suppliers form expectations about the cost of participation, the intensity of competition and the probability that effort will translate into a winning outcome. These design choices are not neutral administrative decisions. Each creates incentives that shape supplier behaviour before a single bid is submitted.

1. WJ6UWY

   North, D. C. (1990). Institutions, institutional change and economic performance. Cambridge University Press. https://doi.org/10.1017/CBO9780511808678

Procurement procedures can therefore be read as governance arrangements designed to address a layered principal-agent problem. Citizens delegate spending authority to public bodies, which then contract with suppliers under asymmetric information about costs and quality. Formal procedures act as commitment devices: they constrain the buyer's discretion, making the process predictable and contestable, but they also limit the buyer's ability to adapt requirements when needs evolve or when market conditions differ from what the tender design anticipated. The rigidity that protects against favouritism simultaneously raises compliance costs for participants and reduces the buyer's capacity to respond to information revealed during the process. This is not a design flaw but a structural trade-off inherent in rule-bound procurement, and the balance point differs across procedure types, contract values and market segments.

A practical implication of this framing is that procurement outcomes reflect two distinct sources of variation. The first is market structure: the size, capabilities and competitive dynamics of the available supplier pool in a given sector and geography. The second is institutional design, how a specific tender is structured, what it demands from participants, how it evaluates offers and what remedies it provides. The same sector can produce very different participation outcomes under different procedural configurations. Detailed documentation requirements combined with short submission deadlines, for instance, may systematically deter smaller or specialised firms even when they are technically capable, consistent with evidence that participation barriers disproportionately affect firms with limited bidding capacity. Conversely, simplified procedures with clear criteria and adequate timelines can attract broader participation from the same supplier pool. The variation is not random, it is produced by identifiable design features that suppliers observe and respond to.

1. 9KVSWB

   Flynn, A. (2025). Research on SME involvement in public procurement: A review, critique and conceptual framework. Journal of Purchasing and Supply Management

Viewing procurement as a designed institutional environment clarifies why participation costs arise even when the technical specification is formally available to all potential suppliers. The rule-bound nature of competitive tendering, standardised documentation formats, sequential procedural stages, fixed deadlines, formal communication channels and the possibility of rejection at multiple evaluation gates, creates direct formal burdens in the form of time, administrative effort and compliance risk. These burdens are not incidental to the process but are produced by the very features that ensure transparency and equal treatment. Beyond formal burden, evaluation models that incorporate qualitative scoring introduce a second type of cost: uncertainty about how evaluators will interpret and weight proposals. When criteria are broad, scoring rubrics are underspecified, or the balance between price and quality is unclear, suppliers cannot reliably predict whether a strong proposal will be recognised as such. This uncertainty reduces the expected return on bid investment even for capable firms. A third cost type emerges where procedure design permits negotiation, dialogue or significant buyer discretion. In such settings, suppliers may perceive that the stated evaluation rules do not fully govern the actual selection, raising concerns about process integrity that are difficult to verify from outside. These three cost channels: (1) formal burden, (2) evaluation uncertainty and (3) integrity risk, are not exhaustive, but they capture the primary mechanisms through which institutional design features are translated into the perceived transaction costs that shape the bid/no-bid decision. The theoretical development of these components and their integration into the supplier's decision calculus follows in subsequent sections of this chapter.

This institutional lens clarifies what reform means analytically. A reform is not merely a change in legal text, it is a change in the incentive and information environment that suppliers face when deciding whether to bid. If reform reduces the perceived costs of participation, increases evaluation predictability, or improves the match between requirements and market capabilities, entry should rise. If it increases documentary burden, adds ambiguity, or raises perceived procedural risk, entry may fall. Whether any given reform succeeds depends on whether the design features it changes are the ones that currently bind. This is the bridge between legal changes and competition outcomes, and it motivates the identification of specific tender-level features that suppliers can observe before deciding whether to participate.

Section 1.1 established that procurement is a designed institutional environment whose rules shape participation incentives. The question that follows is how individual suppliers respond to those rules. This thesis models the supplier as a rational economic agent making a participation decision under uncertainty: faced with a tender, the firm assesses whether the expected return from bidding exceeds the cost of participation, given its available alternatives. This is not a claim that every supplier performs an explicit calculation, it is an analytical device that identifies which variables matter for the participation decision and how changes in the institutional environment shift behaviour. The framework therefore requires two theoretical inputs. The first is a model of the information problem that makes participation costly in the first place, why quality is difficult to observe, why demonstrating it is expensive, and why these difficulties vary across market segments. The second is a model of the supplier's decision rule, how expected benefits, participation costs and outside options combine to produce a bid or an abstention. The remainder of this chapter develops both.

The starting point is that participation costs in procurement are not primarily administrative but informational. If buyers could costlessly observe supplier quality, competitive tendering would require little more than a price quote: firms would submit offers, the buyer would select the best combination of price and verified quality, and the cost of participating would be negligible. Participation becomes expensive precisely because quality is not directly observable. Suppliers must invest in producing evidence of their capability such as documentation, qualifications, methodology descriptions, reference portfolios and even specific qualitative proposals. And buyers must invest in designing evaluation architectures capable of interpreting that evidence. The costs on both sides are generated by the information gap between what matters (actual delivery quality) and what is observable at the point of selection (representations of future delivery). This framing positions information asymmetry not as one theoretical lens among several but as the foundational problem from which all participation costs derive. The theoretical tools for analysing how markets manage this problem, and how they fail, come from information economics, beginning with Akerlof's (1970) analysis of market failure under quality uncertainty.

The fundamental economic problem that procurement institutions are designed to solve is quality uncertainty under asymmetric information. Akerlof's 1970 paper "The Market for 'Lemons': Quality Uncertainty and the Market Mechanism," published in the Quarterly Journal of Economics, established one of the most influential results in modern economics, that information asymmetry between buyers and sellers does not merely produce inefficient pricing but can cause markets to unravel entirely. Akerlof in the paper established that the buyer's rational response to the uncertainty is to price the expected average quality. This undercompensates sellers of genuinely high-quality goods relative to their true value. Crucially, Akerlof demonstrated that this is not a static misprice but a self-reinforcing dynamic: as higher-quality sellers find the average price insufficient and withdraw, the composition of the remaining market shifts downward, which validates and deepens the buyer's original caution. The result, in the limiting case, is a market populated only by the lowest-quality participants or no market at all. The paper's influence has been immense precisely because the mechanism it identifies is not specific to any one market, used cars in his initial analysis. It applies to any transaction where one party holds private information about quality that the other cannot verify before exchange, a structure that recurs across insurance, labour markets, financial products and, as this thesis argues, public procurement.

1. CNP7MV

   Akerlof, G. A. (1970). The market for 'lemons': Quality uncertainty and the market mechanism. The Quarterly Journal of Economics, 84(3), 488–500. https://doi.org/10.2307/1879431

The application of information economics to public procurement is well established. Laffont and Tirole (1993), in what remains the standard principal-agent treatment of procurement and regulation, formalised the problem as one in which the government cannot observe the supplier's true cost type before contracting, creating adverse selection, nor fully monitor the supplier's effort after contracting, creating moral hazard. Their framework demonstrated that optimal procurement contract design is fundamentally a response to these information problems and that the structure of evaluation criteria and administrative procedures are all mechanisms for managing what the buyer cannot directly observe. The information problem in procurement is, however, structurally distinct from Akerlof's original formulation in one important respect. In the used car market, the buyer responds to quality uncertainty by adjusting the price downward. In procurement, the price is proposed by the supplier, and the buyer's task is not to set a price but to assess whether the quality claimed behind a given price is credible, on the basis of documentation, methodology descriptions, reference portfolios and proposed technical approaches that are, by nature, representations of future delivery rather than evidence of past performance. This distinction matters because it shifts the analytical focus from price formation to quality verification, and specifically to the mechanisms through which quality can be credibly communicated under asymmetric information. Information economics offers two complementary answers: signaling, where the informed party invests in making quality visible, and screening, where the uninformed party designs conditions that induce self-revelation. Both operate simultaneously in procurement, and their interaction defines the cost structure that suppliers face when deciding whether to bid.

1. QV6KDD

   Laffont, J.-J., & Tirole, J. (1993). A theory of incentives in procurement and regulation. MIT Press.
2. A7L54W

   Khoman, S. (2017). Asymmetric information: A case study in potential public procurement pitfalls.

Signaling theory originates in Spence's 1973 paper "Job Market Signaling," published in the Quarterly Journal of Economics. While the paper used the labour market as its empirical setting, its contribution is a general model of how informed parties in any market can credibly communicate unobservable attributes such as productivity to uninformed parties through costly observable actions. The central insight is that a signal is informative only if its cost is systematically related to the underlying attribute it purports to reveal, specifically, if producing the signal is cheaper for those who genuinely possess the valued quality than for those who do not. Under this condition, different types rationally choose different signal levels, and the uninformed party can infer the unobservable attribute from the observable choice. This is the separating equilibrium: the signal's cost structure, not its intrinsic content, produces the information transfer. Spence established two further properties that define the model's scope. First, the signal need not be directly productive, it functions through its sorting capacity, not through any contribution to the quality of the transaction itself. Second, when the cost-quality correlation weakens or disappears (when low-quality actors can produce the signal at comparable cost) separation fails, and both types are pooled together, and the market reverts to the adverse selection condition that Akerlof identified. The model therefore specifies both the conditions under which information problems are resolvable and the conditions under which they are not, making it a framework for diagnosing market outcomes rather than merely describing them.

1. QSUQDJ

   Spence, M. (1973). Job market signaling. The Quarterly Journal of Economics, 87(3), 355–374. https://doi.org/10.2307/1882010

The procurement analogue is direct. When a contracting authority publishes a tender with qualification thresholds, documentation requirements and evaluation criteria, it is defining what signal a supplier must produce to be considered. The supplier's bid that is encompassing technical proposals, methodology descriptions, evidence of past performance, team qualifications and compliance documentation. The bit requirements are not merely an administrative submission but a costly signal to be assembled to demonstrate unobservable quality to a buyer who cannot verify it directly. The signaling logic holds if the cost of producing a credible bid is systematically related to actual supplier capability: a firm with genuine expertise, relevant experience and established processes can assemble the required evidence at lower cost than a firm that lacks these attributes but attempts to present itself as equivalent. Under this condition, the tender's requirements function as a separating mechanism, capable suppliers find the investment worthwhile relative to the expected contract value, while less capable suppliers self-select out.

M. Spence in his 2002 Nobel lecture "Signaling in Retrospect and the Informational Structure of Markets," published in the American Economic Review, further developed several results that bear directly on the procurement problem. The most consequential for this thesis is the demonstration that signaling equilibria are not unique. Multiple separating equilibria can coexist, each internally consistent, each sustained by self-confirming beliefs. The mechanism is straightforward: the buyer's beliefs about what signal level indicates quality determine the requirements they set, those requirements determine the returns suppliers face, the returns determine which suppliers participate, and the observed participant pool confirms the buyer's original beliefs. Because the beliefs create the conditions that validate them, any internally consistent belief-requirement pair constitutes an equilibrium, including pairs where requirements are set substantially higher than the minimum necessary for effective quality separation. Spence showed that there is no inherent market mechanism that selects the efficient equilibrium from among the feasible ones. Applied to procurement, this means that a contracting authority requiring extensive documentation, multiple certifications and detailed methodology proposals may be sustaining a self-confirming equilibrium in which only suppliers who can afford these signal levels participate, the buyer observes adequate quality from this pool, and concludes that the requirements are appropriately calibrated, but never observing the capable suppliers who were deterred by signal costs that exceeded the minimum necessary for separation. The equilibrium is stable and internally rational, but it is not efficient: a lower-cost signaling equilibrium could achieve equivalent quality separation with broader participation.

1. F66NMM

   Spence, M. (2002). Signaling in retrospect and the informational structure of markets. American Economic Review, 92(3), 434–459. https://doi.org/10.1257/00028280260136200

Signaling addresses how the informed party communicates quality. The complementary problem is how the uninformed party elicits information it cannot directly observe. This is the subject of screening theory, originally formalised by Rothschild and Stiglitz in their 1976 paper "Equilibrium in Competitive Insurance Markets: An Essay on the Economics of Imperfect Information," published in the Quarterly Journal of Economics. Where Spence's model gives the initiative to the informed party who chooses what signal to produce, Rothschild and Stiglitz shift agency to the uninformed party who designs a menu of options structured so that different types, by choosing differently, reveal their private information. The foundational insight is that when the uninformed party cannot distinguish types directly, it can design contracts or conditions whose terms are differentially attractive to different types, so that the act of choosing itself constitutes disclosure. In their original formulation, the uninformed party offers a set of options that vary along dimensions which matter differently to different types, structured so that each type, acting in its own interest, selects the option that reveals its private information. The menu does not require the uninformed party to know anyone's type in advance, the structure of the choices produces the separation through self-selection. Rothschild and Stiglitz further showed that pooling equilibria, where both types select the same option, are unstable in competitive markets: a rival can always design an offer that attracts only the more profitable type, breaking the pool. The implication is that in competitive screening markets, separation is the only stable outcome, but the cost of achieving it falls disproportionately on the better type, who must accept less favourable terms than they would receive under full information.

1. 2J8EG5

   Rothschild, M., & Stiglitz, J. E. (1976). Equilibrium in competitive insurance markets: An essay on the economics of imperfect information. The Quarterly Journal of Economics, 90(4), 629–649. https://doi.org/10.2307/1885326

In procurement, the screening party is the contracting authority. The "menu" it designs is not a set of alternative contracts offered simultaneously but the architecture of a single tender: procedure type, qualification thresholds, documentation requirements, evaluation criteria structure and weightings, and the terms of the resulting contract. Each of these design choices shapes which supplier types find participation attractive and which do not. A tender that requires extensive past performance references, detailed methodology proposals, and team CVs with specific certifications is, in screening terms, a menu designed to be differentially costly across supplier types, thus firms with established track records and deep expert benches can satisfy these requirements from existing material, while firms without this portfolio must construct it from scratch or decline to participate. The authority may not conceptualise its tender design as a screening menu, but the economic effect is the same: the structure of requirements determines which types find participation worthwhile and which self-select out, without the buyer needing to observe capability directly.

In practice, signaling and screening do not operate in isolation but simultaneously, and the market outcome in any given tender reflects their interaction. The standard framing assigns screening to the buyer and signaling to the supplier, but the information asymmetry in procurement is two-sided, and both parties engage in information-revealing and information-interpreting behaviour. The buyer designs the formal screening architecture but also signals its own attributes through tender design. Suppliers, reciprocally, produce costly signals through their bid documentation, but they also read the buyer's signals, inferring from observable tender features whether this contracting authority is worth the investment of participation. This is not screening in the Rothschild and Stiglitz sense, since the supplier does not design conditions that induce the buyer to reveal its type, but rather Bayesian updating: the supplier observes the tender's characteristics and revises its beliefs about the buyer's competence and trustworthiness accordingly. The complexity of procurement markets lies in this mutual assessment under mutual uncertainty. However, the two sides are not symmetric in what this thesis investigates. The buyer's design choices are treated as the institutional environment, observable, given, and exogenous to any individual supplier's decision. The supplier's response to that environment is the object of analysis. The central question is how a supplier, facing a specific screening architecture and forms beliefs about the buyer's type from observable signals, translates that assessment into a participation decision. The theoretical development that follows therefore takes the buyer's design as fixed and models the supplier's decision calculus: under what cost and probability conditions does a rational supplier choose to produce the signal the tender demands, and under what conditions does it decline?

1. 58WZPY

   Harsanyi, J. C. (1967). Games with incomplete information played by 'Bayesian' players. Management Science, 14(3), 159–182.

The information economics toolkit developed above signaling, screening, and their interaction identifies the mechanisms through which participation costs arise. The question now is how these mechanisms enter an individual supplier's decision. A supplier observing a published tender does not engage with signaling theory in the abstract, it faces a concrete assessment: what will this bid cost to produce, what is the probability of winning, and is the expected return sufficient relative to what the same resources could earn elsewhere? The theoretical task of this section is to formalise that assessment in terms that connect directly to the information economics framework developed above. The key construct is signaling cost K(s) - the cost a supplier incurs to produce the signal a given tender demands, which is not an administrative input but the operational expression of the information problem itself. When K(s) is high relative to expected returns, the signaling architecture is failing to attract participation, when K(s) varies systematically across supplier types for the same tender, the architecture is producing asymmetric entry barriers. The section develops K(s) as a relational construct, derives the three perceived transaction cost components as distinct failure modes of the signaling-screening system and formalises the bid/no-bid decision rule, as well as identifies the behavioural deviations that modify the model baseline.

1. HXH5GP

   Spence, M. (2002). Signaling in retrospect and the informational structure of markets. American Economic Review, 92(3), 434–459. https://doi.org/10.1257/00028280260136200

The central cost construct in this framework is K(s), the signaling cost a supplier incurs to produce signal “s” for a specific tender. K(s) is derived directly from Spence's model but adapted to capture a property that is particularly consequential in procurement: it is relational rather than absolute. K(s) is not a fixed property of the tender, nor a fixed property of the supplier, but a function of the gap between what the tender demands and what the supplier can already credibly present. A firm whose existing credentials, past project documentation, team certifications and methodology templates closely match what the tender's evaluation architecture requires faces low K(s), the signal is largely pre-assembled, and the marginal cost of adapting it to this specific tender is modest. An equally capable firm whose experience is equivalent in substance but documented in different formats, organised around different project typologies, or credentialled through different certification systems faces substantially higher K(s) for the same tender, not because it is less capable but because its existing signal portfolio is more distant from what the screening architecture recognises. This relational structure is the mechanism through which formally neutral tender requirements produce asymmetric entry barriers across firm types. An incumbent supplier who has won contracts from the same buyer before accumulates a signal portfolio that is progressively better matched to that buyer's evaluation architecture, reducing K(s) with each successive tender. A new entrant of equivalent capability must construct the signal from scratch, absorbing the full cost of portfolio assembly that the incumbent has already amortised across prior participations. The asymmetry is not in directly related to supplier capability but in signal cost and it is self-reinforcing, because each participation further reduces the incumbent's K(s) while non-participation leaves the entrant's portfolio undeveloped.

Signaling theory explains why participation costs exist and how they produce sorting effects across supplier types. It does not, however, specify what those costs consist of in operational terms. Transaction cost economics in the thesis is applied to provide this vocabulary. Williamson (1985) identified the costs of transacting - searching, negotiating, documenting, monitoring, enforcing and adapting agreements, as the determinants of how economic activity is organised across governance forms. In procurement suppliers incur these costs to identify opportunities, interpret requirements, assemble evidence, produce proposals and manage legal and procedural risks. Three TCE concepts explain why these costs reach the levels they do in procurement specifically. First, asset specificity, the degree to which a supplier's bid investment is tailored to a single tender and cannot be redeployed, this makes bid preparation costs largely sunk, unrecoverable if the bid loses, and therefore sensitive to any procedural feature that shifts effort upward. Second bounded rationality, the limited cognitive and informational capacity of procurement officers designing requirements, this explains why tender specifications are often imprecise, evaluation criteria vague, and scoring rubrics underspecified, generating uncertainty that suppliers must absorb. And third the opportunism risk, the possibility that one party exploits information advantages at the other's expense, this motivates the verification layers, compliance requirements and procedural safeguards that raise formal participation costs while simultaneously explaining why suppliers may doubt the process operates as stated. This thesis treats signaling theory as the primary explanatory architecture and TCE as the institutional vocabulary that makes the framework empirically operational: Spence's model explains the equilibrium structure, Williamson's concepts explain the cost environment within which signaling operates.

1. X3MN4T

   Williamson, O. E. (1979). Transaction-cost economics: The governance of contractual relations. Journal of Law and Economics, 22(2), 233–261. https://doi.org/10.1086/466942

The costs that determine participation, however, are not the objective transaction costs that an external analyst might calculate but the costs as suppliers perceive them. Some participation costs are measurable in principle, such as staff hours allocated to bid preparation, fees for external legal review, the direct expense of obtaining required certifications. But the bid/no-bid decision is driven by what suppliers believe these costs will be before committing resources and by how they anticipate procedural risks whose outcomes are uncertain at the point of decision. Whether a given deadline is perceived as manageable or prohibitive depends on the supplier's current workload, its familiarity with the buyer's documentation conventions, and its judgement about whether the timeline signals a genuine competitive process or a procedure whose outcome is already pre-determined. The same formal requirement can be perceived as routine overhead by one supplier and as a categorical deterrent by another, depending on experience, capability, portfolio and the available alternatives. This thesis therefore treats perceived transaction cost (PTC) as the actionable construct, the subjective assessment that mediates between observable tender features and the participation outcome. PTC is not a distortion of objective cost but an economically rational response to incomplete information: suppliers cannot know the true cost of participation or the true probability of winning before they commit, so they estimate both on the basis of observable signals and prior experience, and act on those estimates. The construct is consistent with the signaling framework, K(s) is itself a perceived cost, since the supplier must estimate the gap between its existing signal portfolio and what the tender demands before it can assess whether that gap is worth closing.

1. DQ7RWK

   Rindfleisch, A., & Heide, J. B. (1997). Transaction cost analysis: Past, present, and future applications. Journal of Marketing, 61(4), 30–54

The signaling-screening system described in section 1.2 can fail in three distinct ways, and each failure mode generates a corresponding component of perceived transaction cost. The first failure occurs when the cost of producing the required signal is excessive relative to the contract's expected value, meaning that, when the documentation, qualification evidence, methodology descriptions and compliance materials demanded by the tender require an investment that a rational supplier cannot justify given the uncertain return. This is bid preparation cost: the direct and opportunity costs of assembling the signal the screening architecture requires. It is driven primarily by formally observable tender characteristics, documentation scope, technical complexity, submission timelines, the number and specificity of required annexes, and it is the PTC component most directly addressable through procedural simplification.

The second failure occurs when the signal, once produced, cannot be reliably read. If the buyer's evaluation architecture lacks the capacity to interpret and reward quality signals consistently, because criteria are vague, scoring rubrics are underspecified, qualitative and price weights are unclear, or evaluators lack domain competence, then the return on signaling investment becomes unpredictable regardless of the signal's quality. This is evaluation uncertainty: the supplier's epistemic uncertainty about outcome predictability, conditional on good-faith procedural compliance. A supplier may be confident in the quality of its own proposal and still face genuine uncertainty about whether that quality will be recognised, because the mapping from signal to evaluation outcome is opaque. In Spence's terms, this corresponds to low signal-reading capacity “q”, the buyer's demonstrated ability to distinguish signal levels is insufficient, and suppliers who can detect this deficiency rationally discount the expected return on their signaling investment.

The third failure occurs when the supplier doubts that the screening mechanism operates as stated, when the concern is not that evaluators are incompetent but that the process may deviate from its formal rules in ways that systematically advantage certain bidders. This is evaluation integrity risk: the supplier's institutional risk perception that selection may be governed by considerations other than those published in the tender documentation. Unlike evaluation uncertainty, which presumes that the buyer is conducting the process in good faith but lacks the capacity to do so consistently, integrity risk concerns the possibility that the rules themselves are not binding, that the outcome is influenced by relationships, prior arrangements, or preferences that the formal process conceals rather than prevents. In the framework's terms, this corresponds to low perceived institutional integrity θ. It is the most informal of the three components, inferred not from tender documentation but from reputational signals, historical award patterns, market narratives and the supplier's accumulated experience with a given buyer or sector.

The gate model's structure and the relative intensity of its components vary with the evaluation method the contracting authority selects. Under price-quality criteria, the most economically advantageous tender (MEAT), both gates operate at full intensity: the supplier must produce elaborate quality signals, the buyer must possess domain competence to interpret them, and the institutional integrity of the qualitative evaluation is consequential because it involves judgement rather than mechanical comparison. Under lowest-price-only criteria, the model simplifies rather than disappears. K(s) reduces to the cost of formal qualification compliance and price calculation, since the supplier is not required to produce or has more limited application documentation. Gate 2 is therefore structurally lighter. Gate 1, however, does not vanish, it shifts. The signal-reading capacity q becomes less relevant because comparing prices requires no domain expertise, but the integrity assessment θ may intensify: when the only differentiator is price, the supplier's concern shifts to whether the technical specification was written neutrally or narrowed to favour a particular supplier who can offer a lower price through incumbency advantages or prior knowledge of the buyer's actual requirements. The model therefore predicts different gate profiles across evaluation methods, MEAT tenders are predicted to exhibit both Gate 1 and Gate 2 barriers at full intensity, while lowest-price tenders are predicted to exhibit attenuated Gate 2 but potentially concentrated Gate 1 θ risk. This variation is not a limitation of the framework but an analytically useful feature: if the empirical analysis confirms that participation patterns differ by evaluation method in the direction the model predicts, this constitutes evidence that the signaling mechanism, not just general procedural burden, is the operative driver.

These three components and the way their relative intensity shifts across evaluation methods also differ in how they can be empirically accessed. Bid preparation cost is predominantly formal and can be approximated through administrative tender data. Evaluation uncertainty sits at the boundary, partly formal, derivable from observable criteria structures and weighting patterns, and partly informal, shaped by the supplier’s experience of how similar criteria have been applied in practice. Integrity risk is predominantly informal, accessible only through direct elicitation. This gradient from formal to informal is consequential for both research design and reform analysis. Different reform instruments operate on different cost channels: simplification and standardisation primarily compress bid preparation cost, clearer evaluation frameworks reduce uncertainty, and credible oversight and predictable remedies address integrity risk by strengthening institutional trust.

The three PTC components enter the supplier's participation decision through a two-gate structure that combines sequential categorical assessment with graded probability estimation. The formulation reflects a core property of Spence's framework: the separating equilibrium either exists or it does not, and participants either fall on the viable side of the separation or they do not. But within these binary boundaries, suppliers form graded assessments of how likely they are to benefit from participation. The model therefore operates as a sequence of gates, each of which can produce categorical non-participation if the assessment falls below a subjective threshold, but which otherwise generates a probability estimate that carries forward into the final decision calculus.

The first gate assesses whether the tender's evaluation mechanism will produce a merit-based outcome. It combines two sub-components. The first is signal-reading capacity q, the supplier's assessment of whether the buyer's evaluation architecture has the criteria clarity, scoring transparency and domain competence to distinguish a strong proposal from a weak one. The second is institutional integrity θ, the supplier's assessment of whether the process will act on what it reads, or whether the outcome is influenced by considerations that the formal rules do not capture. These sub-components are conceptually distinct q is a competence judgement, θ is a trust judgement, but experientially they merge into a single assessment: will a good bid be recognised and rewarded? Together they produce P_screen = f(q, θ), the supplier's perceived probability that merit will determine the outcome. Critically, P_screen has a threshold property. If it falls below a subjective minimum, if the supplier perceives the process as fundamentally non-competitive or the evaluation as incapable of recognising quality, the response is categorical non-participation regardless of contract value or signaling cost. The gate closes and the supplier does not proceed to any further assessment. Above the threshold, P_screen enters the decision calculus as a probability modifier that discounts the expected return from participation.

The second gate is conditional on the first: only if P_screen exceeds the supplier's threshold does the cost-benefit calculation become relevant. At this stage the supplier assesses whether the expected return from bidding justifies the cost of producing the required signal, given competitive conditions and available alternatives. The decision rule is:

Bid if: P_screen × P_compete × (V − C) − K(s) > OC

P_screen carries forward from Gate 1 as the perceived probability that merit determines the outcome. P_compete is a distinct assessment of the supplier's perceived competitive win probability, conditional on the screening mechanism working as it should. P_compete is a function of the supplier's signal strength s relative to the expected signals of competing bidders and the anticipated number of competitors n. A supplier may fully trust the process but still estimate a low probability of winning because the incumbent's signal portfolio advantage is large, or because five other capable firms are expected to bid. P_screen and P_compete multiply because they represent conditionally independent assessments: P_screen asks whether quality will determine the outcome, P_compete asks whether, given that it does, this supplier's quality will be sufficient. V − C is the expected contract margin. K(s) is the signaling cost as developed above - relational, determined by the gap between the supplier's existing signal portfolio and what the tender demands. OC is the outside option, the expected return from deploying equivalent resources in the next-best alternative, whether a commercial engagement, another tender or retained capacity. OC is not zero: suppliers choose between bidding and the most productive use of constrained resources, and for firms with active commercial pipelines the procurement-specific signal cost must be justified against what winning equivalent revenue through other channels would require.

The sequential structure of the gates is consistent with how participation decisions are typically described in practitioner accounts and qualitative procurement research. A supplier scanning a new tender does not immediately begin costing the bid. The first assessment is a scan for viability. If this assessment fails the threshold, the supplier stops. If it passes, the supplier evaluates its competitive position. Only then does the supplier proceed to the resource commitment calculation that Gate 2 formalises. This ordering has a direct implication for reform analysis. The 2026 PIL reform primarily targets Gate 2 parameters. The model predicts that these measures will increase participation in sub-markets where Gate 1 is already passing and Gate 2 is the binding constraint. In sub-markets where suppliers are failing at Gate 1 (where the perceived screening viability is below threshold) procedural simplification will not reach the constraint that actually binds. Identifying which gate is operative in which sub-market is therefore the central diagnostic task that the empirical chapters are designed to perform, and the mixed-methods design is structured around this distinction: quantitative analysis addresses the observable Gate 2 parameters, qualitative interviews address the Gate 1 assessments that administrative data cannot capture.

The gate model as formalised above assumes that suppliers assess P_screen, P_compete, K(s) and OC rationally, if imprecisely. Decision research, however, identifies systematic behavioural deviations that affect participation in ways the rational baseline does not fully capture. This thesis incorporates three such deviations as conceptual extensions used to structure qualitative inquiry rather than to calibrate a formal behavioural model. The first is loss aversion. Kahneman and Tversky's (1979) prospect theory demonstrates that individuals weight losses more heavily than equivalent gains. In the participation decision, the signaling cost K(s) is a certain and immediate expenditure while the return is probabilistic and deferred, the supplier pays now and may win later. Loss aversion predicts that suppliers weight K(s) more heavily than the rational model implies, reducing participation below what the expected-value calculation would suggest, particularly for lower-value contracts where the certain cost looms larger relative to the uncertain prize. The second is ambiguity aversion. When Gate 1 parameters are not merely uncertain but genuinely unknowable, thus when criteria are so vague that the supplier cannot form even a rough estimate of P_screen, or when the buyer's behaviour is so inconsistent that no prior experience provides useful calibration, the supplier faces Knightian uncertainty rather than quantifiable risk. Under ambiguity aversion, the response to unknowable probabilities is more severe than the response to merely unfavourable ones: suppliers who cannot estimate whether the process will reward quality withdraw more aggressively than suppliers who estimate a low but definite probability of merit-based selection. This is the behavioural mechanism through which weak evaluation quality deters participation beyond what the gate model predicts, suppliers are not miscalculating but correctly identifying that the inputs to the calculation are absent. The third is threshold effects. Rather than continuously trading off expected returns against signaling costs, suppliers may apply categorical screening rules that classify tenders as viable or non-viable before engaging in any detailed assessment. When K(s) falls below an acceptance threshold, the cost has negligible influence on the decision, the supplier categorically refuses to participate regardless of contract value. This is consistent with bounded rationality and with how organisations manage decision volume, a firm receiving dozens of tender notifications cannot perform a full expected-value calculation for each and instead applies coarse filters that eliminate most opportunities before detailed evaluation begins. Importantly, the threshold is not fixed but shifts with the outside option: when commercial alternatives are plentiful and OC is high, the viability threshold falls and tenders that would otherwise attract bids are filtered out. These behavioural dimensions are investigated through the qualitative interviews, which probe for asymmetric cost weighting, categorical screening rules, and the conditions under which suppliers report abandoning the calculation entirely in favour of an immediate rejection.

1. TC3W83

   Kahneman, D., & Tversky, A. (1979). Prospect theory: An analysis of decision under risk. Econometrica, 47(2), 263–292. https://doi.org/10.2307/1914185
2. SFF8S7

   Knight, F. H. (1921). Risk, uncertainty and profit. Houghton Mifflin. (Also: Ellsberg, D. (1961). Risk, ambiguity, and the Savage axioms. The Quarterly Journal of Economics, 75(4), 643–669. — you have this in your literature collection)

The gate model identifies which constraint binds, but a further result from Spence (2002) clarifies what effective reform of that constraint looks like. Spence (2002) demonstrates that in any separating equilibrium, the private return to signal investment exceeds its direct contribution to productive quality. The return has two components: a productive component, where signal production genuinely improves the proposal and demonstrates real capability, and a signaling premium, where the investment serves competitive differentiation, redistributing contract awards among suppliers without adding substantive value to the deliverable. Because suppliers equate their marginal signaling cost to the total return including both components, they invest past the point where productive value alone would justify the expenditure. In procurement terms, this means that part of what suppliers spend on tender preparation genuinely improves proposal quality, while part is overinvestment driven by the competitive logic of signaling itself, elaborate formatting, redundant evidence, supplementary documentation that demonstrates effort rather than capability. The reform implication is precise. Effective quality separation requires a cost differential between supplier types, it requires that high-capability firms find signal production cheaper than low-capability firms. It does not require a high absolute cost.

1. H7CD2Q

   Spence, M. (2002). Signaling in retrospect and the informational structure of markets. American Economic Review, 92(3), 434–459. https://doi.org/10.1257/00028280260136200

The policy question is therefore not whether qualification requirements should exist but whether the absolute compliance burden is calibrated to the minimum sufficient for separation or has been inflated by institutional inertia, risk aversion and accumulated procedural layering beyond what information transfer requires. The 2026 reform, in these terms, is an attempt to compress the signaling premium, by reducing absolute K(s) through standardisation, credential reuse and threshold elevation, while preserving the cost differential that enables the buyer to distinguish supplier types. The gate model predicts that this compression increases participation in Gate 2-binding sub-markets by making signal production affordable for suppliers who were previously deterred not by screening viability concerns but by the sheer cost of entry. Whether the compression simultaneously weakens the separation, allowing low-quality suppliers to pass through at lower cost, is an empirical question that depends on whether the reformed requirements maintain the cost-quality correlation that Spence's model identifies as the necessary condition for informative signaling.

The gate model developed in section 1.3 describes how an individual supplier decides whether to bid. The market-level question is what happens when many suppliers apply the same logic simultaneously to the same sub-market over time. Competition in procurement is commonly proxied by participation: the number of bids per tender, the share of single-bid tenders and patterns of repeated winners. While these measures do not fully separate genuinely thin markets, where few capable suppliers exist, from institutionally deterred markets, where capable suppliers exist but choose not to participate, a substantial empirical literature establishes that participation levels have measurable economic consequences: each additional bidder is associated with lower contract prices, with particularly strong effects at low bidder counts and diminishing marginal returns thereafter. At the same time, participation is not costlessly beneficial without limit, since each additional bidder incurs signaling costs. In practical terms, persistent single bidding is rarely economically benign, while the policy target is not maximum participation but a shift from very low competition toward a mid-range where the price and quality benefits of additional bidders outweigh the aggregate cost of bid preparation.

To capture the possibility that participation patterns are not random fluctuations but stable states sustained by reinforcing expectations, this thesis uses equilibrium language in a pragmatic sense. An equilibrium, as applied here, refers to a relatively stable participation level in a defined sub-market (a CPV sub-category, buyer type or contract value tier) where supplier expectations and buyer behaviour mutually reinforce observed outcomes. Suppliers learn which tenders and which buyers are "worth bidding for," buyers learn what level of competition to expect and calibrate their requirements accordingly, and both adapt their strategies to the patterns they observe. Such feedback loops can sustain low-competition states even without any change in the underlying supplier pool's capabilities, the market's structure is stable not because it is efficient but because the expectations sustaining it are internally consistent.

The self-confirming property of signaling equilibria, developed in section 1.2, provides the theoretical basis for this persistence. The mechanism, where buyer beliefs determine requirements, requirements determine participation, and participation confirms beliefs, operates at the sub-market level through the gate model's feedback structure. Low participation in a sub-market produces concentrated winner patterns. Incoming suppliers observe these patterns and revise their Gate 1 assessments downward, either inferring low evaluation competence q from the buyer's apparent inability to attract a diverse pool, or inferring low institutional integrity θ from repeated awards to the same firms. Reduced P_screen triggers Gate 1 failures for marginal bidders, which further depresses participation and reinforces the concentration that prompted the inference. The equilibrium is stable not because it is efficient but because each participant's beliefs are validated by the outcomes those beliefs collectively produce. Critically, the same supplier pool could sustain a higher-participation equilibrium under different initial conditions, but no individual supplier has an incentive to deviate unilaterally, because the screening environment they face is shaped by the aggregate behaviour of all other suppliers.

The adverse selection dynamic identified by Akerlof (1970) has a direct sub-market implication when read through the gate model. If the suppliers who exit first are those with the strongest outside options and the greatest capacity to detect unfavourable screening conditions, as the model predicts, since these suppliers face the highest OC and the lowest tolerance for Gate 1 failure, then the average quality of the remaining bidder pool declines as participation falls. This compositional shift is the procurement equivalent of Akerlof's market-for-lemons deterioration, driven not by price adjustment but by the selective withdrawal of higher-capability firms. Whether buyers respond to this deterioration in ways that compound or correct it, and whether the resulting equilibrium is self-reinforcing or self-correcting, is an empirical question that the thesis investigates through the sub-market classification developed in Chapter 5.

1. HQZSV4

   Akerlof, G. A. (1970). The market for 'lemons': Quality uncertainty and the market mechanism. The Quarterly Journal of Economics, 84(3), 488–500. https://doi.org/10.2307/1879431

The empirical benchmarks on participation-price effects provide the basis for classifying sub-markets into competition tiers. The specific threshold definitions and classification criteria are developed in the analytical chapters that follow.

The theoretical framework developed in the preceding sections applies in principle to any public procurement market where information asymmetry generates participation costs. The thesis tests it, however, in a specific empirical setting: knowledge-intensive business services (KIBS), a category encompassing IT consulting, research and development, and professional advisory services. The choice of setting is not arbitrary. Knowledge-intensive business services, a term established in the innovation economics literature by Miles et al. (1995) and subsequently developed by Muller and Doloreux (2009), are defined by a cluster of economic characteristics that distinguish them from standardised goods or routine service provision. Their outputs are customised rather than uniform, co-produced with the client rather than delivered unilaterally, and dependent on the expertise, judgement and problem-solving capacity of individuals rather than on replicable production processes. The quality of the deliverable emerges through the engagement itself and cannot be fully determined in advance by either party. These properties do not merely make KIBS procurement difficult in practice, they systematically amplify every mechanism the theoretical framework identifies. The empirical scope, data sources and market characterisation of knowledge-intensive services procurement in Latvia are developed in the following chapter.

1. VPUQ76

   Miles, I., Kastrinos, N., Flanagan, K., Bilderbeek, R., den Hertog, P., Huntink, W., & Bouman, M. (1995). Knowledge-intensive business services: Users, carriers and sources of innovation (EIMS Publication No. 15). European Commission.
2. LQT6CB

   Muller, E., & Doloreux, D. (2009). What we should know about knowledge-intensive business services. Technology in Society, 31(1), 64–72. https://doi.org/10.1016/j.techsoc.2008.10.001

The economic classification of goods by information availability that is introduced by Nelson (1970) and extended by Darby and Karni (1973), provides the analytical basis for understanding why KIBS amplifies information asymmetry. Search goods are those whose quality a buyer can assess before purchase through inspection. Experience goods are those whose quality becomes apparent only through consumption. Credence goods are those whose quality remains difficult to verify even after delivery, because the buyer lacks the expertise to evaluate what was provided or because outcomes depend on contextual factors that prevent straightforward attribution. Knowledge-intensive services sit firmly in the credence category. A buyer who commissions a strategic analysis, an IT architecture design or a policy evaluation cannot fully specify what the "right" output looks like before the work begins, because the value of the service lies precisely in the supplier's capacity to identify and solve problems that the buyer could not have anticipated. Nor can the buyer straightforwardly verify quality after delivery. This means that the information asymmetry identified in section 1.2 as the foundational source of participation costs is at its most extreme in KIBS procurement. The buyer cannot assess quality before, during or after the transaction through direct observation alone, it is dependent, at every stage, on signals and proxies whose relationship to actual delivery quality is indirect and contestable.

1. 45A7G4

   Darby, M. R., & Karni, E. (1973). Free competition and the optimal amount of fraud. The Journal of Law and Economics, 16(1), 67–88. https://doi.org/10.1086/466756
2. 6D4MPE

   Nelson, P. (1970). Information and consumer behavior. Journal of Political Economy, 78(2), 311–329. https://doi.org/10.1086/259630

The credence property amplifies both gates of the participation model in distinct ways. Gate 1 is stressed because evaluating knowledge services proposals requires domain competence, assessing whether a proposed methodology is appropriate or whether a team's composition matches the assignment's complexity is a subject-matter judgement, not a procedural one. When procurement officers lack this expertise, signal-reading capacity q is structurally low regardless of the evaluation framework's formal design. Suppliers who have observed how similar evaluations have been conducted recognise this deficiency and discount P_screen accordingly. Gate 2 is stressed because the proxy signals that credence services demand are inherently more elaborate and costly to produce than those required for goods or standardised services. K(s) for a knowledge services tender is structurally higher than for a comparable-value goods tender, not because the procedure is more complex but because bridging the information gap requires more elaborate signaling. The combination of fragile Gate 1 and elevated Gate 2 costs is why the framework predicts that KIBS procurement exhibits lower participation and more persistent low-competition equilibria than sectors where quality is more directly assessable.

The gate model's predictions for KIBS are further sharpened by a misalignment between how knowledge services are procured and how their transaction properties suggest they should be governed. Kraljic (1983) classified purchases along two dimensions, supply risk and profit impact, to guide procurement strategy. Knowledge services frequently combine high measurement difficulty with non-trivial supplier scarcity, particularly in specialised domains such as cybersecurity consulting, advanced data analytics or niche policy research where the pool of genuinely qualified providers is small. This combination places them closer to strategic or bottleneck items in Kraljic's framework, categories that call for collaborative supplier relationships, negotiated scope development and flexible contract structures. Yet institutional procurement frameworks, driven by the transparency and equal treatment requirements discussed in section 1.1, often lead contracting authorities to treat knowledge services as routine purchases amenable to open competitive procedures with standardised documentation and heavy price weighting. This is precisely the governance mismatch that Bajari and Tadelis (2001) identified: competitive fixed-price tendering is efficient when scope can be specified ex ante, but knowledge services are transactions whose optimal scope frequently emerges through execution. The consequence for the gate model is specific. When a buyer applies commodity procurement logic to a strategic service, the evaluation criteria tend to overweight price and formal compliance at the expense of the qualitative dimensions (methodological sophistication, team judgement, adaptive capacity) that actually determine delivery quality. Suppliers perceive this imbalance: they bear high delivery and reputational risk while competing on thin margins under criteria that do not capture what they consider their actual value proposition. This perception can depresses both P_screen, because the evaluation architecture appears incapable of rewarding genuine quality, and the expected margin V − C, because aggressive price weighting compresses returns. The mismatch therefore operates on both gates simultaneously, weakening screening viability while reducing the prize that would justify the elevated signaling cost.

1. 8MDYSU

   Kraljic, P. (1983). Purchasing must become supply management. Harvard Business Review, 61(5), 109–117.

The international dimension reaches its sharpest expression in knowledge-intensive services. The general cost differentials that international suppliers face relative to domestic ones, such as language, procedural familiarity, local reference networks and the difficulty of assessing buyer reliability from a distance, are compounded in KIBS by the credence properties of the service. When quality is directly observable, as with standardised goods or construction to specification, an international supplier can compete primarily on price, and the additional costs of cross-border entry may be offset by production cost advantages. When quality is a credence attribute, the supplier must produce signals that are not only technically adequate but legible to the buyer’s specific evaluation conventions. A methodology description that is sound in substance may fail to signal quality if it is structured around frameworks unfamiliar to national evaluators, references projects in jurisdictions the buyer cannot assess, or uses professional terminology that does not map onto the scoring rubric’s categories. The signal portfolio mismatch that defines K(s) as relational is therefore systematically larger for international suppliers, not because they lack capability but because their existing signal portfolio was assembled for different screening architectures in different national procurement traditions.

Gate 1 is also differentially affected. An international supplier entering a Latvian sub-market has substantially less access to the informal information through which domestic suppliers assess screening viability. A domestic firm can draw on direct experience with a specific buyer, contacts in the professional community who have bid on similar tenders previously, and publicly discussed or informally circulated knowledge about award patterns and evaluation conduct. An international firm must form its Gate 1 assessment from formal tender documentation alone, which, as section 1.3 established, captures only the formal dimension of screening viability and leaves the q and θ assessments largely unresolvable. Under ambiguity aversion, the rational response to this information deficit is not a modestly discounted P_screen but a categorically cautious one, the supplier cannot distinguish a legitimate opportunity from a non-competitive process and may default to non-participation as the lower-risk choice. The framework therefore predicts that international suppliers face systematically higher K(s) and lower P_screen for the same tender, producing lower participation rates that reflect not capability differentials but the structural properties of cross-border signaling in credence-heavy markets. This prediction is tested through the thesis's treatment of supplier origin as a cross-cutting analytical variable, comparing participation patterns and PTC profiles between domestic and internationally registered firms within the same sub-markets.

A further consequence of the credence property is that it reshapes what suppliers invest in when they do choose to participate. When evaluation cannot assess substantive quality directly because the deliverable is a future performance rather than an inspectable product, the evaluation architecture necessarily relies on proxies: documentation quality, procedural compliance, volume of supporting evidence, visible certifications, the apparent thoroughness of the proposed methodology. Suppliers, as rational agents, optimise for what the evaluation actually measures rather than for what the buyer ultimately needs. DiMaggio and Powell (1983) described this convergence as institutional isomorphism, organisations adopt practices that signal legitimacy within a given institutional field, regardless of whether those practices improve substantive performance. In KIBS procurement, isomorphism takes a specific form: suppliers invest in producing increasingly elaborate documentation, accumulating certifications whose relevance to delivery quality is marginal, and formatting proposals to match evaluation conventions rather than to communicate genuine insight. The process rewards legibility over substance, and experienced suppliers learn to allocate effort accordingly. This dynamic connects directly to the signaling premium identified through Spence (2002) in section 1.3. The isomorphic investment is precisely the redistributive component of signaling, the portion that serves competitive differentiation within the evaluation architecture without contributing to the quality of the deliverable. It inflates K(s) for all participants, but disproportionately burdens suppliers whose comparative advantage lies in substantive expertise rather than in procedural fluency. The result is a systematic bias in the bidder pool: firms that are skilled at producing procurement-legible signals are overrepresented relative to firms whose capability is equivalent or superior but whose signal production is oriented toward delivery rather than evaluation. This is not corruption or incompetence on the buyer's part, it is a structural consequence of evaluating credence services through observable proxies, and it represents a cost of the procurement governance form that no procedural reform can entirely eliminate, only compress.

1. DRJLTN

   DiMaggio, P. J., & Powell, W. W. (1983). The iron cage revisited: Institutional isomorphism and collective rationality in organizational fields. American Sociological Review, 48(2), 147–160. https://doi.org/10.2307/2095101

The justification for focusing on knowledge-intensive services is not that they are uniquely problematic but that they represent the conditions under which every mechanism developed in this chapter operates at maximum intensity. Information asymmetry, signaling cost, screening fragility, governance mismatch, cross-border cost differentials and isomorphic pressures are all present in any procurement market — in KIBS they are simultaneously at their most acute. This makes the sector a strong test rather than a special case: if the gate model explains participation patterns where its predictions are sharpest, its relevance to less informationally demanding categories follows by extension. The working propositions that structure the empirical chapters — that PTC effects are stronger for lower-value contracts, that higher PTC reduces bidding probability, and that sub-markets with predominantly informal PTC are less responsive to the 2026 reform — are formulated and tested in this environment, with the expectation that the underlying mechanisms generalise beyond it.

Latvia’s public procurement system is anchored in the Public Procurement Law (PIL), complemented by Cabinet regulations that operationalize procedures, publication rules, and complaint deposits, and embedded in the EU’s harmonised procurement regime (notably the 2014 directives on “general,” “sectoral,” concessions, and defence procurement).  The institutional centre of this system is the procurement oversight body (IUB), a direct administrative authority subordinate to the Ministry of Finance, which operates the national procurement notice and complaint databases and serves as Latvia’s first-instance review venue for procurement challenges (i.e., ex ante/ex post legality control through the complaints function). 

1. V7BLR4

   https://www.iub.gov.lv/en/regulations
2. GYGBSG

   https://www.iub.gov.lv/en/search-notices , likums, MK noteikumi ?

A crucial governance feature—directly relevant for any empirical design—is that procurement transparency and enforceability are channelled through (i) mandatory publication of procurement notices in the national database, and (ii) a formal challenge mechanism that allows suppliers to contest procurement documents, decisions, or outcomes within legally defined time limits.  In parallel, EU integration is not only “legal harmonisation,” but also a shared informational infrastructure: above-threshold notices are routed into the EU Official Journal/TED ecosystem, and the CPV classification provides a standardised way to describe contract subject matter—critical for making cross-buy and cross-country comparisons. 

1. 6R8BDQ

   https://www.iub.gov.lv/en/search-notices , likums, MK noteikumi ?
2. CMKXF6

   https://single-market-scoreboard.ec.europa.eu/business-framework-conditions/public-procurement_en direktiva, EU politikas dokumenta atsauces?

In macroeconomic terms, public procurement is economically significant in Latvia, consistent with the EU-wide perspective that procurement accounts for a material share of GDP. The European Commission’s procurement performance pages use an EU-wide “~15% of GDP” reference point for the scale of procurement in modern economies, and Latvia-specific official sources report similarly high salience at the national level.  For example, Latvia’s supreme audit institution reports a 2023 procurement value of EUR 5.4 billion (excl. VAT), about 14% of GDP, and the procurement authority’s own communications report a 2024 total procurement volume of EUR 5.45 billion, about 13% of GDP.  These magnitudes justify treating procurement not as a “narrow administrative topic,” but as a core public-finance allocation mechanism—exactly the kind of setting where competition, buyer discretion, and information problems can produce first-order welfare effects. 

The EU internal market objective deserves closer attention because it bears directly on one of this thesis's analytical dimensions. The EU procurement directives, beginning with the 2004 framework and consolidated in Directive 2014/24/EU, were designed explicitly to enable cross-border supplier participation by harmonising procedures, mandating publication in the Official Journal above threshold values, and prohibiting discriminatory technical specifications. The underlying economic logic is straightforward: opening national procurement markets to suppliers from other member states should increase competition, reduce prices and improve quality through a larger and more diverse bidder pool. In practice, however, cross-border participation remains marginal. European Commission data consistently show that direct cross-border procurement accounts for only around 5% of above-threshold contract awards across the EU, and the figure is lower still for smaller member states with less internationally visible procurement markets. Latvia is no exception: despite formal compliance with the directive framework, the overwhelming majority of bids in knowledge-services tenders come from domestically registered firms. This gap between institutional design and observed behaviour suggests that the formal removal of legal barriers has not eliminated the participation costs that international suppliers face. Language, familiarity with national procedural conventions, access to local reference networks and the ability to assess buyer reliability from a distance all constitute additional cost components that domestic suppliers do not bear to the same degree. Whether these differential costs are large enough to produce systematically different participation patterns is an empirical question that the thesis investigates by treating supplier origin as a cross-cutting analytical variable.

1. RNVJ2L

   European Commission. (2021). Single Market Scoreboard: Public Procurement. https://single-market-scoreboard.ec.europa.eu/
2. KDNTY7

   European Parliament & Council of the European Union. (2014). Directive 2014/24/EU on public procurement. Official Journal of the European Union, L 94, 65–242.

Latvia is in the middle of a structural reform initiative that explicitly aims to reduce bureaucracy, raise procurement discretion/flexibility, and shift the governance model toward data-driven oversight and post-control. The Ministry of Finance frames the reform goals in terms of improving value for money, increasing competition, increasing transparency through broader publication of lifecycle information, and modernising governance through digital infrastructure and performance monitoring (KPI-style indicators and reference prices).  A key design choice is the planned upward shift of national regulation thresholds to EU threshold levels: the Ministry states that the system would raise the effective thresholds from EUR 10,000 to EUR 143,000 for goods/services and from EUR 20,000 to EUR 5,538,000 for works.

1. TXH4TE

   https://www.fm.gov.lv/lv/publisko-iepirkumu-sistemas-reforma

Substantively, the reform package couples threshold lifting with simplification measures and a rebalancing of oversight instruments. On simplification, the Ministry reports (i) a reduction of mandatory exclusion grounds (described as “2 mandatory” rather than “12”), and (ii) an expected reduction in procedure duration by about 25%, presenting both as mechanisms to reduce administrative load (for contracting authorities) and friction costs (for suppliers).  On institutional governance, the reform would reposition IUB as the central management authority for the public procurement system, including responsibility to maintain and develop the Electronic Procurement System (EIS) and the Publication Management System, while other institutions would focus on narrower specialisations (e.g., ICT procurement focus for the state digital agency and sector specific competence centers).

1. KRR34U

   https://www.fm.gov.lv/lv/publisko-iepirkumu-sistemas-reforma

The legislative timeline matters. Official parliamentary communications show the reform law was taken up in the legislature with a first-reading stage in October 2025 (conceptual support), and later proceeded through a second-reading stage in February 2026.  The Ministry of Finance states implementation is planned to enter into force “gradually,” conditional on parliamentary adoption of amendments to the PIL and related procurement laws and subsequent Cabinet regulation amendments.  This sequencing supports a thesis narrative that treats 2026 not as a single break-point but as a reform window with anticipatory behavioural incentives and evolving institutional constraints.

(improve flow) The empirical significance of participation levels is well established. Studies of EU procurement consistently find that each additional bidder is associated with lower contract prices, with especially strong effects when competition increases from one to three or four bidders and diminishing marginal returns thereafter. A recent Danish government study found that moving from one to four bids was associated with roughly 10–13% lower prices, while the marginal effect declined once participation reached approximately five bidders. A large-scale study of Turkish procurement auctions found that procurement costs fell until approximately six to eight bidders, with sectoral variation, offering a policy-relevant benchmark for the participation range where competition benefits dominate bidding costs.

1. U3V6XY

   Aagaard, K. B., & Linaa, J. G. (2024). The impact of competition for public contracts on public finances. Danish Competition and Consumer Authority.
2. GLK77F

   Onur, I., & Tas, B. K. O. (2019). Optimal bidder participation in public procurement auctions. International Tax and Public Finance, 26, 595–614

At the European Union level, competition in public procurement is routinely assessed through participation-based indicators. The European Commission's Single Market Scoreboard treats the single-bidder rate as one of its most critical metrics, assigning it triple weight in its composite indicator due to its direct link with competition, transparency and market access. Persistent single-bidder outcomes are interpreted not as isolated tender-specific anomalies but as indicative of structural participation barriers. Latvia's single-bidder rate of approximately 21.9%, considerably above the EU average of roughly 14%, places it among the member states where structural barriers are most likely operative, a positioning that motivates the present investigation into the mechanisms producing this outcome.

1. E2B2DE

   European Commission. (2018). Single Market Scoreboard: Public Procurement. https://single-market-scoreboard.ec.europa.eu/

Knowledge-intensive business services occupy a structurally distinct position within Latvia's procurement market. Analysis of IUB open data for the eForms publication period (October 2023 – April 2026) shows that the three KIBS CPV divisions — IT services and consulting (CPV 72), research and development (CPV 73), and professional and management services (CPV 79) — account for 14,493 lot-level procurement records, representing 6.8% of all 213,620 lot-level records in the national procurement database and 7.7% of total estimated procurement value. The value share exceeding the count share reflects the above-average contract size in IT services procurement: CPV 72 alone accounts for approximately €1.04 billion in estimated value, roughly three-quarters of all KIBS procurement value, driven by large-scale information system development contracts. Research and development services (€101 million) and professional services (€259 million) are smaller by value but substantially more numerous — professional services alone comprise 58% of all KIBS lot-level records.

The composition of contracting authorities purchasing KIBS reveals a concentration among central state institutions and larger municipalities. The Latvia Investment and Development Agency leads by volume, followed by the Ministry of Welfare, the Riga City municipality, Jelgava and Daugavpils municipalities, and major universities (Rīga Stradiņš University, Riga Technical University, University of Latvia). This pattern is consistent with the nature of knowledge services: central agencies procure IT infrastructure and policy consulting, municipalities procure professional services for urban development and administration, and universities procure research-adjacent services. The buyer profile matters for the gate model because institutional procurement capacity — the competence to design evaluation criteria and manage complex service specifications — varies substantially across these authority types.

The value distribution of KIBS procurement is heavily skewed toward smaller contracts. Among the lot-level records where estimated value is reported, the median stands at approximately €42,000, with the interquartile range spanning €18,000 to €150,000. Over half of all valued KIBS lots fall below €50,000, and nearly three-quarters fall below the EU services procurement threshold of €143,000. This distribution differs markedly across CPV divisions: IT services show a median of €100,000 with wide dispersion (reflecting both small software maintenance contracts and multi-million-euro system development procurements), while professional services cluster at a median of €36,000 and research services at approximately €50,000. The prevalence of sub-threshold contracts has a direct implication for the empirical design: a substantial share of KIBS procurement falls below the EU threshold and therefore does not appear in TED, making the IUB national dataset essential for capturing the full participation landscape.

The evaluation method applied to KIBS tenders provides an initial structural indicator of how contracting authorities balance price against quality in knowledge-services procurement. Across all KIBS lot-level records, 22.9% apply the most economically advantageous tender criterion (MEAT) with explicit quality weighting, while 43.6% use price as the sole or dominant criterion. The remaining records correspond to notice types where evaluation criteria are not yet specified (prior information notices, planned contract announcements). The MEAT adoption rate varies substantially across CPV divisions in a pattern consistent with the theoretical expectation that evaluation mechanism quality correlates with service complexity: R&D services (CPV 73) show the highest MEAT rate at 43.5%, reflecting the inherent difficulty of specifying research outputs in purely price terms, while professional services (CPV 79) show the lowest at 28.9% with 62% applying price-only criteria. IT services (CPV 72) occupy an intermediate position at 26.2% MEAT. From the perspective of the two-gate model, the predominance of price-only evaluation in professional services suggests that Gate 1 screening conditions differ across sub-markets: where buyers default to lowest price, signal affordability (Gate 2) becomes the binding constraint, but evaluation quality q may simultaneously be low, meaning that even suppliers willing to invest in bid preparation face a mechanism that cannot distinguish their quality advantage.

At the EU level, TED contract award data for 2023 positions Latvia's above-threshold KIBS procurement as a moderately competitive market by participation volume: the average number of tenders received per KIBS contract award stands at 7.58, compared with an EU-27 aggregate of approximately 4.8 across all KIBS awards. This figure, however, must be read with caution. The TED average includes only above-threshold procurements where publication in the Official Journal is mandatory, and Latvia's high field coverage rate (92% of KIBS award records report bidder counts) means the average is well-grounded statistically but not directly comparable with countries where coverage is substantially lower. Moreover, the average masks distributional asymmetry: competitive markets with high mean bidder counts can still contain a substantial tail of single-bid or low-competition outcomes that the mean does not reveal. Latvia's overall single-bidder rate of approximately 21.9%, as reported in the European Commission's Single Market Scoreboard, sits well above the EU average of roughly 14% — and KIBS procurement, with its higher information asymmetry and qualification complexity, is likely to exhibit single-bid rates at or above the national average, though KIBS-specific single-bid rates are not separately reported in the Scoreboard and will be derived from the tender-level data in section 2.3.

Cross-border participation provides a complementary lens on market accessibility.

TED data show that in 2023, only 1.9% of tenders received in Latvian above-threshold KIBS procurement originated from firms registered in other EU member states — marginally below the EU-27 average of 2.0%. The contrast with Estonia is instructive: despite comparable market size and geographic proximity, 13.6% of tenders in Estonian KIBS procurement came from other EU firms, nearly seven times the Latvian rate. Lithuania, at 3.1%, also exceeds Latvia. This pattern is consistent with the gate model's prediction that participation costs are not uniform across institutional environments: language barriers, familiarity with national procedural conventions, and the ability to assess buyer reliability from a distance all constitute cost components that vary across jurisdictions. The low cross-border rate in Latvia does not, in itself, constitute evidence of formal barriers — Latvia's procurement framework is fully compliant with EU directives — but it does suggest that the effective cost of participation for international suppliers remains high relative to the expected return, exactly the condition the model captures through the Gate 2 inequality.

The conjunction of these features — knowledge services characterised by acute information asymmetry and credence-good properties, a procurement landscape where the majority of contracts fall below EU thresholds and more than three-fifths of professional services tenders are evaluated on price alone, single-bidder rates substantially above the EU average, and cross-border participation near the bottom of the EU distribution — makes Latvia's KIBS procurement market a productive empirical setting for testing the two-gate participation model. The variation across CPV divisions, contracting authority types and evaluation methods provides the within-sample heterogeneity needed to distinguish Gate 1 from Gate 2 effects, while the imminent 2026 PIL reform, which targets precisely the cost and complexity parameters the model identifies, creates a natural policy experiment against which the model's predictions can be assessed.

The gate model’s structure determines the research design. Gate 2 parameters — signaling cost, contract value, competitive intensity — leave observable traces in administrative tender data. Gate 1 parameters — perceived evaluation competence and institutional integrity — are subjective assessments that no administrative dataset records. Testing the full gate structure therefore requires both quantitative analysis of what tender data can capture and qualitative investigation of what it cannot.

The research follows a sequential explanatory mixed-methods design (Creswell & Plano Clark, 2018) in which a quantitative phase establishes empirical patterns that a subsequent qualitative phase is designed to explain. A purely quantitative study could test whether observable tender characteristics predict participation levels but could not distinguish between two sub-markets that exhibit identical low participation for different reasons: one where capable suppliers find the signaling cost prohibitive, and another where capable suppliers do not trust the evaluation process. The regression residuals in the first case would be small; in the second, they would be large, because the binding constraint lies at Gate 1 where the data has no direct purchase. A purely qualitative study could investigate how suppliers perceive participation costs but could not establish whether these perceptions correspond to measurable patterns across the full population of tenders. The sequential structure assigns each phase a specific diagnostic function: the quantitative phase produces a participation map and identifies where observable features fail to predict, the qualitative phase targets these sub-markets specifically. The integration produces the central deliverable: a classification of sub-markets by binding constraint type that generates conditional reform predictions.

1. B64CNN

   Creswell, J. W., & Plano Clark, V. L. (2018). Designing and conducting mixed methods research (3rd ed.). Sage.

Data sources and variable construction

The empirical analysis draws on two complementary data infrastructures that together cover both the administrative and substantive dimensions of Latvian knowledge-intensive services procurement during 2024–2025. Their relationship, and the role each plays in the analytical design, requires explicit mapping before the sample construction logic is presented.

Two distinct systems serve Latvian procurement transparency. The Publication Management System (PVS), operated by the Procurement Monitoring Bureau (IUB), receives and publishes eForms notices, the standardised electronic forms mandated by EU Regulation 2019/1780, for all procurement procedures that cross formal publication thresholds. The PVS feeds the IUB data visualisation dashboard and the open data service at open.iub.gov.lv, which provides machine-readable daily JSON exports of all published notices from October 2023 onwards.

1. Z9AW59

   EU Regulation 2019/1780

A separate system, the Electronic Procurement System (EIS) at eis.gov.lv, operated by the State Regional Development Agency, serves as the transactional platform through which contracting authorities conduct procurement procedures and suppliers submit bids. The EIS captures a broader universe than the PVS: for 2024, the EIS records 17,068 unique procurement procedures compared with approximately 11,500 competition-stage eForms notices in the PVS for the same period. The difference is accounted for almost entirely by three procedure categories conducted through the EIS but not generating formal eForms publications, that are small procurements below PIL thresholds (5,053 procedures in 2024), unregulated procurements (1,291), and pre-procurement market consultations (3,037). These three categories together account for 9,381 procedures, precisely the gap between the two systems.

This thesis draws on both sources for different analytical purposes. The PVS open data, with its structured eForms fields covering notice type, CPV classification, evaluation criteria, procedure type, and contracting authority metadata, provides the basis for the market landscape characterisation and the descriptive statistics in section 2.3. The EIS dataset, which records lot-level detail including planned contract values, submission deadlines, and procurement status for all procedures including those below publication thresholds, serves as the primary source for the tender-level quantitative analysis. Where the two sources overlap — formal PIL and SPSIL procedures — they describe the same procurement events through complementary lenses: the PVS captures what was published and how, while the EIS captures how the procedure was conducted and other detail.

The primary quantitative dataset is a combined EIS export covering the 2024 and 2025 procurement years, comprising 64,452 rows at lot level and 35,588 unique procurement identifiers (Iepirkuma_ID). The dataset was extracted from the EIS open data portal and processed in two analytical layers. The first layer operates at the procurement level — rows deduplicated by Iepirkuma_ID — and uses administrative metadata fields (status, procedure type, CPV classification, contract value, submission deadlines) that are uniform within each procurement. The second layer returns to the lot-level structure for variables that vary across lots within a single procurement, primarily planned contract values and lot-specific subject classifications.

The secondary source is the procurement documentation itself, accessed through hyperlinks embedded in the EIS export and through the IUB eForms publication system. Document-level extraction provides the deeper variable set required for the gate model operationalisation: bidder count, evaluation criteria weights and structure, qualification threshold specificity, documentation complexity, the number and nature of required annexes, and award decision records. The extraction of variables from procurement documentation follows a structured protocol grounded in content analysis reliability principles (Krippendorff, 2018) and adapted for automated extraction in line with recent methodological developments. Tender documents are published primarily as PDF and .doc files, typically in Latvian, with variable formatting across contracting authorities. The protocol proceeds in four stages: manual coding of a gold standard reference set; iterative prompt development for the Anthropic Claude Haiku API with structured output schemas (Lunn et al., 2025); full-pipeline application with audit logging; and validation on a random 20% subsample against source documents, with agreement rates reported by variable type following Krippendorff's reliability-per-variable principle.

The research sample is constructed through a sequential filtering procedure applied to the 35,588 unique procurements in the EIS export. The logic proceeds in four steps, with Steps 3 and 4 producing parallel samples that serve different analytical functions. Table 2.3 reports the count at each stage.

Step 1 retains only procurements with a concluded or award-decided status — Līgums noslēgts (contract concluded), Noslēgts (concluded), or Lēmums pieņemts (decision taken) — on the grounds that only procurements that have reached an award decision generate the observable bidder behaviour constituting the dependent variable. This reduces the sample from 35,588 to 26,279 procurements, excluding terminated (Pārtraukts, 3,063), active (Izsludināts, 2,595), discontinued (Izbeigts, 1,917), and other in-progress statuses.

Step 2 restricts to the two procedure types that are analytically tractable for the research question and together constitute the governed KIBS market under study: open tender (Atklāts konkurss) and small procurement (Mazie iepirkumi). This restriction is substantively motivated on three grounds. First, these are the only two procedures that operate under sufficiently standardised regulatory conditions — PIL and its below-threshold provisions respectively — to permit cross-procurement comparison of participation behaviour and evaluation architecture. Second, all other procedure types present structural information deficits that preclude quantitative analysis: negotiated procedures (Sarunu procedūra, 888 excluded), restricted tenders (Slēgts konkurss, 621), Annex 2 service procurements (928), and the remaining regulated variants operate under bespoke rules and generate non-comparable documentation structures. Third, from the perspective of KIBS market governance, open tenders and small procurements together define the boundary within which the signalling and participation dynamics theorised in the gate model actually operate — the former as the standard competitive mechanism above thresholds, the latter as its structural analogue below thresholds within the same contracting authority universe. Supplier consultations (Apspriede ar piegādātājiem, 4,422 excluded) and unregulated procurements (Neregulēts iepirkums, 1,770) are excluded additionally because they do not constitute formal award procedures. This step yields the analysis universe of 17,552 procurements, steps 3 and 4 operate in parallel on this base.

Step 3 applies CPV-based classification to define the administrative KIBS boundary: the core CPV sample (3A) retains 1,630 procurements whose primary CPV code falls within divisions 72 (IT services and consulting), 73 (research and development services), or 79 (professional and management services). The extended CPV sample (3B) adds a further 159 procurements where any additional CPV code falls within these divisions, yielding 1,789 procurements in total. These samples define the outer administrative boundary of the KIBS market as captured by the procurement classification system and serve as the baseline for descriptive analysis.

Step 4 applies a KIBS text-based classification developed as a complementary identification strategy, described in the following sub-section. The strict text-based KIBS rule (4A) identifies 833 procurements and the extended rule (4B) identifies 1,290. Cross-referencing against the CPV samples reveals that 652 procurements appear in both the strict text rule and the extended CPV sample — the high-confidence KIBS core — while 181 procurements classified as strict KIBS by text fall outside the CPV boundary, and 94 procurements within the extended CPV boundary are classified as non-KIBS by the text rule.

The analytical sample for the quantitative regression analysis is the extended CPV sample (3B, n = 1,789). The text-based classification serves as a robustness check and identifies boundary cases for qualitative follow-up in the document-extraction stage.

CPV codes provide a necessary but insufficient boundary for knowledge-intensive services procurement. The CPV taxonomy was designed for administrative classification rather than theoretical precision: division 79 (Business services) encompasses printing and postal procurement activities alongside management consulting, and divisions 72 and 73 include software licensing and routine maintenance alongside substantive IT consulting and applied research. Conversely, genuinely knowledge-intensive procurements occasionally appear under other CPV divisions — particularly 71 (architectural and engineering services), 80 (education), and 85 (health) — when the knowledge-intensive component is bundled into a broader service contract or when contracting authorities exercise discretion in code assignment.

To address this, a KIBS-specific keyword dictionary was developed drawing on the theoretical definition of knowledge-intensive business services as activities whose primary output is the application of specialised knowledge to generate new information, solutions, or capabilities for the client organisation (Miles et al., 1995; Strambach, 2008). The dictionary operationalises this definition through three families of positive signals and one family of negative disambiguating signals, applied to the procurement title field (Iepirkuma_nosaukums) and the aggregated lot title field (Iepirkuma_dalas_nosaukums). Table 2.4 presents the full dictionary.

The classification proceeds in two tiers. The strict KIBS rule classifies a procurement as KIBS if three conditions hold simultaneously: the subject type field (Iepirkuma_prieksmeta_veids) is Pakalpojums (service), at least one strong KIBS signal group is present or a combination of development and digital/system signals fires, and negative context signals do not dominate. The strict rule is conservative by design, accepting false negatives in preference to false positives given the title-only information available.

Table 2.3. Sequential Sample Reduction: EIS 2024–2025 Procurement Data

Note. Steps 3A/3B and Steps 4A/4B are parallel outputs derived independently from the Step 2 analysis universe (n = 17,552). Diagnostic rows D1–D3 report cross-method overlap; they do not represent additional filtering. Status inclusions at Step 1: Līgums noslēgts (15,432), Noslēgts (9,193), Lēmums pieņemts (1,654). Procedure exclusions at Step 2 include all regulated procedure types other than Atklāts konkurss and Mazie iepirkumi.

Table 2.4. KIBS Text-Based Classification Dictionary

Note. Asterisks denote wildcard stems (e.g., pētīj* matches pētījums, pētīja, pētīts). All terms are applied to Latvian-language text fields. Strong KIBS signal groups carry greater weight than supporting/development groups in the strict classification rule (Rule 4).

The possible KIBS rule relaxes the dominance condition for negative signals and includes procurements where only supporting-level signals fire, capturing ambiguous cases such as IT platform development that may or may not constitute knowledge-intensive services depending on the full contract structure. The union of strict and possible classifications covers 2,123 procurements across the Step 2 universe.

The text method has a known limitation: the EIS export contains procurement titles and lot titles but not full technical specifications, meaning classification operates on brief and often formulaic descriptions rather than the substantive contract content. The strict rule is calibrated to this constraint and accepts classification from titles alone only when the language is unambiguous. The 181 procurements identified as strict KIBS outside the CPV boundary (diagnostic row D2) and the 94 CPV-boundary procurements classified as non-KIBS (row D3) represent the priority targets for document-level validation in the extraction stage.

The second source is the procurement documentation itself provides deeper insight into the variable data set comprised of: tender specifications, evaluation criteria descriptions, qualification requirements, clarification records and award decision notices. Document extraction in the second layer produces the full variable set for every tender that enters the analytical sample: bidder count, evaluation criteria weights and structure, qualification threshold specificity, documentation complexity, the number and nature of required annexes, and

The extraction of variables from procurement documentation follows a structured protocol grounded in content analysis reliability principles (Krippendorff, 2018) and adapted for automated extraction in line with recent methodological developments. Tender documents in the Latvian procurement system are published as PDF and doc files, typically in Latvian, with variable formatting across contracting authorities. Recent studies have demonstrated that large language models can achieve human-comparable accuracy on structured data extraction tasks when guided by predefined schemas, with Konet et al. (2024) reporting 96.3% accuracy and Schmidt et al. (2024) approximately 80% accuracy across domains, both emphasising that accuracy varies by variable type and that human validation remains essential.

1. 79YV3L

   Krippendorff, K. (2018). Content analysis: An introduction to its methodology (4th ed.). Sage.
2. AYB5AZ

   Konet, A., Thomas, I., Gartlehner, G., Kahwati, L., Hilscher, R., Kugley, S., Crotty, K., Viswanathan, M., & Chew, R. (2024). Performance of two large language models for data extraction in evidence synthesis. Systematic Reviews, 13(1). https://doi.org/10.1186/s13643-024-02609-x
3. S9W584

   Schmidt, L., et al. (2024). Exploring the use of a large language model for data extraction in systematic reviews: A rapid feasibility study. arXiv preprint arXiv:2405.14445.

The extraction proceeds in four stages. First, a gold standard is established through manual coding of a reference set. Second, the extraction pipeline is developed using the Anthropic Claude Haiku API with structured prompts specifying target variables, definitions, output formats and disambiguation rules, following the iterative refinement approach validated by Lunn et al. (2025). Third, the calibrated pipeline is applied to the full analytical sample with audit logging. Fourth, validation is performed on a random subsample of approximately 20% by manual comparison against source documents, with agreement rates reported by variable type following Krippendorff’s principle that reliability must be established per variable. Variables falling below acceptable agreement thresholds are either re-extracted with adjusted prompts or excluded with the limitation reported.

1. KTZ4BU

   Lunn, E., et al. (2025). Harnessing large-language models for efficient data extraction in systematic reviews: The role of prompt engineering. Campbell Systematic Reviews, [vol]. https://doi.org/10.1002/cesm.70058
2. Z7N78H

   Mathes, T., Klassen, P., & Pieper, D. (2017). Frequency of data extraction errors and methods to increase data extraction quality: A methodological review. BMC Medical Research Methodology, 17(1), 152. https://doi.org/10.1186/s12874-017-0431-4

The variables entering the quantitative analysis are organised by their function in the gate model. The dependent variables are bidder count (a count variable) and a binary single-bid indicator. Independent variables are grouped by the gate mechanism they proxy: Gate 2 indicators capture observable dimensions of signaling cost K(s), including documentation complexity, timeline pressure, procedure type, contract value and amendment dynamics; partial Gate 1 indicators capture formal evaluation architecture dimensions, including the price-only versus price-quality distinction and quality weight proportions. Controls include contracting authority type, CPV sub-category, year of publication and value tier. Supplier origin enters both as a control and as an interaction term. Tables 2.1 and 2.2 map each theoretical construct to its empirical proxy and specify the full variable dictionary respectively.

[TABLE 2.1: Model construct mapping — INSERT FROM CURRENT DRAFT]

[TABLE 2.2: Variable dictionary — INSERT FROM CURRENT DRAFT]

Analytical approach

The quantitative analysis begins with a descriptive baseline mapping participation across sub-markets defined by CPV sub-category, buyer type and contract value tier. Sub-markets are classified into competition tiers — low, moderate and high — using the benchmarks established in section 2.1. The core econometric test models bidder count using Poisson or negative binomial regression (depending on overdispersion), with a complementary logistic regression on the single-bid indicator. The specification follows the gate structure: Gate 2 proxies and partial Gate 1 indicators enter as independent variables, with evaluation method treated as both a stratification variable and a control. If K(s) proxies show stronger effects in MEAT tenders than in price-only tenders, as the theory predicts, this constitutes evidence that the signaling cost mechanism is the operative driver.

The regression results are informative not only through coefficients but through residuals. Sub-markets where the model explains low participation well are K(s)-binding candidates. Sub-markets where participation is substantially lower than predicted are Gate 1 candidates, and these become the targeting frame for the qualitative phase.

The qualitative phase uses semi-structured interviews with a purposively selected sample of knowledge-service suppliers, stratified by firm size, sector (IT, R&D, professional services) and supplier origin, including both active participants and firms that have reduced or ceased bidding. The interview protocol follows a three-phase structure: open decision-process elicitation (before the gate framework is introduced), vignette-based gate probing (using anonymised tenders from the quantitative dataset, particularly those where the regression model underperforms), and direct model validation. Coding applies a priori gate-mapped codes (G1-q, G1-θ, G2-K(s), G2-OC, G2-P_compete) with allowance for emergent codes, following Miles, Huberman and Saldaña (2014).

1. 83USHL

   Miles, M. B., Huberman, A. M., & Saldaña, J. (2014). Qualitative data analysis: A methods sourcebook (3rd ed.). Sage.

Methodological boundaries and ethical considerations

The dataset covers the pre-reform period (2023–2025) and does not include ex-post observations of how participation changes after the 2026 PIL amendments take effect. The reform predictions are therefore ex-ante assessments. The cross-sectional design captures a snapshot rather than dynamic adjustment. The LLM-assisted extraction introduces measurement uncertainty that varies by variable type. The qualitative phase rests on an exploratory sample sufficient for mechanism validation but not for prevalence claims. Cross-border participation rates in Latvian knowledge-services procurement are low, which may limit the robustness of domestic-versus-international comparisons.

All interview respondents provide informed consent. Interview data is anonymised — neither supplier names, buyer names nor specific tender identifiers appear in the thesis. Tender vignettes used in interviews are anonymised before presentation. Where ethics review is required by the University of Latvia’s institutional procedures, it is obtained prior to commencing the interview phase.

The descriptive baseline establishes the empirical landscape against which the regression analysis and qualitative validation operate. It proceeds in four steps: the competition profile by CPV division, which is the primary sub-market segmentation; the buyer-side variation, which captures institutional capacity differences; the procedure-type comparison, which tests whether formal simplification alone shifts participation; and bid price dispersion, which provides an indirect indicator of specification ambiguity and evaluation uncertainty. Throughout, the interpretation connects observed patterns to the gate model’s predictions about where participation constraints bind.

Table 2.5. Competition profile of KIBS procurement by CPV division (lot-level, n = 1,241)

Note. Sample restricted to KIBS tenders (CPV 72, 73, 79) with concluded status and available bidder count. Dispersion CV = standard deviation of bid prices divided by mean, computed for multi-bid tenders only (n in parentheses). Buyer hist. SB rate = buyer’s historical single-bid rate in prior 12 months (n in parentheses indicates coverage). Source: gold standard extraction from EIS/IUB procurement documentation, 2024–2025.

Table 2.5 presents the headline finding. The overall KIBS single-bid rate stands at 34.6%, approximately 1.6 times the Latvian national average of 21.9% and 2.5 times the EU average of roughly 14%. This confirms the central premise developed in section 1.5: knowledge-intensive services, as credence goods with acute information asymmetry, systematically depress supplier participation relative to the broader procurement market. The effect is not uniform across CPV divisions, and the gradient is analytically informative.

IT services (CPV 72) exhibit the most severe competition deficit: half of all tenders receive a single bid, the mean bidder count is 2.27, and only 18.5% of tenders attract four or more bidders. This is consistent with the theoretical expectation that IT procurement combines the highest signaling cost K(s) — driven by elaborate technical specifications, team composition requirements, and methodology proposals — with evaluation architectures that require substantial domain competence to interpret. The combination of elevated Gate 2 cost and structurally fragile Gate 1 conditions produces the pattern the model predicts: a sub-market where both gates constrain simultaneously.

Research and development services (CPV 73), though a smaller sub-market (n = 70), show a comparable pattern: 41.4% single-bid rate and only 11.4% reaching four or more bidders. The low bid price dispersion in multi-bid R&D tenders (median CV = 0.102, the tightest of the three divisions) suggests that where competition does exist in R&D, it is among suppliers with similar cost structures — consistent with specialised markets where the pool of genuinely qualified providers is narrow. This is the “thin market” variant: low participation may reflect genuine supply-side scarcity rather than institutional deterrence, a distinction the qualitative phase is designed to resolve.

Professional services (CPV 79), the largest sub-market at 955 tenders, show the most favourable competition profile: 30.5% single-bid rate, 3.06 mean bidders, and 32% reaching four or more bidders. The relatively broader participation reflects the heterogeneity within CPV 79, which encompasses both routine services (translation, printing, security) where K(s) is low and quality is more directly assessable, and substantive advisory services (management consulting, strategic evaluation) where credence properties are fully present. The aggregate statistics therefore mask within-division variation that the regression analysis in the following sub-section will decompose.

Table 2.6. Competition profile by buyer type (KIBS, n = 1,118)

Note. Buyer classification based on contracting authority type. n varies from Table 2.5 due to missing buyer-type classification for 123 tenders. “Gate model note” indicates theoretical expectation for the buyer type’s evaluation capacity. Source: gold standard extraction, 2024–2025.

Buyer-type variation provides the first evidence for the institutional capacity channel theorised in the gate model. Municipalities, the buyer type with the weakest expected procurement capacity, show the second-highest single-bid rate (38.9%) and the lowest share of 4+ bidder tenders (20.6%) among major categories. State-owned enterprises, which combine commercial procurement experience with institutional resources, show the strongest profile among non-specialised buyers: 24% single-bid rate and 41.3% reaching four or more bidders. The pattern is consistent with the Gate 1 prediction: suppliers assess buyer evaluation competence (q) and calibrate participation accordingly. Buyers with higher expected domain competence attract more bidders, not because they face different supplier pools but because suppliers perceive a higher probability that quality signals will be correctly read.

The hospital category (n = 25, mean 5.88 bidders, 16% single-bid) is an instructive outlier. Hospital KIBS procurement typically involves standardised IT maintenance or routine professional services with well-specified deliverables and established supplier markets. These are precisely the conditions under which the gate model predicts high participation: low K(s), high q (because specification clarity reduces evaluation ambiguity), and transparent competitive conditions. The contrast between hospital procurement and the overall KIBS profile illustrates that the participation deficit is not intrinsic to knowledge services as a category but emerges from the interaction between service complexity and evaluation architecture quality.

Company-type buyers (private entities conducting regulated procurement, n = 56) show the highest single-bid rate at 46.4%. This likely reflects a combination of narrow supplier markets for specialised industrial services and lower visibility of private-sector tenders in supplier monitoring systems, both of which compress the effective bidder pool.

Table 2.7. Competition by procedure type (KIBS, n = 1,211)

Note. Restricted to tenders with identifiable procedure type. Preparation days = calendar days from publication to submission deadline (median). Source: gold standard extraction, 2024–2025.

The procedure-type comparison yields a result that is modest in magnitude but analytically significant. Small procurements, which operate under lighter regulatory requirements and shorter timelines, show a higher single-bid rate (36.6%) than open tenders (32.0%), and a lower share of 4+ bidder outcomes (26.0% versus 31.0%). This is counterintuitive if procedural complexity were the binding constraint: simpler procedures should attract more bidders, not fewer. The finding is consistent with the gate model’s prediction that when Gate 2 cost (K(s)) is reduced but Gate 1 conditions (q, θ) are simultaneously weak — as in smaller, less scrutinised procurements where evaluation architectures are less developed — procedural simplification does not reach the constraint that actually binds. The median preparation time difference (18 versus 20 days) is small, suggesting that timeline pressure is not the primary mechanism separating the two procedure types. This pattern has direct reform implications: the 2026 PIL reform’s threshold elevation will shift many current small procurements out of regulated oversight entirely. If the binding constraint in these sub-markets is already at Gate 1 rather than Gate 2, deregulation may reduce transparency without improving competition.

Table 2.8. Bid price dispersion (CV) by CPV division (multi-bid KIBS tenders, n = 763)

Note. Coefficient of variation = SD(bid prices) / mean(bid prices), computed only for tenders with two or more bids. Higher CV indicates greater dispersion in submitted prices. Source: gold standard extraction, 2024–2025.

Bid price dispersion, measured as the coefficient of variation of submitted bid prices within each tender, provides an indirect indicator of specification ambiguity. When suppliers submit vastly different prices for nominally the same service, it signals that the technical specification is not constraining interpretations: bidders are pricing different understandings of scope, effort, and deliverable quality. This is precisely the condition under which evaluation uncertainty (low q) intensifies, because the evaluator must compare proposals that are not, in substance, responding to the same brief.

The CPV-level pattern is consistent with this interpretation. R&D services (CPV 73) show the tightest price dispersion (median CV = 0.102, IQR 0.060–0.158), suggesting that where competition exists in research procurement, it is among suppliers with convergent scope interpretations and cost structures. Professional services (CPV 79) show the widest dispersion (median CV = 0.227, IQR 0.092–0.525), consistent with the heterogeneity of the division: a tender for “management consulting services” admits a far wider range of scope interpretations than a tender for a specific software module. The wide interquartile range within CPV 79 — spanning from near-zero to above 0.5 — confirms that the division contains both well-specified and poorly-specified sub-markets. IT services (CPV 72) sit between the two (median 0.201, IQR 0.098–0.377).

The dispersion measure enters the analytical framework not as a direct Gate 1 variable — it captures an outcome of specification quality rather than a supplier perception — but as a diagnostic flag. Tenders with high dispersion (CV above the 75th percentile, approximately 0.49) are candidates for the qualitative phase’s Gate 1 investigation: they are tenders where the evaluation architecture’s capacity to compare bids meaningfully is likely compromised by scope ambiguity, and where suppliers who anticipate this ambiguity may discount P_screen accordingly.

A final pattern in the descriptive data speaks to the equilibrium persistence mechanism theorised in section 1.4. The buyer’s historical single-bid rate — the share of that buyer’s prior KIBS tenders that received only one bid — shows a statistically significant positive association with current-tender single-bid outcomes (Pearson r = 0.105, p = 0.002, n = 883). Buyers in the highest historical single-bid band (50% or above) show a 41.1% single-bid rate on current tenders, compared with 28.9% for buyers in the lowest band (0–15%). The relationship is modest in magnitude, indicating that buyer history is one signal among several rather than a deterministic predictor. Nevertheless, the pattern is consistent with the self-confirming equilibrium mechanism: buyers with historically low-competition outcomes continue to attract low competition, suggesting that suppliers’ beliefs about buyer quality are informed by observable track records. Whether this reflects accurate assessment of persistent buyer-side deficiencies or inertial belief updating that could be broken by targeted intervention is a question the qualitative phase will probe.

Taken together, the descriptive baseline establishes four empirical regularities that structure the regression analysis and qualitative investigation. First, the KIBS single-bid rate is substantially above both national and EU benchmarks, confirming the theoretical prediction that credence-good properties amplify participation barriers. Second, the competition gradient across CPV divisions (IT worst, professional services least constrained) maps onto the information asymmetry gradient the model identifies. Third, procedure-type simplification does not, by itself, resolve the competition deficit — suggesting that the binding constraint often lies at Gate 1 rather than Gate 2. Fourth, buyer type and buyer history are associated with participation outcomes in directions consistent with the institutional capacity and equilibrium persistence mechanisms the model predicts. The regression analysis that follows tests whether these patterns hold after controlling for confounders, and the residual analysis identifies the sub-markets where observable tender features fail to predict participation — the candidates for Gate 1 investigation in the qualitative phase.

Regression analysis: Gate 2 determinants of participation

[Poisson/NB + logit results; MEAT vs full sample; supplier origin interactions; proposition testing]

Residual analysis and qualitative targeting

[Sub-markets where observed participation << predicted; vignette selection; bridge to 2.4]

[TO BE WRITTEN — depends on completing interviews]

Sample and interview process

[Final sample, stratification, conduct]

Gate mechanisms in supplier decision-making

[Open elicitation findings; vignette responses; G1-q, G1-θ, G2-K(s) evidence; domestic vs international]

Cross-case patterns

[Gate-failure profiles by sub-market type; sector and firm size effects]

[Classification matrix: sub-market × binding constraint. K(s)-binding where regression explains low participation and interviews confirm cost barriers. q-binding where residuals are large and interviews reveal evaluation competence concerns. θ-binding where residuals are large and interviews reveal integrity concerns. Mixed profiles.]

[Reform instruments mapped to gate parameters. Conditional predictions by sub-market type: K(s)-binding → reform reaches; q-binding → reform does not reach; θ-binding → reform does not reach. H3 test. International supplier differential.]

[Beyond the 2026 reform: additional instruments for q and θ. Evaluator competence development. Transparency mechanisms. Differentiated policy by sub-market type. Limitations of ex-ante assessment.]