Abstract
Rising pharmaceutical expenditure poses a persistent governance challenge for public health systems in low- and middle-income countries, yet empirical evidence on pricing in Business-to-Government (B2G) settings remains scarce. This scarcity reflects a fundamental constraint where administrative procurement records are rarely accessible for independent analysis. This study addresses that gap by exploiting rare access to 26,652 medication transaction records from Thailand’s Comptroller General’s Department (CGD) for 2019–2024. Using acute pain medications, a semi-logarithmic hedonic regression decomposes price variation into product-level and institutional-level determinants. Internal product characteristics, including pharmacological class, dosage form, formulation technology, and brand status, account for the majority of price variation. Conventional market mechanisms, such as competitive pressure and procurement volume, generate no meaningful price reduction once product composition is controlled. Institutional governance factors, however, are associated with significant residual effects. Hospitals under non-Ministry of Public Health ministries are associated with price premiums of 13–29% for equivalent products, while Government Pharmaceutical Organization procurement is associated with a 9.6% price reduction. These findings indicate that product characteristics are the dominant drivers of pharmaceutical price variation in B2G markets, while institutional governance factors are associated with residual price differences after controlling for product composition. These residual differences may have relevant implications for procurement practices in fragmented public health systems. Note on causal inference: All reported associations are based on observational cross-sectional data and should not be interpreted as evidence of causal relationships. The regression coefficients reflect conditional correlations between procurement prices and their covariates, controlling for observed product and institutional characteristics.
Keywords
1. Introduction
Rising pharmaceutical expenditure has become a defining policy challenge for health systems worldwide. Policymakers have responded with an array of cost-containment instruments, including external and internal reference pricing, value-based pricing, and pharmacoeconomic evaluation, yet these measures frequently fall short of their intended goals, particularly in low- and middle-income countries (LMICs). In such settings, constrained market size limits competitive dynamics, while weak regulatory enforcement, inadequate governance, and corruption along the pharmaceutical supply chain collectively erode policy effectiveness.1,2 A further barrier is the persistent scarcity of empirical research on pharmaceutical markets and pricing tailored to LMIC contexts.1,3 Taken together, these conditions underscore the urgency of robust government engagement in designing efficient, evidence-based pharmaceutical pricing mechanisms.
The theoretical basis for government intervention in pharmaceutical markets rests on well-documented market failures that prevent purely market-driven outcomes from being socially optimal. 4 Chief among these failures is information asymmetry: patients generally lack the clinical expertise to make fully autonomous treatment decisions, placing significant discretionary authority with prescribing physicians and creating an imbalance between patients, prescribers, and pharmaceutical firms. 4 This asymmetry is compounded by aggressive pharmaceutical marketing, which can distort prescribing toward costlier and not necessarily more effective therapies. 5 Opaque pricing structures, confidential rebate arrangements, and multi-layered supply chains further impede transparency, making it difficult for patients and providers alike to assess true drug costs or make meaningful price comparisons. Government intervention, through mandated price disclosure, marketing regulation, and independent comparative effectiveness information, offers a corrective to these systemic distortions.
The hedonic pricing model provides a rigorous analytical framework for understanding pharmaceutical price formation in such imperfect markets. By deconstructing observed prices into the implicit valuations of a drug’s constituent attributes, such as clinical efficacy, safety profile, route of administration, dosing convenience, and long-term health outcomes, hedonic regression reveals what market participants are effectively willing to pay for each characteristic. 6 This attribute-level decomposition is directly relevant to health technology assessment (HTA) bodies and policymakers seeking to ground reimbursement and reference-pricing decisions in quantifiable measures of therapeutic value. While hedonic methods have been widely applied across industries, from real estate and automobiles to agriculture and technology, their pharmaceutical applications have been largely confined to exploring efficacy-safety-price relationships or assessing cross-national price disparities for comparable products. 7
Existing hedonic pricing studies in pharmaceuticals predominantly examine Business-to-Business (B2B) transactions, where market competition and product differentiation are assumed to be the primary drivers of prices negotiated between manufacturers and private buyers such as wholesalers or private hospitals.8,9 This leaves a significant gap in the literature concerning Business-to-Government (B2G) pharmaceutical procurement, where the pricing environment is fundamentally different. B2G transactions are directly shaped by explicit government policies, including tendering systems, national essential medicine lists, and reference price ceilings, as well as by institutional factors including procurement agency mandates, ministry affiliations, hospital classification, and governance vulnerabilities such as corruption. These elements are not adequately captured by hedonic models calibrated to competitive B2B markets, and they remain insufficiently examined in the empirical literature.
A structural reason for this gap is that B2G pharmaceutical pricing research requires access to government administrative procurement databases, which are records typically held internally by state agencies and rarely released for independent academic analysis. Unlike retail or insurance claims data, which are sometimes commercially available, actual transaction-level B2G procurement records reflect the prices at which government entities settle with suppliers after tendering and negotiation. These data are therefore the only reliable evidence base for studying the true cost of public pharmaceutical purchasing, as opposed to published list prices or reference prices that may deviate substantially from settled transaction prices. Their inaccessibility has been a fundamental constraint on the development of this literature. The present study is among the first in the LMIC context to exploit such a dataset, made possible through access to the Comptroller General’s Department (CGD) of Thailand, a central government financial authority whose records capture actual transaction prices across all public hospital procurement for civil servant medical reimbursements.
Thailand offers an especially instructive case for examining B2G pharmaceutical pricing dynamics. The country has experienced sustained growth in pharmaceutical expenditure, which rose from 4.3% of GDP (680 billion Thai Baht) in 2021 to 6.2% of GDP (830 billion Thai Baht) in 2022, with average annual growth of 7–8% recorded even before the COVID-19 pandemic. The government sector is responsible for approximately 60–70% of total medicine consumption, with a comparable share of market value attributable to imported finished medicines, making public procurement a central lever for cost management. In response, the Thai Ministry of Public Health, through its Subcommittee for Development of the National List of Essential Medicines (NLEM), collaborates with the Health Intervention and Technology Assessment Program (HITAP), the Royal Medical Councils, and the National Health Security Office (NHSO) to set value-based reference prices and negotiate prices for high-cost medications.10-12
This study addresses the identified research gap by applying hedonic regression to the B2G pharmaceutical market in Thailand, using a large administrative dataset of 26,652 internal medication transactions recorded by the Comptroller General’s Department (CGD) across public hospitals from 2019 to 2024. The use of CGD data constitutes a methodological contribution in its own right by providing actual government transaction prices, not list prices, catalogue prices, or insurance reimbursement rates that are rarely available for econometric analysis in LMIC health systems. This centralized procurement record, covering civil servant and government employee medical expenditure across multiple ministries and hospital tiers, provides a uniquely comprehensive window into the mechanisms of public-sector drug pricing. The analysis focuses on analgesic and NSAID medications, a pharmacologically diverse and widely procured category, and examines how both internal product characteristics (active pharmaceutical ingredients, dosage forms, formulation technology) and external institutional factors (ministry affiliation, hospital class, Government Pharmaceutical Organization procurement) jointly determine prices in a government-controlled market. By integrating these dimensions, the study contributes new evidence on the complex interplay between pharmaceutical attributes, institutional bargaining power, and government policy in shaping drug prices, with direct relevance to Thailand and transferable value for other LMICs facing similar challenges.
The remainder of the paper is structured as follows. Section 2 reviews the literature on hedonic regression and its applications in pharmaceutical markets. Section 3 describes the data and presents a descriptive analysis. Section 4 details the hedonic regression model and reports empirical findings. Section 5 concludes with policy implications and directions for future research.
2. Literature Review on Hedonic Regression Models and Their Application in the Pharmaceutical Industry
2.1. Foundations of Hedonic Pricing
Hedonic regression is a well-established method in empirical economics, premised on the idea that the market price of a differentiated good reflects the implicit prices consumers attach to its underlying characteristics. 13 Formally, the approach regresses observed prices on a vector of product attributes, often alongside a time dummy to control for temporal variation, enabling the estimation of each attribute’s marginal contribution to match with price. 14 The resulting implicit prices can be interpreted as the marginal willingness to pay (WTP) for incremental improvements in a specific characteristic, holding all others constant. Hedonic models thus serve multiple analytical purposes: deconstructing price variation, constructing quality-adjusted price indices, evaluating consumer surplus changes from product innovations, and informing policy on product valuation. 15
The methodology’s most classical application is in housing economics, where property prices are deconstructed into the implicit values of structural features (bedrooms, floor area), neighborhood attributes (school quality, crime rates), and environmental amenities. 16 From this foundation, the framework has been extended cover a broad range of markets including automobiles, consumer electronics, agricultural commodities, and pharmaceuticals.2,17 In each domain, the key analytical insight is the same, namely, that observed transaction prices encode latent valuations of product characteristics that can be recovered through regression.
2.2. Hedonic Pricing in the Pharmaceutical Industry
Within pharmaceuticals, hedonic regression has been applied primarily to construct price indices that account for quality change and to identify the factors that explain price differentials across products, markets, or countries. Cross-national price comparisons have been a particularly active area. Studies using quasi-hedonic methods in the United States have shown that international price disparities reflect not only regulatory differences but also variations in product attributes and their implicit valuations.8,9 Danzon and Chao (2000) 9 demonstrate that countries with more stringent price controls or greater reliance on generic competition consistently achieve lower drug prices, suggesting an association between regulatory stringency and lower market price levels. Conversely, in markets with limited regulation, generic competition exerts downward pressure on prices by expanding buyer choice and increasing price sensitivity.
A second strand of research uses hedonic methods to assess how clinical characteristics, including efficacy, safety, and innovation, are priced by the market. A notable early contribution by Cockburn and Anis (2001) 18 applied hedonic regression to disease-modifying antirheumatic drugs (DMARDs), finding counterintuitive results that authors attributed to confounding factors such as orphan drug designations and the specialized cost structures of niche therapeutic markets, underscoring the complexity of pharmaceutical price determination and the need to account for broader institutional and market context.
Wu et al (2014) 3 extended this literature to the Chinese pharmaceutical market, employing quasi-hedonic regression to examine how market competition and manufacturer origin affect drug prices. Their findings confirmed that prices decline with the number of generic and therapeutic substitutes, but rise with therapeutic class breadth, reflecting the premium attached to broader treatment coverage. Notably, locally manufactured drugs commanded lower prices than those from multinational firms, a result that persisted after controlling for product attributes and pointed to the role of brand origin and market positioning as pricing determinants.
2.3. Research Gap: B2G Pharmaceutical Pricing
Despite these contributions, the existing hedonic pricing literature in pharmaceuticals is characterized by a systematic emphasis on Business-to-Business (B2B) transactions in relatively competitive markets. The implicit assumption is that innovation, efficacy, and market competition are the dominant price drivers, an assumption that may be appropriate for private-sector procurement but is poorly suited to government purchasing contexts. B2G pharmaceutical transactions are governed by a different logic, namely, that prices are shaped by centralized tendering rules, national reference price ceilings, essential medicine policies, and the procurement mandates of specific government agencies. The institutional landscape, including which ministry oversees a hospital, how procurement is organized, and whether centralized agencies such as a government pharmaceutical organization mediate purchase, can exert substantial influence on prices independently of product-level attributes.
This institutional dimension has received limited attention in the hedonic pricing literature for two compounding reasons. First, and most fundamentally, B2G pharmaceutical procurement data are government administrative records that are rarely made available for independent academic analysis. Actual transaction-level data, reflecting prices settled through tendering and negotiation between government buyers and pharmaceutical suppliers, are held internally by procurement agencies, finance ministries, or centralized reimbursement authorities, and are typically not released to researchers. This inaccessibility has been a binding constraint on the field since without actual B2G transaction prices, it is not possible to study the determinants of what governments actually pay for medicines, as opposed to published list prices or reference prices that may deviate substantially from settled transaction prices. Second, even where procurement data exist, the conceptual framing of most hedonic studies has not explicitly incorporated government institutional factors, such as ministry affiliation, hospital classification, or centralized procurement program participation, as explanatory variables [18]. As a result, there is minimal empirical evidence as to how these governance dimensions affect pharmaceutical prices in government-dominated markets.
The present study addresses both dimensions of this gap. By securing access to the Comptroller General’s Department (CGD) database, a government financial system that records actual reimbursement-based transaction prices for civil servant medical care across Thailand’s public hospital network, the study obtains the kind of administrative B2G procurement data that has constrained prior research. This access is non-trivial because the CGD dataset captures prices at the point of actual government payment rather than at the point of price registration or list publication, and it spans all ministry-affiliated hospitals, enabling direct comparison of procurement prices across institutional settings that would not be possible with commercially available or publicly disclosed data. The analytical framework developed here, which explicitly models ministry affiliation, hospital type, and GPO procurement status alongside product characteristics, is designed to exploit this institutional richness and to produce findings that are directly relevant to procurement governance reform.
Thailand’s pharmaceutical system exemplifies precisely the kind of complex B2G environment that existing models have not fully addressed. With the government accounting for the majority of pharmaceutical consumption and procurement fragmented across multiple ministries, each with distinct management structures, bargaining capacities, and institutional incentives, price determination cannot be understood through product attributes alone. This study addresses this gap by developing a hedonic regression framework that explicitly incorporates both internal product characteristics and external institutional factors, using Thailand’s public hospital procurement data as a novel empirical setting. In doing so, it contributes to a more complete and policy-relevant understanding of pharmaceutical pricing in government-controlled markets, with findings that hold broader lessons for LMIC health systems confronting similar structural challenges.
Two additional theoretical frameworks are relevant to the B2G pharmaceutical procurement context. First, monopsony theory predicts that a single large buyer—in this case the government—possesses sufficient market power to drive prices below competitive equilibrium levels. Thailand’s public sector, accounting for 60–70% of pharmaceutical consumption, would theoretically qualify as a dominant buyer capable of extracting price concessions. However, this monopsonistic potential is only realized if procurement is sufficiently centralized and coordinated. The institutional fragmentation observed across ministries—with distinct tendering systems, formulary lists, and bargaining capacities—dilutes this potential market power and is consistent with the higher prices observed among non-MoPH institutions. Second, principal–agent theory highlights the governance risks inherent in delegated procurement decisions. When individual hospital procurement officers act as agents for the central government (the principal), informational asymmetries, incomplete incentive alignment, and limited monitoring may permit procurement behavior that does not minimize prices. The statistically significant ministry and hospital-type effects identified in this study are consistent with principal–agent inefficiencies operating at the institutional level, though causal attribution remains beyond the scope of the current analysis.
3. Data
This study draws on 26,652 internal medication procurement transaction records from Thailand’s public hospitals over the period 2019–2024. The data were compiled by the Comptroller General’s Department (CGD), which administers medical expense reimbursements for civil servants and government employees. As a centralized administrative record of actual transaction prices—rather than catalogue or list prices—the CGD dataset provides a uniquely transparent basis for analyzing pharmaceutical pricing in the Business-to-Government (B2G) context.
It is important to note that the CGD dataset reflects pharmaceutical procurement reimbursed under the Civil Servant Medical Benefit Scheme (CSMBS), which covers approximately 4.5 million civil servants and their dependents. This represents one of three major public health coverage schemes in Thailand, alongside the Universal Health Coverage scheme and the Social Security Scheme. CSMBS beneficiaries generally have access to a broader formulary and may receive higher-cost products than UHC beneficiaries, potentially reflecting upward price pressure specific to this scheme. The findings of this study are therefore most directly applicable to the CSMBS procurement context and generalization to the broader Thai pharmaceutical market should be made with caution.
The dataset specifically encompasses medication procurement records from the universe of Thailand’s public hospitals reporting to the CGD financial system. While these facilities fall under different ministerial jurisdictions, they collectively represent the state-controlled hospital network. This comprehensive coverage allows the analysis to move beyond hospitals managed solely by the Ministry of Public Health. By utilizing transaction data from the CGD centralized financial database, the study captures the diverse procurement activities of hospitals across all participating ministries including Defense, Interior, and Higher Education.
The analysis is restricted to analgesic medications for acute pain management: paracetamol (acetaminophen) and a structured selection of nonsteroidal anti-inflammatory drugs (NSAIDs). This category was chosen for three reasons. First, it constitutes one of the highest-volume procurement categories across public hospitals, ensuring sufficient data depth. Second, it encompasses meaningful variation in regulatory status as paracetamol is available over the counter, whereas most NSAIDs are classified as dangerous drugs under Thai law, creating natural variation in procurement constraints. Third, the category includes both commodity generics and technologically differentiated formulations (modified-release and enteric-coated products), allowing the hedonic framework to simultaneously identify the price effects of pharmacological innovation, formulation technology, and institutional governance factors. These properties make the category well-suited as a testing ground for a B2G hedonic pricing model.
Regarding the nature of the CGD price data, the CGD database captures prices recorded in the government’s centralized financial accounting system at the point of actual payment disbursement to suppliers. These are settled transaction prices following the completion of the tendering process and invoice verification and they represent the amounts that participating hospitals actually paid.
However, two important caveats apply. First, the CGD data may not capture all post-transaction adjustments, including confidential volume rebates, retrospective clawback arrangements, or supplementary discount agreements negotiated directly between hospitals and suppliers outside the formal tendering record. To the extent that such arrangements exist and vary systematically across ministries or hospital types, price measures may partly reflect administrative recorded prices rather than final net prices. Second, procurement recording practices may vary across ministries. These sources of measurement heterogeneity are acknowledged as limitations of the dataset and results should be interpreted accordingly.
3.1. Price Variation Across Products and Institutions
Procurement Price Distribution per Unit of Analgesic and NSAID Medicines, 2019–2024 (Thai Baht)
Source: Authors’ calculation. Data from the Comptroller General’s Department (CGD).
Note. EC = Enteric Coated; MR = Modified Release.
Among oral solid formulations, non-selective COX-II inhibitors (diclofenac, ibuprofen, mefenamic acid) command median unit prices below 1 THB, consistent with their status as mature, highly competitive generics. Paracetamol in combination with tramadol carries a higher median of 5.24 THB, reflecting the added regulatory complexity of opioid-containing products. Selective COX-II inhibitors (celecoxib: 16 THB; etoricoxib: 23.72 THB) occupy a substantially higher price tier, commensurate with their more recent patent history and pharmacological novelty. Parenteral formulations display the widest absolute price range: parecoxib injection records a median of 161.57 THB and a maximum exceeding 570 THB, reflecting near-monopoly competition in the intravenous selective COX-II segment. The wide within-category variation, particularly visible in maximum prices, further suggests that institutional and governance factors introduce price dispersion that product characteristics alone cannot account for.
3.2. Variable Classification
To structure the empirical analysis, price determinants are classified into two categories: internal attributes, which capture the intrinsic pharmacological and technological characteristics of each drug product, and external attributes, which capture institutional and market-level factors operating through the procurement environment. This distinction maps directly onto the five-model specification strategy described in Section 4.
Internal attributes are operationalized along three dimensions. API class groups drugs by chemical structure following Rang et al (2011), 19 acetaminophen and its combinations (paracetamol-orphenadrine, paracetamol-tramadol), propionic acid derivatives (ibuprofen), phenylacetic acid derivatives (diclofenac), and sulfonyl/sulfonamide coxibs (celecoxib, etoricoxib, parecoxib). Dosage form distinguishes eight categories: tablets, oral solutions and suspensions, injections, oral drops, topical gels, external sprays, eye drops, and suppositories. Formulation modification classifies products as conventional (unmodified), enteric-coated (EC), or modified-release (MR). These technologies alter bioavailability or gastric tolerability and are expected to command price premiums over standard formulations. Finally, brand status (original brand versus generic) is included to isolate any residual brand premium that persists after controlling for pharmacological and formulation characteristics.
Product comparability in this study is operationalized through three observable attributes. These include the API and dosage regimen combination, such as diclofenac sodium 50 mg tablet, the dosage form category, and the formulation modification class encompassing conventional, enteric-coated, or modified-release. This approach controls for major sources of product differentiation visible in procurement records. It does not, however, distinguish between products from different manufacturers within the same API-form-modification group. Nor does it capture differences in packaging size, excipient composition, or bioequivalence documentation, which could introduce residual price variation within groups. The brand status variable partially addresses manufacturer-related heterogeneity, but within-generic price variation across manufacturers remains an uncontrolled source of residual heterogeneity. Readers should interpret price comparisons across equivalent products with this limitation in mind.
External attributes are drawn from three sources. Hospital type follows the Ministry of Public Health’s 2022 classification framework, which ranks facilities by size, case-mix complexity (Case Mix Index), and financial risk into five tiers: Centers of Excellence (A), Standard Hospitals (S), General Hospitals (M), Family Hospitals (F), and University Hospitals (U). University hospitals are treated separately, given evidence that pharmaceutical company sponsorship and academic detailing can influence drug selection and pricing in academic medical centres.6,17
Ministry affiliation captures systematic differences in bargaining capacity, procurement protocols, and governance structures across the five ministry groups. Market competition is proxied by the number of Trade Pharmaceutical Unit (TPU) codes registered for each dosage regimen in Thailand, which is a measure of competitive pressure at the product level, varying from monopoly (parecoxib 40 mg injection) to 19 registered suppliers (paracetamol 500 mg tablet). GPO procurement status identifies transactions channeled through the Government Pharmaceutical Organization, a state enterprise with a preferential mandate to supply public hospitals. GPO-routed procurement is hypothesized to reduce prices through centralized price negotiation and bulk purchasing.
Descriptive Statistics of Variables Used in the Hedonic Pricing Model (N = 26,652)
Source: Authors’ compilation. Data from the Comptroller General’s Department (CGD), 2019–2024.
Note. EC = Enteric Coated; MR = Modified Release; GPO = Government Pharmaceutical Organization.
4. Econometric Estimation and Empirical Findings
4.1. Model Specifications and Estimations
The hedonic pricing function is estimated in semi-logarithmic (log-linear) form, with the natural logarithm of the per-unit procurement price as the dependent variable. This functional form is well-established in hedonic price analysis
20
and is appropriate here for two reasons. First, pharmaceutical unit prices are heavily right-skewed and span several orders of magnitude; the log transformation compresses this range and improves residual properties. Second, and more importantly for policy interpretation, the semi-log specification allows each regression coefficient β to be converted into a precise percentage price premium or discount relative to the reference category using the transformation (e
β
− 1) × 100. All percentage effects reported in Section 4.2 use this exact transformation rather than the raw coefficient. The estimating equation is:
Hedonic Regression Results – Determinants of Pharmaceutical Procurement Price (Dependent Variable: Ln Unit Price)
Note. Reference categories: Mefenamic acid (API); Suppository (dosage form); Modified release (modification); Generic (brand); University Hospital (hospital type); Independent agency (ministry). Coefficients for categorical dummies represent approximate percentage price differences relative to the reference category. Year fixed effects included in all models.
***p < 0.001.
**p < 0.01.
*p < 0.05.
To account for within-hospital correlation in procurement transactions, all reported standard errors are clustered at the hospital level. Robust clustered standard errors are larger than conventional OLS standard errors and provide a more conservative basis for statistical inference. All empirical findings reported in Section 4.2 are robust to this correction. To assess multicollinearity among the explanatory variables, between API class, dosage form, and formulation modification, Variance Inflation Factors (VIFs) were computed for the full model specification. All VIF values fell below 10 with a mean VIF of 3.8, which indicates that multicollinearity does not pose a material threat to the stability or interpretability of the estimated coefficients.
The high adjusted R 2 in Model V of 0.953 reflects the substantial and genuine explanatory power of product-level attributes in a hedonic pricing framework. In cross-sectional hedonic models with rich product fixed effects, R 2 values above 0.90 are commonly reported in the pharmaceutical pricing literature and are consistent with the theoretical expectation that product composition is the primary determinant of observed price variation. This is not evidence of overfitting in the predictive sense because the model is not trained to minimize out-of-sample prediction error but rather to estimate the marginal implicit prices of product attributes using the full sample. All reported associations are based on observational cross-sectional data and should not be interpreted as evidence of causal relationships. The absence of random assignment or a credible natural experiment precludes causal inference. The regression coefficients reflect conditional correlations between procurement prices and their covariates while controlling for observed product and institutional characteristics.
4.2. Findings
Model I establishes a statistically significant negative relationship between procurement quantity and unit price which is superficially consistent with a downward-sloping demand curve. However, this relationship collapses entirely once product characteristics are introduced in Model III where the coefficient becomes negligible and statistically indistinguishable from zero. The reversal is explained by product-mix composition. High-volume procurement records in this dataset are dominated by cheap generic tablets such as paracetamol and ibuprofen, so the apparent negative association in Models I and II conflates compositional differences with a genuine price-quantity relationship. After controlling for what is being purchased, procurement volume exerts no independent effect on price. This finding is consistent with price-taking behavior in a regulated B2G market where unit prices are determined through tendering protocols and reference price ceilings rather than through volume-driven bilateral negotiation.
Market competition, proxied by the count of registered TPU competitors for each dosage regimen, is statistically significant across models but economically negligible. In Model II, each additional competitor is associated with a price reduction of approximately 2.4%. Once internal product attributes are controlled in Model III, however, the sign reverses to a small positive association of 1.1% per additional competitor. This counterintuitive reversal is consistent with product heterogeneity within TPU categories. While products that attract many competitors are often those with well-established therapeutic value and market demand, these characteristics are also associated with higher prices once composition is held constant. The consistently low R 2 in Models I and II of approximately 0.35 confirms that market-level forces leave the vast majority of price variation unexplained.
The addition of internal product attributes in Model III produces a dramatic improvement in explanatory power where the R 2 rises from 0.35 to 0.953. This step-change demonstrates that pharmacological class, dosage form, formulation technology, and brand status are the dominant drivers of price variation in the Thai acute pain medication market. By contrast, market-level factors contribute comparatively little to observed price dispersion.
Among API classes, selective COX-II inhibitors command extraordinary premiums. Etoricoxib carries a price premium of approximately 2,122% which means it is procured at more than 22 times the per-unit price of the reference drug. This reflects its status as a novel inhibitor with limited generic penetration during the study period. Parecoxib also presents a distinct case because it faces near-monopoly market conditions for its principal dosage regimen. Celecoxib sustains a large premium despite the availability of generic versions which is consistent with lingering brand loyalty and the persistence of physician familiarity effects in drug selection even in a regulated procurement context. Paracetamol combined with tramadol occupies a distinctly higher price tier than any non-opioid paracetamol product, which is attributable primarily to tramadol’s classification as a controlled substance under Thai law. Standard paracetamol and diclofenac are priced substantially below reference, which is consistent with their status as mature, high-volume generics facing intense competition across multiple manufacturers.
Dosage form effects are of comparable magnitude to API effects and are clinically interpretable. Relative to suppositories, tablets command the deepest discount in the entire model reflecting fundamental production-cost advantages and the extraordinary intensity of generic competition. Spray formulations and eye drops occupy the highest price tiers among non-tablet forms which reflects sterile manufacturing requirements and specialized packaging that substantially raise per-unit costs. Formulation modification effects reveal a clear technology premium hierarchy where modified-release technology is established as the most expensive formulation type. The brand premium remains large and remarkably persistent as original brand products are priced, on average, 101.8% above their generic counterparts even after controlling exhaustively for all other observable product characteristics.
Model IV introduces hospital-type and ministry-affiliation dummies alongside the product controls from Model III. The hospital-type results are striking in their uniformity. Not one of the facility-class coefficients achieves statistical significance at the 5% level. The scale and clinical specialization of the procuring hospital confer no detectable bargaining advantage over drug suppliers in this regulated market.
Ministry affiliation presents a sharply contrasting picture. Hospitals affiliated with the Ministry of Defense are associated with procurement prices 28.8% higher than the reference for equivalent products. Hospitals under the Ministry of Interior are associated with prices 16.5% higher. Ministry of Higher Education hospitals are associated with prices 14.5% higher. GPO-channeled procurement is associated with a 9.6% lower unit price. The Ministry of Public Health, by contrast, procures at prices statistically indistinguishable from the independent agency reference.
Hospitals affiliated with the Ministry of Defense are associated with procurement prices 28.8% higher than the reference for equivalent products. Hospitals under the Ministry of Interior are associated with prices 16.5% higher. Ministry of Higher Education hospitals are associated with prices 14.5% higher. GPO-channeled procurement is associated with a 9.6% lower unit price.
The institutional associations identified above, in particular the price premiums associated with non-MoPH ministry affiliation, should be interpreted as conditional associations rather than causal estimates. Several sources of residual confounding may influence the observed differentials. First, hospitals affiliated with non-MoPH ministries may have systematically different patient case-mixes, therapeutic protocols, or urgency-driven procurement patterns that lead them to purchase higher-cost formulations or smaller quantities per transaction. Second, ministry-specific administrative structures may channel procurement through different supplier networks not fully captured by the GPO status or TPU competition variables included in the model. The ministry affiliation coefficients should therefore be understood as capturing the residual price differential associated with institutional affiliation after controlling for observable product and hospital characteristics, rather than as evidence of a direct governance effect on prices.
The introduction of GPO procurement status in Model V provides evidence that centralized procurement delivers a near-10% price reduction uniformly across all product categories and institutional settings. The pattern of attenuation in ministry premiums when GPO status is added confirms that part of the observed ministry premiums reflects differential rates of GPO utilization. Nevertheless, the residual ministry premiums that remain in Model V are substantial, indicating that GPO adoption alone is insufficient to equalize procurement prices across ministries. The residual premium represents genuine governance and infrastructure deficits, differences in procurement planning, and institutional accountability that cannot be resolved merely by directing more transactions through the GPO channel.
Year dummy fixed effects reveal that apparent price increases in 2020–2021 were not driven by genuine inflation but by compositional shifts toward higher-cost medicines during the pandemic. By 2022–2023, moderate but significant year effects re-emerge which is consistent with genuine inflationary pressures in pharmaceutical supply chains during the post-pandemic period.
In conclusion, three principal findings emerge from this analysis, each carrying distinct implications for pharmaceutical policy. First, internal product characteristics—API class, dosage form, formulation technology, and brand status—account for approximately 95% of all price variation in the Thai acute pain medication market. Selective COX-II inhibitors command premiums exceeding 1,000% over reference-class generics; modified-release formulations attract the largest formulation premium; and original-brand products sustain a premium of over 100% relative to generics even after exhaustive product controls. These results establish that pharmaceutical innovation and product differentiation are overwhelmingly the dominant drivers of procurement price in this market. Policymakers seeking to manage pharmaceutical expenditure must therefore engage directly with product-level substitution and formulary management strategies, rather than relying solely on market-level instruments such as promoting competition or increasing procurement scale.
Second, conventional market mechanisms such as competitive pressure and procurement volume do not meaningfully reduce prices in this regulated B2G environment. Neither the number of registered competitors nor the quantity procured exerts an economically significant independent effect on price once product characteristics are held constant. This finding underscores the fundamental limits of market-based cost-containment approaches in a setting where prices are governed by government tendering and reference pricing, and where product differentiation within the analgesic class is both pharmacologically and regulatorily substantial.
Third, institutional governance factors—specifically ministry affiliation and GPO procurement status—generate significant and policy-actionable price effects that operate independently of product characteristics. Hospitals affiliated with non-MoPH ministries pay premiums of approximately 13–29% for the same drugs, reflecting procurement governance deficits that differ systematically across ministerial structures. GPO procurement delivers a consistent and significant 9.6% price reduction but only partially offsets ministry-level premiums, pointing to the need for governance reform targeted at procurement infrastructure within non-MoPH ministries rather than channel-based interventions alone. These residual institutional price differences—conditional on product composition—may have relevant implications for pharmaceutical expenditure management across Thailand’s fragmented public health system, as elaborated in Section 5.
5. Conclusion and Policy Recommendations
5.1. Conclusion
This study applies a hedonic regression framework to 26,652 internal medication procurement transactions recorded by Thailand’s Comptroller General’s Department between 2019 and 2024. By focusing on the acute pain medication market, the analysis isolates and quantifies the contributions of internal product attributes, institutional governance factors, and market-level forces to observed price variation.
The central finding is that internal product characteristics account for approximately 95% of all price variation in this market. Within this set of characteristics, pharmacological innovation is the single most consequential dimension. Selective COX-II inhibitors command premiums exceeding 1,000% over mature generic NSAIDs, with etoricoxib priced at more than 22 times the reference on a per-unit basis. Modified-release technology carries the largest formulation premium and original brand products sustain a price more than double that of generic equivalents even after exhaustive product controls [15]. This is consistent with persistent information asymmetries in prescribing that regulation has not fully corrected.
The study also produces findings that challenge conventional assumptions about how markets and institutions operate in pharmaceutical procurement. Contrary to the predictions of standard demand theory, procurement volume exerts no meaningful effect on unit price once product composition is controlled. This indicates that scale-based bargaining leverage does not operate in this regulated B2G setting. Similarly, market competition, as measured by the number of registered TPU competitors, contributes negligibly to price reduction. This underscores the limits of relying on competitive entry where reference pricing and tendering set effective price ceilings. Hospital size and clinical specialization prove equally inert as bargaining determinants.
To be precise about the hierarchy of price determination, product-level attributes account for the overwhelming majority of price variation with an R 2 contribution of approximately 0.91 from product dummies alone. Institutional and governance factors explain a statistically significant but quantitatively secondary share of residual price variation. Ministry affiliation remains a powerful and consistently significant predictor where hospitals under the Ministries of Defense, Interior, and Higher Education are associated with price premiums of 13–29% for equivalent products.
The central policy implication is not that governance replaces product composition as a driver of pharmaceutical costs but that within any given product category, governance arrangements generate meaningful and addressable price differentials. Specifically, procurement fragmentation across ministries and limited use of centralized GPO channels are associated with these disparities. While the Government Pharmaceutical Organization procurement channel delivers a reliable 9.6% price reduction, GPO utilization alone is insufficient to close the institutional price gap facing non-MoPH ministries. Strengthening procurement coordination across ministries and expanding GPO coverage represent tractable policy levers for reducing this residual institutional price variation.
5.2. Policy Recommendations
Four targeted recommendations follow from the empirical findings. First, formulary stewardship should be strengthened to address the dominant role of pharmacological class in price determination. The premiums commanded by selective COX-II inhibitors, particularly etoricoxib and parecoxib in routine outpatient settings, are difficult to justify on cost-effectiveness grounds relative to non-selective alternatives. Tighter therapeutic indication guidelines for these agents, mandatory generic substitution for off-patent molecules, and hospital-level formulary audit mechanisms would address the largest source of avoidable pharmaceutical expenditure identified in this analysis.
Third, the 13–29% institutional price premiums paid by non-MoPH ministry hospitals reflect procurement governance deficits that can be partially addressed through cross-ministerial harmonization. Permitting or mandating hospitals affiliated with the Ministries of Defense, Interior, and Higher Education to participate in MoPH national tender agreements and to combine with targeted efforts to increase GPO utilization in those settings would directly narrow the institutional price gap identified in Models IV and V.
Fourth, the CGD administrative dataset is a substantially underutilized resource for pharmaceutical price surveillance. Formalizing access to CGD transaction records for ongoing price monitoring and tracking unit price trends by product, ministry, and hospital type would enable early detection of price drift and provide the evidentiary basis for targeted corrective action. Published price benchmarks would also reduce the information asymmetries that allow institutional premiums to persist.
5.3. Limitations
Three limitations merit acknowledgement. First, the analysis is restricted to a single therapeutic category and cannot be assumed to be generalized to cover drug classes with different innovation profiles, such as oncology agents or biologics, where the determinants of B2G pricing may differ substantially. Extending the hedonic framework to high-cost specialty medicines is a priority for future work. Second, the absence of clinical outcome data at the transaction level means the study can identify what procurement prices imply about the implicit valuation of drug attributes but cannot assess whether those valuations are consistent with cost-effectiveness benchmarks. Integration with HTA data would enable a value-based evaluation of the premiums recovered here. Third, the CGD dataset covers only civil servant medical expense reimbursements (CSMBS) and does not capture procurement under the Universal Coverage Scheme (UCS) administered by the NHSO, limiting direct generalizability to the full public sector pharmaceutical market. Future research comparing price dynamics across these parallel procurement systems would provide a more complete picture of pharmaceutical expenditure governance in Thailand.
Two proxy limitations also deserve explicit acknowledgement. The competition measure—the number of registered Trade Pharmaceutical Unit (TPU) codes per dosage regimen—captures the breadth of product registration rather than actual competitive dynamics at the tendering stage. Realized competitive pressure is more accurately reflected by the number of active bidders, the distribution of bids, or the Herfindahl-Hirschman Index of winning supplier shares—data that are not available in the CGD dataset. The null result for TPU count in Models III–V should therefore be interpreted as the absence of a registration-breadth effect, not as evidence that competition is irrelevant to pharmaceutical pricing more broadly. Similarly, the null result for procurement quantity may reflect the regulatory constraints of public tendering systems where price ceilings and reference prices limit the scope for volume-based negotiation. Whether quantity discounts operate more substantially in less constrained procurement environments remains an open empirical question.
Regarding the scope of generalizability, the findings should be interpreted within the specific therapeutic scope of this study. Analgesic and NSAID products are predominantly mature generics with well-established supply chains and competitive manufacturers. Therapeutic categories characterized by higher clinical uncertainty—including oncology, biologics, and high-cost innovative drugs—are governed by markedly different pricing logics involving patent exclusivity, outcome-based reimbursement agreements, and complex multilateral negotiations. Institutional governance effects identified here may operate differently, or more intensely, in those settings, and their investigation constitutes an important direction for future research.
Footnotes
Acknowledgements
The authors express their sincere gratitude to the Comptroller General’s Department (CGD) of Thailand for providing access to the medication procurement transaction records used in this study. We also thank the National Institute of Development Administration (NIDA) for supporting the research environment necessary for this analysis. Any remaining errors or omissions are the sole responsibility of the authors.
Author Contributions
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.