DEGREE PROJECT IN TECHNOLOGY AND ECONOMICS, SECOND CYCLE, 30 CREDITS STOCKHOLM, SWEDEN 2019

Government favoritism in public procurement: evidence from Romania

DANIEL PUSTAN

KTH ROYAL INSTITUTE OF TECHNOLOGY SCHOOL OF INDUSTRIAL ENGINEERING AND MANAGEMENT

This page intentionally left blank

Government favoritism in public procurement: evidence from Romania

by

Daniel Pustan

Master of Science Thesis TRITA-ITM-EX 2019:728 KTH Industrial Engineering and Management Industrial Management SE-100 44 STOCKHOLM

Regeringens favorisering inom offentlig upphandling: fallet Rumänien

Daniel Pustan

Examensarbete TRITA-ITM-EX 2019:728 KTH Industriell teknik och management Industriell ekonomi och organisation SE-100 44 STOCKHOLM

Master of Science Thesis TRITA-ITM-EX 2019:728

Government favoritism in public procurement: evidence from Romania

Daniel Pustan

Approved Examiner Supervisor 2019-12-31 Matti Kaulio Kristina Nyström Commissioner Contact person

Abstract

In Romania, the consideration that politicians use their influence to control the public procurement market is axiomatic. It is no surprise that the country ranks high in perception- based surveys or the low participation of firms on the procurement market. The more difficult task is to demonstrate the existence of restrictions to procurement contracts in order to benefit preferred companies. That is, to measure the extent to which the market is captured by favored companies.

Employing data on all public procurement contracts in Romania for the period 2009 – 2015, this paper examines government favoritism in public procurement exerted by political parties. Using a dynamic panel data approach (Dávid-Barrett and Fazekas 2019), the companies are classified based on their winning pattern with respect to government change. Favoritism is observed if winning companies within the government period are also associated with a higher risk of corruption measured by two alternative approaches.

The findings confirm that procurement market is captured in a low to moderate proportion (24%) and that the market display patterns of systematic favoritism. This may signal certain progress registered by Romania to combat political corruption. Arguably, the insensitivity of perception indicators with respect to this progress is, at least partly, due to media coverage of the on-going corruption investigations related to the past.

Key-words Grand corruption, High-level corruption, Political corruption, Clientelism, Favoritism, Governance, Corruption measurement, Public procurement, Romania

Examensarbete TRITA-ITM-EX 2019:728

Regeringens favorisering inom offentlig upphandling: fallet Rumänien

Daniel Pustan

Godkänt Examinator Handledare 2019-12-31 Matti Kaulio Kristina Nyström Uppdragsgivare Kontaktperson

Sammanfattning

I Rumänien, finns det en allmän uppfattning om att politiker använder sitt inflytande för att kontrollera den offentliga upphandlingsmarknaden. Det är ingen överraskning att landet rankas högt i perceptionsbaserade undersökningar rörande korruption eller att företags deltagande inom upphandlingsmarknaden är lågt. Ett svårare uppdrag är dock att påvisa och bevisa förekomsten av begränsningar kring upphandlingskontrakt med syfte att gynna vissa företag. Med andra ord så föreligger en utmaning att, genom mätning, påvisa i vilken utsträckning marknaden inkluderar favoriserade företag kontra hur den exkluderar övriga företag.

Med hjälp av uppgifter om samtliga kontrakt gällande offentlig upphandling i Rumänien under åren 2009 – 2015, undersöker denna avhandling regeringens och dess politiska partiers favorisering. Företagen i upphandlingarna klassificeras med hjälp av dynamiskpaneldata (Dávid-Barrett och Fazekas 2019), baserat på des vinnande mönster kopplat till regeringsbyte. Favorisering kan observeras om vinnande företag inom regeringsperioden även är förknippade med en högre risk för korruption som mätts genom två alternativa metoder.

Resultaten bekräftar att upphandlingsmarknaden fångas i en låg till måttlig andel (24%) av favoriserade företag och att marknaden visar mönster av systematisk favorisering. Resultaten kan dock signalera om visst framsteg som Rumänien har uppnått för att bekämpa politisk korruption. Det kan argumenteras att perceptionsbaserade indikatorer fångade inte upp dessa framsteg, åtminstone delvis, på grund av mediatäckningen av pågående korruptionsutredningar i Rumänien relaterade till det förflutna.

Nyckelord Stora korruption, korruption på hög nivå, politisk korruption, klientelism, favorisering, styrning, korruptionsmätning, offentlig upphandling, Rumänien Government favoritism in public procurement: evidence from Romania ​ ​7 ​

Table of Contents

1. Introduction 13

2. Literature review 18

2.1 Definition and classification 18

2.2 Measures of corruption 23

3. Institutional framework 27

3.1 Legislative setting 28

3.2 Public procurement institutions 30

3.3 Government characteristics 31

4. Method 34

5. Data 41

6. Results 44

7. Robustness analysis 54

8. Conclusions 58

References 61

Annexes 70

KTH ROYAL INSTITUTE OF TECHNOLOGY 8 Daniel Pustan ​ ​

List of figures

Figure 1 The bi-dimensional clientelistic structure

Figure 2 The comprehensiveness of regulation in Romania

Figure 3 Timeline of cabinets and parties for the period 2009 - 2015

Figure 4 Mean regression residuals by two-percentiles of relative procedure period (contract date - call for tender date), linear prediction

Figure 5 Mean regression residuals by two-percentiles of relative notification of award period (award notice date - contract date), linear prediction

Figure 7 CRI-arithmetic by company group, half-year period

Figure 8 Combined market shares of company types, half-year period

Figure 9 CRI arithmetic compared to perception-based surveys (CPI, WGI, GCI) ​ ​ Figure 10 CRI-PCA compared to perception-based surveys (CPI, WGI, GCI)

Figure 11 Single bid compared to perception-based surveys (CPI, WGI, GCI)

Figure 12 Outliers for relative contract value < 2

Figure 13 CRI-PCA patterns by company group, half-year period ​ ​ Figure 14 CRI-arithmetic patterns by company group, one year period

Figure 15 CRI-PCA patterns by company group, one year period

Figure 16 CRI-arithmetic patterns by company group, changes in legislation

Figure 17 CRI-PCA patterns by company group, changes in legislation

KTH ROYAL INSTITUTE OF TECHNOLOGY Government favoritism in public procurement: evidence from Romania ​ ​9 ​

List of tables

Table 1 Amendments to GEO 34/2006 during the studied period

Table 2 Romania’s government cabinets during 2009-2015

Table 3 Main characteristics of the dataset

Table 4 System GMM linear dynamic panel regression explaining company market success

Table 5 Logit regression results on contract level, 2009-2015, number of winners ≥ 2

Table 6 CRI component weights

Table 7 Bivariate Pearson correlation of CRI and single bid with perception survey indicators

Table 8 Linear regression explaining relative contract value and CRI

Table 9 Summary of selected studies using objective indicators of corruption

Table 10 Corruption techniques and indicators

Table 11 Summary of selected studies on firm’s political connections

Table 12 Descriptive statistics

KTH ROYAL INSTITUTE OF TECHNOLOGY 10 Daniel Pustan ​ ​

Abbreviations

ANAP National National Public Procurement Agency ANRMAP National Authority for Regulating and Monitoring Public Procurement BI Business International CIN Company Identification Number CNSC National Council for Solving Complaints CPI Corruption Perception Index CPV Common Procurement Vocabulary CRI Corruption Risk Index CRI Corruption Risk Index CVM Cooperation and Verification Mechanism DLAF Department for the Fight Against Fraud DNA National Anticorruption Directorate FDI Foreign Direct Investments GCI World Economic Forum GEO Government Emergency Ordinance ICRG The International Country Risk Guide IMD The Institute for Management Development NAPP National Agency for Public Procurement NGO Non governmental organization NUTS Nomenclature of Territorial Units for Statistics PC Conservative Party PCA Principal Component Analysis PDL Democrat Liberal Party PNL National Liberal Party PSD Social Democratic Party SEAP Electronic System for Public Procurement TED Tenders Electronic Daily UCVAP Unit for Coordination and Verification of Public Procurement UDMR Democratic Alliance of Hungarians in Romania UNPR National Union for the Progress of Romania USL Social Liberal Union WCY World Competitiveness Yearbook WEF World Economic Forum WGI Global Competitiveness Index

KTH ROYAL INSTITUTE OF TECHNOLOGY Government favoritism in public procurement: evidence from Romania ​ 11 ​ ​

Acknowledgements

On the very outset of this paper, I would like to extend my sincere and heartfelt obligation towards all the personages who have helped me in this endeavor. Without their active guidance, help, cooperation and encouragement, I would not have made headway in the project.

I would like to express my gratitude to my supervisor, Kristina Nyström for her patience and valuable advice for structuring the ideas of this paper. I am extremely thankful to Mihály Fazekas - co-author of the paper on which the methodology of this study relied on - for the kind responds to my questions.

I extend my gratitude to Sebastian Buhai for the precious guidance related to the topic of this research and to Hayk Sargsyan for his useful comments and encouragement when most needed.

I am deeply indebted to my former manager, Larisa Lăcătuș, for creating the conditions to expand my knowledge alongside my professional development. On the same note, I gratefully acknowledge the assistance of Karin Asplund with the Swedish translation.

Lastly but not least, my highest appreciation goes to my wife, Maria, who experienced firsthand this long and strenuous process. Your limitless understanding and unconditional support made it possible for this journey to end successfully.

Any omission in this brief acknowledgment does not mean lack of gratitude.

st Stockholm, December 31 ​, 2019 ​

Daniel Pustan

KTH ROYAL INSTITUTE OF TECHNOLOGY 12 Daniel Pustan ​ ​

1. Introduction

Corruption is estimated to cost the European Union economy €120 billion per year which is almost the equivalent of the EU annual budget (European Commission 2014). The true cost of ​ ​ corruption, however, includes arguably inestimable social costs such as: poorer income distribution, environmental protection as well as undermining trust in democratic institutions (European Commission n.d.). ​ ​

Thirty years after communism fell in Romania, it seems that the fight against corruption remains an ongoing battle which affects the country on many levels. In terms of country perception, Romania ranks amongst the most corrupt countries in the European Union (Hall 2019) while favoritism appears to be the rule for the majority of government decisions (Schwab ​ 2016). Additionally, Romania is held back from joining the free movement Schengen area ​ because of the lack of progress in combating corruption as reflected by the Cooperation and Verification Mechanism (CVM) .1

The civil society became harshly conscient of the fatal consequences of corruption following Colectiv nightclub fire in November 2015. A fire that tragically ended the lives of many, only because, allegedly, corrupt practices led to systematic violation of the rules which, if respected, could have saved lives. Protests that followed denouncing political elites’ tolerance or association to corruption led to Victor Ponta’s resignation, the first Prime Minister accused of corruption while still in charge (BBC 2015). ​ ​

Considering the external and internal pressure - by means of European Commission reports and civil protests - everything indicated that the message was understood and corruption would be irremediably defeated in a battle with the government as protagonist.

Nothing could be further from reality. Liviu Dragnea, chief of the ruling Social Democratic Party (PSD) and president of the Chamber of Deputies in Parliament who already had a suspended sentence, was risking prison sentence in his ongoing corruption trial. (France24 2018). In an ​ ​ attempt to spare the PSD chief and other high profile politicians from the ongoing investigations or prison sentences, the Romanian Government issued an Emergency Ordinance 13/2017 (to enter into effect 10 days after publication in the Official Gazette).

1 T​ he Cooperation and Verification Mechanism is the mechanism that European Commission uses to monitor reform implementation in the area of justice and rule of law.

KTH ROYAL INSTITUTE OF TECHNOLOGY Government favoritism in public procurement: evidence from Romania ​ ​13 ​

Massive protests followed against what was perceived as legalization of corruption acts inducing the government to abolish the controversial Act prior to its entry into force. Should the Emergency Ordinance came into effect just for one day, Liviu Dragnea could have escaped the three and a half years of prison that he was sentenced to in May 2019 (BBC 2019). ​ ​

Unfortunately, political corruption does not represent just a few isolated cases. Only in 2017, other 12 high level officials were prosecuted among which: ministers, deputies, Secretary of State, Senators and government Secretary-General National (Anticorruption Directorate Activity Report 2017). ​ ​

In this context, public procurement is especially vulnerable to corruption (European Commission 2014; OECD 2016). Perception based reports advance that public procurement is ​ ​ ​ ​ manipulated in proportion of 80% (Euractiv 2014). This severe trust deficit is probably one of ​ ​ the main causes of the low share of firms taking part in the bidding process. In fact, the number of unique bidding companies during the seven years period of observation was 15,568 (1.3%) out of a 1.2 million active companies (Oprea 2017). By comparison, in Sweden the number of ​ ​ companies participating in public procurement bids during 2013 was 212,000 representing 18.80% of the total number of registered companies (Lukkarinen, Morild and Jönsson 2015). ​ ​

Furthermore, the low number of referrals received from public administrations (9% in 2017) as well as the rare occasions in which local authorities decide not to become claimants in order to recover incurred losses (National Anticorruption Directorate 2017), might be an indicator of ​ ​ participation in a larger network of favoritism or the resignation that such networks exist and there is not much to do against them.

The overflow of the anti-corruption theme outside the institutional framework into society has caused deep rifts among many walks of life. In these last years, many have seen in the Social Democrat Party (PSD) the very essence of corruption, a party that was ripping the resources of the state shamelessly. At the same time, for others it represented the ultimate guardian for the ordinary citizen ripped off by multinationals in their run to make profits at any cost.

Perceptions became less reliable while sliding towards a binary, oversimplified division of reality - between “us” and “they” (Bogdan 2018). Polarization deepened and fact based arguments lost ​ ​

KTH ROYAL INSTITUTE OF TECHNOLOGY 14 Daniel Pustan ​ ​ prominence in an ideological battle, arguably, between traditional and modern values (David 2015). ​

The necessity to regain a common ground between the polarized sides became obvious. This, in turn, required a measurement tool that focuses on objective behavior rather than opinions and perceptions.

For a long time, measurement of corruption relied mostly on perception based indicators. However, perception may not be of much help in this scenario. Furthermore, the little variation of indicators makes them susceptible of not being responsive to change, in spite of significant changes in Romania from one year to another. (Oman and Arndt 2006). ​ ​

Companies associated with high profile politicians are often successful on the procurement market, earning the label of “companies subscribed to contracts with the State'' (Befu 2015; ​ ​ Marcovici 2015; Andrei 2018). In fact, Mungiu-Pippidi (2015) highlights that political ​ ​ ​ ​ ​ ​ connections are the main indicator of favoritism in Eastern Europe. But do high level politicians use their influential capacity to divert contracts towards favored companies?

According to Gherghina and Volintiru (2017) clientelistic model, parties are interested to engage ​ ​ in reciprocal relationships by redirecting public contracts towards favored private contractors in exchange for benefits - which in most cases are monetary. Parties will use this financing to sustain or increase the electoral support and consequently consolidate their power.

Figure 1 The bi-dimensional clientelistic structure ​ Source: Gherghina and Volintiru 2017, p. 122 ​ ​

KTH ROYAL INSTITUTE OF TECHNOLOGY Government favoritism in public procurement: evidence from Romania ​ ​15 ​

Therefore, if the answer is affirmative as anticipated, what size of the procurement market is captured by the political elites?

To my knowledge, the only research on Romania’s procurement market with objective indicators were performed by Doroftei (2016) and Doroftei and Dimulescu (2015). They examined ​ ​ ​ ​ contracts over €1,000,000 in the construction sector for the period between 2007 and 2013. Both papers conclude that single bidding and political connections explained 32% of the number of contracts awarded during the entire research period and 44% in pre-election years.

Using a more sophisticated measurement of corruption, Dávid-Barrett and Fazekas (2019) ​ ​ found that 50%-60% of procurement market in Hungary suffer from systematic favoritism while only 10% of companies were politically favored in the United Kingdom. Their approach consisted of two steps: (1) analyze the influence that government change has on companies procurement outcome and (2) investigate the risk of corruption associated with the winning of contracts. The sample comprised period 2009-2012 and was limited to contracts: (i) awarded by national authorities, (ii) above the mandatory EU-threshold and (iii) only applied to competitive markets.

Following Dávid-Barrett and Fazekas (2019) method, this paper explores empirically to what ​ ​ extent the governmental political parties in Romania exert control over the public procurement process; in other words, it quantifies the level of government favoritism in Romania. The analysis extends to all public procurement contracts in Romania during 2009-2015 period. The sample was not restricted to contracts awarded only by national central administration. This decision was taken considering the centralization of power in Romania both at the level of public administration as well as that of party organization.

The contribution to the research literature is fourfold. First, to my knowledge it is the first time ​ ​ to apply Dávid-Barrett and Fazekas (2019) innovative methodology on public procurement in ​ ​ Romania. Secondly, it expands the method by performing a cross-governmental analysis with a ​ ​ focus on the main party in government. Thirdly, this study introduces the Principal Component ​ ​ Analysis as an alternative to simple average method for weighting the indicators composing the Corruption Risk Index. Lastly, it provides a corrected large dataset suitable for scientific ​ ​ research.

KTH ROYAL INSTITUTE OF TECHNOLOGY 16 Daniel Pustan ​ ​

The method follows three steps. First, companies are classified based on the residuals after ​ ​ estimating the panel regression using Arellano-Bond system GMM transformation. Second, ​ ​ Corruption Risk Index is calculated both as: (1) as a weighted average (Dávid-Barrett and Fazekas 2019) and (2) using the Principal Component Analysis approach. Thirdly, the market ​ ​ ​ ​ share of aggregated suspicious companies is calculated and companies are aggregated by group to identify patterns of systematic favoritism.

This paper analyzes a dataset of over 650,000 observations and finds that 24% of the Romanian procurement market is associated with systematic government favoritism. However, it is noteworthy, that the study only captures specific clientele for each of the two parties. It does not reflect favoritism in situations where companies buy influence from both parties or public administration officials that withstand government change.

Furthermore, highly innovative companies may be mislabeled as “surprise winners” as both exhibit similar pattern behavior (Dávid-Barrett and Fazekas 2019). This risk is reduced for this ​ ​ specific dataset by the presence of “lowest price” award criterion in 94% of the procedures during the analyzed period. Typically, due to the high costs of research, innovative companies do not compete for price and they are not expected to take part in these tenders. Two other limitations are discussed further.

The Common Procurement Vocabulary (CPV) codes are the basis for the CPV division (2 digit level of CPV codes) and CPV product group (3 digit level of CPV codes) which in turn serve as proxies for the market. Although generally accepted as a market indicator, it is not a perfect instrument due to flaws in the CPV code structure such as: (1) inconsistent hierarchical tree-structure, (2) inconclusive classification levels, (3) unmatching between subordinate and superordinate level and (4) not mutually exclusive codes (Attström et al. 2012). ​ ​

Aggregating the data based on a half year period “is optimal for retaining a high level of granularity while also taking into account the erratic character of many public procurement markets” (Dávid-Barrett and Fazekas 2019, p. 13). However, the random length of the cabinets’ ​ ​ period does not perfectly match the half year aggregation. To circumvent this problem, an alternative model was analyzed where the cabinet variable was replaced by variables consisting in legislation changes. Both models yield comparable results.

KTH ROYAL INSTITUTE OF TECHNOLOGY Government favoritism in public procurement: evidence from Romania ​ ​17 ​

In view of the ensuing inefficient distribution and allocation of resources, corruption and in particular clientelism constitutes, perhaps, the antithesis of sustainable development. Clientelism skews the distribution of state resources (Ferreira Rubio and Andvig 2019) with ​ ​ potential negative consequences in all three aspects: environmental, social and economic. Among the social consequences are the increase of inequality (Ferreira Rubio and Andvig 2019) ​ ​ and erosion of confidence in institutions. Corruption affects the most vulnerable groups as infrastructure projects may be vastly financed to steer money towards favorite companies in detriment of education and health (Mungiu-Pippidi 2015). ​ ​

Corruption is also negatively correlated to the growth in genuine wealth per capita. This is clearly illustrated by Aidt (2010) with Denmark. If levels of corruption in Denmark would ​ ​ suddenly increase to those of the countries that make it to the bottom of Corruption Perception Index, the development would become unsustainable.

Narrowing it further down, the implications of this paper on sustainability - described as making decisions in the present so as not to impair future real incomes (Markandya and Pearce 1988) - ​ ​ are highly practical. The economic impact is given by the importance of public procurement which rises to almost 50% of public spending in developing countries. Within the social aspect, it represents an invitation for society to focus on facts in an attempt to find a common ground where this was lost. It promotes the equality and importance of individuals in the social system while promoting the ranking of ideas. The positive impact on the environment stems from the implicit responsible and efficient use of resources as alternative to particularism.

The remaining of the paper is structured in seven sections. First, the literature is reviewed with ​ special emphasis on the various definitions of corruption and the challenges encountered in measuring it. Second, the institutional framework is discussed including the changes in legislation during the period, the specific institutions related to public procurement and the composition of government. Following, the method section describes the process of suspicious companies identification based on their market success and the subsequent analysis of the conditions under which they as shown by the Corruption Risk Index. The remainder of the paper is organized as follows: data, results, robustness analysis and conclusions.

KTH ROYAL INSTITUTE OF TECHNOLOGY 18 Daniel Pustan ​ ​

2. Literature review

2.1 Definition and classification

Corruption is a term used to cover a wide range of different practices. It appears that to some extent the term is framed by prevalent cultural norms (Werner 1983) as well as depending on ​ ​ the context (Johnston 1996). The difference is obvious when comparing the more developed ​ ​ American democracy with the Romanian mindset of government. The American view strives to limit the authority and divide the power (Huntington 1968). By contrast, the Romanian ​ ​ dominant psychological profile is that of a collectivist person and concentration of power (David 2015). ​

In an interview at the Romanian Business Leaders Summit (Roșca 2018), David argues that ​ ​ favoritism and nepotism are embedded in the Romanian culture and what is considered corruption from a Western perspective, it is simply perceived as a proof of loyalty. David continues by exemplifying with the implementation of a Western institution in Romania - that of a parliamentary cabinet - a parliamentary with a collectivist profile will hire a person who they can trust most - a relative - instead of the best expert (Roșca 2018). ​ ​

This shows that conflict of interests as a specific form of corruption, may also be the unintentional result of institution implementation within a culture different than the one where the institution was born. However, nepotism as a cultural accepted practice is not specific only to Eastern Europe. In India and Africa it is not only acceptable but expected (Leys 1965; Bayley ​ ​ 1966). ​

Huntington (1968) goes as far as to suggest that corruption is a hardly avoidable externality of ​ ​ modernization [of institutions]. Scholars supporting the “grease the wheels” hypothesis highlight potential malfunctions or inefficiencies that corruption may help circumvent the functioning of the institutions and government policies. It can (1) increase the quality of the service (Leys 1965; Bayley 1966) - by keeping the underpaid talented employees in the public ​ ​ ​ ​ service, (2) speed-up processes like waiting in the queue (Huntington 1968; Lui 1985) and (3) ​ ​ ​ ​ foster efficiency as an imperfect auction system in competitions for a limited number of resources or instead of non optimal mechanical quota based systems (Leff 1964; Bayley 1966), ​ ​ ​ ​ (4) act as an insurance against bad public policies ensuring predictability and give a sense of being in control during transition period (Leff 1964; Bayley 1966), (5) non-violent manner to ​ ​ ​ ​ access the establishment and (6) provide political representation or unity in spite of

KTH ROYAL INSTITUTE OF TECHNOLOGY Government favoritism in public procurement: evidence from Romania ​ ​19 ​ irreconcilable ideologies or interests, potentially higher and better investment (Leff 1964; Bayley ​ ​ 1966; Werner 1983). ​ ​ ​

Most empirical research, however, does not support the claim of positive effects of corruption. Dependency on grafts acts as an incentive to deliberately slow down or limit the provision of the service and to block the access to other competent workers in order to secure a large stream of cash (Lui 1985; Kurer 1993). Furthermore, in order to compensate for the bribe, the firm can ​ ​ ​ ​ provide products of lower quality (Rose-Ackerman 1997). ​ ​

The “sand in the wheels” effect of corruption is more obvious when looking at its impact on economic growth. Various studies confirmed the negative relationship between corruption and growth consequence of a decrease in private investment (Mauro 1995) or the quality of the ​ ​ infrastructure (Tanzi and Davoodi 1998) even in countries with a weak rule of law and inefficient ​ ​ government (Méon and Sekkat 2005). Mo (2001) estimated that 1% increase in corruption led to ​ ​ ​ ​ a decrease by 0.72% in private investment.

So, what is corruption? Most often, corruption is associated with bribery in the public eyes. However, it embraces a large palette of forms (bribery, embezzlement, theft and fraud, graft, revolving door, clientelism, nepotism, cronyism, conflict of interests, patronage, etc.) targeting different levels of authority and adapting the appropriate methods of sophistication to context and culture. It is no wonder that there is no consensus among scholars on the definition of corruption.

The general accepted definition, promoted by international organizations such as Transparency International, refer to corruption as “the misuse of entrusted power for private gain” (Rose-Ackerman 2008, p. 551; Andersson and Heywood 2009, p. 746; Xu 2016, p. 86). As ​ ​ ​ ​ ​ ​ already anticipated, this definition is generally associated with bribery in a bureaucratic context (Fazekas, Cingolani and Tóth 2016; Fazekas 2017; Mungiu-Pippidi 2013 acc. Mungiu-Pippidi ​ ​ ​ ​ 2015). Scholars classify this type as petty, bureaucratic or low-level corruption. ​

Another category of corruption - object of this study - takes place at the higher-levels in public positions, benefits of vast discretion and lacks explicit detailed rules (Fazekas, Cingolani and Tóth 2016). This type of corruption, defined as a deviation from the universalism norm (Kurer ​ ​ 2005: Mungiu-Pippidi 2015) is known in the literature as high-level corruption, government ​ ​ ​ favoritism, institutionalised grand or political corruption, (Jain 2001; Graycar 2015; Fazekas, ​ ​ ​ ​

KTH ROYAL INSTITUTE OF TECHNOLOGY 20 Daniel Pustan ​ ​

Cingolani and Tóth 2016; Fazekas, Tóth and King 2016; Fazekas 2017). Thus, government ​ ​ ​ ​ ​ ​ favoritism in public procurement “refers to the allocation and performance of public contracts by bending universalistic rules of open and fair access to government contracts in order to benefit a closed network while denying access to all others.” (Fazekas, Cingolani and Tóth 2016, ​ ​ p. 3; Fazekas, Tóth and King 2016, p. 370). ​ ​

Ultimately, high-level corruption may deviate into state capture if it reaches the stage where institutions or laws are expressly created to benefit a private group. At this point corrupt decisions are implemented in a corrupt-free manner (Søreide 2002; Fazekas 2017). This stage is ​ ​ ​ ​ more characteristic to economies in transition from communism to capitalism and to some extent through lobby groups in developed economies (Graycar 2015). ​ ​

Among the different types of political corruption, clientelism distinguishes as a special form of particularism because of its contingent or “quid pro quo” nature. Clientelism is the distribution of all kinds of public goods in exchange for political support (Hicken 2011). The major risk areas ​ ​ of capture by political elites are the appointments in the public sector, privatization projects, subsidy programmes and public procurement contracts.

Hicken (2011, p. 291) highlights that the difference between clientelist practices and other types ​ ​ of exchange consists in the fact that the former “always comes with strings attached”. These benefits aim to consolidate or increase political power and are typically in the form of voting. However, it may come in the most diverse forms such as: donations, favorable media coverage, revolving door and public procurement contracts. It is in such a scenario where principal-agent models prove their limitations.

Until recently clientelism was understood as a type of principal-agent relationship. However, the reciprocal character of clientelism does not allow for a clear distinction between the role of the principal and that of the agent. The relationship may reverse to the point where politicians end up holding voters accountable for their votes (Stokes 2005). ​ ​

Instead, Gherghina and Volintiru (2017) propose a bidimensional model of clientelism. Thus, ​ ​ political elites use state resources, in particular procurement contracts, to favor or promise favors to private companies in exchange for support, typically donations. These funds are then used as a competitive advantage against other parties to attract votes (Gherghina and Volintiru 2017). ​

KTH ROYAL INSTITUTE OF TECHNOLOGY Government favoritism in public procurement: evidence from Romania ​ ​21 ​

Complementing Gherghina and Volintiru (2017) model, Dávid-Barrett and Fazekas (2019) ​ ​ ​ ​ highlight three stages in which politicians can bend the universalistic norms: formation of the law, implementation by bureaucracy and monitoring of implementation or evaluation. This paper focuses on the implementation stage, one that requires a new violation of law with each transaction (Dávid-Barrett and Fazekas 2019). ​ ​

One direction of literature sheds light on the important role of political connections on companies’ performance. Political connections are identified as one or more of the three indicators: (1) political campaign donations, (2) board seat connections or family ties and (3) stock holdings by politicians.

With few exceptions, companies benefit greatly of political connections2 in three main areas: stock value, financing and procurement contracts. Fisman et al. (2012) finds no influence of the ​ ​ stock value related to Chimney’s personal ties. Furthermore, Jackowicz, Kozłowski and Mielcarz (2014) study reveals a negative correlation between political connections and firm profitability ​ ​ which, they suggest, is likely due to the instability of Poland’s politics.

More connected with the subject of this paper, the studies related to public procurement reveal that political connections have a major impact on contract allocation even in countries with a solid institutional setting such as Denmark (Amore and Bennedsen 2013) or the United States ​ ​ (Goldman, Rocholl and So 2013, Luechinger and Moser 2014). ​ ​ ​ ​

The increase in the value of procurement contracts varies. In the United States, campaign contribution results in additional procurement contracts of about 2.5 times the value of donations (Bromberg 2014). The same figure in Brazil is at least 14 times (Boas, Hidalgo and ​ ​ Richardson 2014) while in the Czech Republic it can reach 100 times (Titl and Geys 2019). ​ ​ ​ ​

Gürakar and Bircan (2019) shows that closed bids are more expensive than open bids especially ​ ​ if the winner is a company political connected to the party in power.

Scholars conclude that politically connected firms are on average, less productive and more likely to win procurement contracts (Zhuravskayay 2012; Brugués F., Brugués J. and Giambra ​ ​ 2018). In particular, donating firms receive more generous contracts or a higher amount of ​

2 For a detailed list of the studies, please refer to Annex C

KTH ROYAL INSTITUTE OF TECHNOLOGY 22 Daniel Pustan ​ ​ smaller contracts, flexible completion dates and less bidding competition (Brogaard, Denes and Duchin 2015; Titl and Geys 2019; Canayaz, Martinez and Ozsoylev 2015) suggesting a potential ​ ​ ​ ​ ​ ​ clientelistic link between politicians and companies.

In a more targeted study, Hyytinen, Lundberg and Toivanen (2018) identifies favoritism ​ ​ towards the in-house units of municipalities in public procurement for cleaning services in Sweden. Nonetheless, they associate favoritism with municipalities’ efforts to support local employment rather than “outright corruption” (p. 422). This conclusion indicates that favoritism is viewed from the petty corruption perspective rather than as a form of particularism.

The only available studies on favoritism with objective indicators on Romania is the paper of Doroftei and Dimulescu (2015) confirmed by the work of Doroftei (2016). The analysis drew on ​ ​ ​ ​ public procurement contracts over €1,000,000 on construction sector in Romania between 2007 and 2013. Both studies conclude that single bidding and political connections explain the number of contracts awarded in proportion of 44% in pre-election years.

It is worth noting that in spite of the clarity of the single bidding indicator, its simplicity makes it vulnerable to circumvention (Fazekas 2017). One basic technique to adopt is for the connected ​ ​ firm to agree in advance on the content of the bid with one or two other firms at their participation in the procurement process (Fazekas, Tóth and King 2016). A corruption case on ​ ​ the airport runway extension in Cluj-Napoca illustrates such a situation. In this case, the politically connected business man admitted it was a bid rigging. In fact, to create the impression of legality, he went so far as to ask one of the participants to dispute the bid. Only because he knew that the company would not stand any chance as it did not qualify (Bacalu 2018). ​

Dávid-Barrett and Fazekas (2019) propose a more complex measurement for favoritism. For a ​ ​ company to be considered favored, it has to have both suspicious winning patterns and higher risk of corruption as measured by CRI. The CRI is a composite indicator that besides the single bid includes another 5 input indicators. Their analysis revealed that favored companies control 50%-60% of public procurement market in Hungary, a large number compared to 10% in the United Kingdom.

KTH ROYAL INSTITUTE OF TECHNOLOGY Government favoritism in public procurement: evidence from Romania ​ ​23 ​

2.2 Measures of corruption

Measuring corruption is a challenging task. Until recently, the corruption measures consisted mainly in surveys/indicators of attitudes, perception and experience of corruption or a mix of these, namely composite indices. The respondents are mostly experts or firms and to a lesser extent citizens (Fazekas, Tóth and King 2016). Reviews of institutions, scientific analyses, audits ​ ​ and investigations of particular cases complement these measures.

Examples of surveys based on expert opinions are the International Country Risk Guide (ICRG) indices by Political Risk Services and Business International Corporation (BI). The first was used by Tanzi and Davoodi (1998) while the former was used for the first time by Mauro (1995). ​ ​ ​ ​ The Global Competitiveness Index assembled by the World Economic Forum (WEF) and the Institute for Management Development (IMD) in the World Competitiveness Yearbook (WCY) are constructed based on business executives opinion. One cross-country experience based survey among citizens is the Global Corruption Barometer constructed by Transparency International. Transparency International developed the most famous composite index, Corruption Perception Index followed by the Corruption Index from the World Bank.

In favor of these indicators stands their extensive use and the strong correlation with each other as well as with other economic and development indicators (Blackburn, Bose and Emranul Haque 2010). Nonetheless, in spite of their broad use, critics abound with respect to surveys and ​ ​ perception indicators.

Andersson and Heywood (2009) emphasize the remarkable similarity in survey questions, ​ ​ respondents’ group sample and type of corruption targeted (most often bribery) These characteristics may explain in good measure the high degree of correlation between these indicators.

Further, perception does not imply having directly experienced corruption in any way. It may be driven by different factors such as media attention (Golden and Picci 2005). It may also be ​ ​ inconsistent in time due to a later discovery and broadcasting of past corruption acts (Kenny 2009) or unconsciously using previous reports on corruption as benchmark (Golden and Picci ​ 2005). ​

KTH ROYAL INSTITUTE OF TECHNOLOGY 24 Daniel Pustan ​ ​

Other researchers add the fact that answers are shaped by factors such as diverse as individual characteristics, ethnic diversity, engagement in social activities (Olken 2009), the phrasing of ​ ​ the questions (Xu 2016), cultural differences between the interviewee and interviewer (Bayley ​ ​ 1966), or expats impressions on their host country (de Maria 2008). ​ ​ ​

Most interviews are focused on executives or opinion leaders which may indicate a representativeness bias, reflecting the views of a specific group not the one of the general population (Fazekas, Tóth and King 2016). In addition, they may reflect only certain types of ​ ​ corruption while omitting others (Svensson 2005). ​ ​

Most indices are weakly correlated with reality (Weber Abramo 2008). In a study comparing ​ ​ villager's perception on corruption with a more objective measure in Indonesia, Olken (2009) ​ ​ finds that villagers were only able to predict a small part (only 8%) of the real level of corruption in the project. This was due to the fact that they were better at detecting corruption in marked-up prices then inflated quantities which generated most corruption in the project.

The main drawback with composite indices is the conceptual imprecision. They are defined by various indices that compose them which in turn can vary in time (Knack 2007). Their ​ ​ reductionist approach makes them insensitive to variations in sectors or regions and oblivious to underlying phenomena indicated by outliers (Lancaster and Motinola 2001). ​ ​

The Corruption Perception Index (CPI) receives particular criticism as having a Western business-orientation (de Maria 2008, Andersson and Heywood 2009) and that it is more ​ ​ ​ ​ reliable to reflect corruption levels in developed rather than developing countries (Golden and Picci 2005). In spite of these disadvantages, the stakes are high for developing countries by ​ ​ conditioning financial support on the country’s ranking taken as a proxy for the country's reform progress (Oman and Arndt 2006). Misinterpretation of the CPI can lead these countries into a ​ ​ “corruption trap” (Andersson and Heywood 2009, p. 747). ​ ​

The risk of misunderstanding the true measurement increases exponentially with indices where corruption is only a part of a larger group of measurement. This is the case of the broadly use International Country Risk Guide (ICRG) which measures “the political risk involved in corruption” (Lambsdorff 2006, 82) not the level of corruption. ​ ​

KTH ROYAL INSTITUTE OF TECHNOLOGY Government favoritism in public procurement: evidence from Romania ​ ​25 ​

Unfortunately, survey of experiences do not provide a higher degree of reliability. Part of the firms in the sample group may be on the market as a consequence of corruption and may have no interest to report or even misreport (Svensson 2005; Kurtz and Schrank 2007). Even when ​ ​ ​ ​ they do report, they mostly reflect petty corruption as citizens do not come in contact often with high-level corruption (Kenny 2009; Fazekas, Tóth and King 2016). ​ ​ ​ ​

The reviews of institutions lack measurement precision and are based upon implicit understanding regarding the functioning of the observed institutions (Fazekas, Tóth and King 2016; Fazekas and Kocsis 2017). The precision of the measurement increases with scientific ​ ​ ​ analyses, audits and investigation of individual cases. However, their resource-intensive character makes them available in a limited number and even less useful for comparison because of their specific goals and risk of state capture (Fazekas, Tóth and King 2016; Fazekas ​ ​ and Kocsis 2017). ​ ​

The risk of state capture is circumvented by Escresa and Picci’s (2017) more objective index of ​ ​ corruption cases based on the Foreign Corrupt Practices Act. Nevertheless, Stephenson (2014) ​ ​ warns about more or less implicit assumptions that may compromise the validity of this index, such as: (1) conflict of interest (state ownership in the economy of the targeted country), (2) biased indicator because of the mix of exports and Foreign Direct Investments (FDI), (3) the number of investigations in a country being influenced by the USA politics (e.g. high risk country are investigated more often) and (4) the endogenous character of exports.

Indeed, in the latest years researchers drawn their attention towards the so-called “objective” indicators of corruption capable to measure behavior instead of relying on perceptions or experiences. One direction of research analyzed the differences in prices for standardized products or services (Di Tella and Schargrodsky 2003; Bandiera, Prat and Valletti 2009; ​ ​ ​ ​ Hyytinen, Lundberg and Toivanen 2018). Other researchers contrasted official reports with ​ ​ surveys and independent audits to highlight corruption in the award of benefits, grants and projects in infrastructure (Reinikka and Svensson 2004; Olken 2006; Olken 2007). A different ​ ​ ​ ​ ​ ​ measure was suggested by Golden and Picci (2005) which consisted in the ratio between the ​ ​ physical infrastructure and the corresponding cumulative expenditures in 20 regions of the country. Finally, as we mentioned earlier, using proxies such as donations or connected board of directors, political connections constitute in a good measure the focus of recent research.

KTH ROYAL INSTITUTE OF TECHNOLOGY 26 Daniel Pustan ​ ​

Kaufmann, Kraay and Mastruzzi (2011) warns with regards to the two problems that stem from ​ ​ any kind of data when measuring corruption. First, the “noise” in the measurement, which represents the difficulty or impossibility to distinguish between corruption, incompetence, ignorance or other source of noise. This is exemplified in Golden and Picci (2005, p. 37) ​ ​ statement when describing corruption: “... money is being lost to fraud, embezzlement, waste, and mismanagement; in other words, corruption is greater.” Second, the imperfect relationship between specific measures of corruption and overall corruption. There is a risk of generalizing corruption based on a certain measured type of corruption.

In particular, these indicators reliability is shadowed by their dependency on the context, high costs to replicate and the use of a single proxy for corruption which only in the most optimistic if not fanciful scenario coincides with the main driver of corruption (Fazekas, Tóth and King 2016). ​

To address these shortcomings, Fazekas, Tóth and King (2016) created the Corruption Risk ​ ​ Index (CRI). This composite indicator of corruption in public procurement is constructed as the weighted average of various indicators of corruption. The indicators, also called red flags, are identified through the review of the literature and qualitative interviews. Then, they are validated through regression analysis (only indicators which prove to be powerful and significant become part of the index). Further, they are divided between corruption outcomes such as “single bid” and corruption inputs such as “the procedure type” or “the weight of non price evaluation criteria”.

The main benefits of this index are not only that it derives from objective data consisting of more than a single indicator but it is also comparable across time and countries. Furthermore, it follows directly from the definition of high-level corruption in public procurement, namely, the bending of prior, explicit rules for the benefit of a limited network (Mungiu-Pippidi 2006; ​ ​ Rothstein and Teorell 2008) and it can be validated with other proxies (Fazekas and Kocsis ​ ​ 2017). ​

An important inconvenience is that the indicators are not ranked based on their importance. Hence, the resulting score is not as explicit as it should be. In fact, the weighted average CRI score may only reflect the number of indicators composing the CRI, in the absence of categorical and continuous variables which receive special treatment.

KTH ROYAL INSTITUTE OF TECHNOLOGY Government favoritism in public procurement: evidence from Romania ​ ​27 ​

It might be adventurous to read a higher score as more prone to corruption as it is not evident that a higher number of indicators equals a higher risk of corruption. The presence of a single indicator - single bidding - may be sufficient to achieve the corrupt goal while the presence of other numerous indicators may be insufficient to steer the contracts for the favorite company. To address this issue, the paper introduces an alternative approach to weighted average, specifically the Principal Component Analysis.

KTH ROYAL INSTITUTE OF TECHNOLOGY 28 Daniel Pustan ​ ​

3. Institutional framework

The crucial role that institutions play for the prosperity of a country was demonstrated by Acemoglu and Robinson (2012). In the same note, most suggested measures to combat ​ ​ corruption refers to features of institutional environment. Hence, a larger body of research developed under the umbrella of “good governance”, a notion closely related with that of institutional environment (Andersson and Heywood 2009, p. 750). ​ ​

Transparency, meritocracy or the separation of bureaucrat’s careers from political influence as well as the quality and quantity of regulation are some important areas of investigation (Islam 2006; Lindstedt and Naurin 2010; Bauhr and Grimes 2013; Bauhr et al. 2019). ​ ​ ​ ​ ​ ​ ​

3.1 Legislative setting

Prior to its adhesion to the European Union as a full member in 2007, Romania transposed the public procurement Directives 2004/17/EC and 2004/18/EC of the European Parliament and of the Council into national legislation. The main legislation governing public procurement in Romania up until 2016 was represented by Government Emergency Ordinance (GEO) 34/2006. The Ordinance has suffered numerous and substantial modifications while into force. Only during the analyzed period I identified 17 changes.

T able 1 Amendments to GEO 34/2006 during the studied period ​ 2009 2010 2011 2012 2013 2014

GEO 19 GEO 76 Law 279 Law 76 GEO 4 GEO 51 Law 84 Law 284 Law 114 GEO 44 Law 9 GEO 72 Law 278 GEO 77 GEO 31 GEO 35 Law 166 Law 193

In addition to the primary legislation, secondary and tertiary legislation apply to public procurement. The secondary legislation refer to Government Decisions consisting of the detailed arrangements on the application of the primary legislation. The orders and instructions issued by the National Agency for Public Procurement (NAPP) compose the tertiary legislation. The existence of this extra layer of legislation highlights the problem of red tape in Romania’s public procurement (Doroftei and Dimulescu 2015). ​ ​

KTH ROYAL INSTITUTE OF TECHNOLOGY Government favoritism in public procurement: evidence from Romania ​ ​29 ​

F igure 2 The comprehensiveness of regulation in Romania ​ Source: Europam.eu

On May 23rd, 2016 a new package of public procurement laws consisting of three laws 98/2016, 99/2016 and 100/2016 replaced the Government Emergency Ordinance (GEO) 34/2006 and its subsequent modification acts. These laws are meant to transpose Directives 2014/23/EU, 2014/24/EU, 2014/25/EU.

Mungiu-Pippidi (2015) shows that two thirds of the contracts awarded to favored companies ​ ​ come from non-EU funds. Corruption with EU funds is foreseeable less prevalent considering that most potential irregularities with these funds are covered by the Law 78/2000 on prevention, discovering and sanctioning of corrupt acts. The national funds do not benefit of such a level of protection. In this case, most prosecutions are undertaken based on the law punishing the abuse or malfeasance of office (Euractiv 2014). ​ ​

This law was amended later, after the period of analysis, specifically on June 15, 2016, by the Constitutional Court decision number 405 which specified that the syntagm “performs faulty” should be interpreted as “performs by breaking the law”. By “the law” is understood only laws, simple Government Acts or Emergency Acts. This way secondary regulations, such as Government Decisions or Ministerial orders, are excluded (Toma 2016). ​ ​

KTH ROYAL INSTITUTE OF TECHNOLOGY 30 Daniel Pustan ​ ​

To decrease the period of delays in contract execution, on June 28, 2014 the government issued an Emergency Act to increase the obstacles for submitting complaints. It requires a “legitimate interest” as well as a guarantee of 1% and up to 100,000 EUR to be paid by complainer. The guarantee was forfeited in the event that the complaint were rejected (regardless of the reason: in substance, inadmissible or by way of a plea) (Avocat.net 2014; Euractiv 2014). On March 19, ​ ​ ​ ​ 2015, the Constitutional Court decided that the Emergency Act 51/2014 was unconstitutional.

3.2 Public procurement institutions

Until mid 2015, the institutional framework consisted of three institutions with competences exclusively in relation to public procurement (Aprahamian and Pop 2016). The main responsible ​ ​ body was the National Authority for Regulating and Monitoring Public Procurement (ANRMAP) with a role in policy and regulatory supervision, controls the manner of the contract award (ex-post) and applies legal sanctions where necessary. The second public body was the Unit for Coordination and Verification of Public Procurement (UCVAP) with attribution in verification of the procedural aspects related to the contract award process (ex-ante control). On May 20, 2015 both agencies were merged into one, namely, the National National Public Procurement Agency (ANAP).

The National Council for Solving Complaints (CNSC) is the third public body specific to public procurement system and represents the administrative, non-judicial first instance to solve complaints lodged under public procurement jurisdiction. As a final remedy procedure, CNSC decision can be challenged before the Court of Appeal located within the same range as the contracting authority’s headquarters (Directorate-General for Regional and Urban Policy (European Commission) and PWC 2016). ​ ​

The Court of Accounts represents the main oversight body. It conducts ex-post audits related to all types of funds reporting their findings to ANRMAP and suspicious irregularities under the criminal law to the National Anticorruption Directorate (DNA). Other entities involved in monitoring public procurement are: the Management Authorities and the Department for the Fight Against Fraud (DLAF), the Competition Council, the Authority for Certifications and Payments and the National Integrity Agency (Doroftei and Dimulescu 2015; Directorate-General ​ ​ for Regional and Urban Policy (European Commission) and PWC 2016) ​ ​

It is also worth mentioning the Electronic System for Public Procurement (SEAP) operated by the Romanian Agency for Digital Agenda (formerly known as the National Management Centre

KTH ROYAL INSTITUTE OF TECHNOLOGY Government favoritism in public procurement: evidence from Romania ​ ​31 ​ for the Informational Society). It is an electronic procurement system with the aim to increase efficiency and transparency in the public procurement system. The platform was upgraded and integrated with relevant institutions in public procurement on April 2, 2018 (Romanian Digital Agenda Agency 2018). ​ ​

3.3 Government characteristics

During the seven years period, Romania had eight Cabinets and three Prime Ministers (though one of them for a very short period of time). The first party, the Democrat Liberal Party (PDL), governed between December 22, 2008 and April 26, 2012 when it was dismissed through a no-confidence motion. The second party, the Social Democratic Party (PSD), was in power between April 27, 2012 and November 17, 2015.

Following parliamentary elections, Boc I cabinet, made of PDL, PSD and the Conservative Party (PC), started its mandate on December 22, 2008 and was dismissed by Parliament through a no-confidence motion on October 13, 2009 (Mediafax 2009). This was soon after PSD and PC ​ ​ alliance left the government on October 1, 2009. Prime Minister Emil Boc and ministries belonging to PDL party continued to ensure the interim until December 23, 2009.

On the aforementioned date, Boc II cabinet was invested by the Parliament as Traian Basescu was re-elected president (Politica Românească 2011). Thus PDL was able to form a ​ ​ parliamentary majority with the National Union for the Progress of Romania (UNPR) - a party formed later by independent parliamentarians who either left or were dismissed from their own party after 2008 parliamentary elections (Ciubotaru 2013). ​ ​

For clarity purposes, it should be mentioned that 2004 was the last year when Presidential and Parliamentary elections coincided. This is due to the fact that the Presidential mandate began to last 5 years while parliamentary continued to be of 4 years. Hence Parliamentary elections took place in December 2008 while presidential elections were held in December 2009 (Centrul pentru Integritate în Business 2013). ​ ​

Cabinet Boc 2 ended its mandate on February 6, 2009 after Emil Boc resigned as prime minister following street protests. After a three days period of interim, cabinet Ungureanu was invested. It had an independent prime minister but the ministers were occupied by the National Liberal Party (PNL), UNPR and the Democratic Alliance of Hungarians in Romania (UDMR). This

KTH ROYAL INSTITUTE OF TECHNOLOGY 32 Daniel Pustan ​ ​ cabinet only lasted for three months as it was dismissed by the Parliament on April 27, 2012 through a non-confidence motion.

After the dismissal of cabinet Ungureanu, Victor Ponta became prime-minister being in charge of four consecutive governments between May 7, 2012 until November 5, 2015 when he resigned following street protests claiming that the tragedy from “Colectiv” club could have been avoided with more efficient measures to combat corruption (Tran 2015). ​ ​

The first Ponta cabinet, formed by the Social Liberal Union (USL) formed by PSD and PNL was in charge until parliamentary elections in December 2012. Cabinet Ponta 2 was formed after the elections and co-opted UNPR in the government. PNL leaving the government lead to government restructuring and co-opting the hungarian minority party, UDMR, which resulted into cabinet Ponta 3. PNL decided to break the coalition suspecting that their ally will not keep the agreement and will launch their own candidate for presidency instead of supporting PNL’s candidate. Cabinet Ponta 4 was formed due to UDMR’s departure from the government as a result of a massive vote in the presidential election of their electorate in favor of Klaus Iohannis, contra-candidate of prime minister in charge to presidency.

After a short interim period generated by the prime-minister resignation as consequence of street protests following Colectiv nightclub fire, the independent Ciolos cabinet was invested. The investment by the parliament took place on November 17, 2015. Ciolos government was in power until December 2016 parliamentary elections.

A summary of government’s succession in presented in table 2. ​ ​

KTH ROYAL INSTITUTE OF TECHNOLOGY Government favoritism in public procurement: evidence from Romania ​ ​33 ​

T able 2 Romania’s government cabinets during 2009-2015 ​ Prime Type of Cabinet Period Party Other parties minister ending 2008, 2009, No-confidence Boc I Emil Boc PD-L PSD+PC (PUR) Dec 22 Dec 23 motion Boc II 2009, 2012, Government Emil Boc PD-L UDMR, UNPR Dec 23 Feb 9 resignation Mihai 2012, 2012, PDL, UNPR, No-confidence Ungureanu Razvan Independent Feb 9 May 7 UDMR motion Ungureanu 2012, 2012, Victor Parliamentary Ponta I PSD PNL, PC (PUR) May 7 Dec 21 Ponta elections 2012, 2014, Victor PNL, PC (PUR), Government Ponta II PSD Dec 21 Mar 5 Ponta UNPR restructuration 2014, 2014, Victor PC, UDMR, Government Ponta III PSD Mar 5 Dec 17 Ponta UNPR, PLR restructuration Ponta IV 2014, 2015, Victor UNPR, ALDE Government PSD Mar 5 Nov 17 Ponta (PLR + PC) resignation

Figure 3 presents the two government periods together with the party composition of ​ government cabinets during the observation period.

F igure 3 Timeline of cabinets and parties for the period 2009 - 2015 ​

KTH ROYAL INSTITUTE OF TECHNOLOGY 34 Daniel Pustan ​ ​

The average period time of governing for a cabinet was 12 months during the 7 years observation period. The shortest period in power was that of cabinet Ungureanu (3 months) while the longest running cabinet was cabinet Boc 2 (25 months). For simplicity, the interim periods were considered jointly with the cabinet period that preceded them. Furthermore, this paper proposed to capture the influence of the main parties (PDL and PSD) along a multi-governmental period which included various cabinets.

KTH ROYAL INSTITUTE OF TECHNOLOGY Government favoritism in public procurement: evidence from Romania ​ ​35 ​

4. Method

This paper uses a quantitative approach defined as “entailing the collection of numerical data, a deductive view of the relationship between theory and research ... and an objectivist conception of social reality” (Bryman 2012, p. 149). In this study, quantitative research is preferred over ​ ​ qualitative research as it allows testing hypotheses based on directly observable behavior. The increase in government transparency and availability of big data on public procurement contracts made this type of research possible.

Further, a particular form of longitudinal analysis, namely, panel analysis was performed after aggregating the data on a half yearly basis. The main benefits of using panel data are that it accounts for heterogeneity and allows to study dynamic factors such as the spells of contract value won by companies (Hill, Griffiths and Lim 2011). ​ ​

Following Dávid-Barrett and Fazekas (2019) two hypotheses are verified. First, it is identified ​ ​ for which companies the change in government affects the value of contracts won. Second, corruption indicators or “red flags” are analyzed to determine whether the change in companies’ winning tender is also associated with a higher risk of corruption.

Companies benefit of favoritism where both conditions are met. Additionally, favoritism is considered systematic if the patterns of winning contracts in the presence of more red flags than the rest of the market can be observed at the level of company type.

Fine shifts in government spending or more sophisticated methods of corruption are the most reasonable explanation when only the first hypothesis is true. If only the second hypothesis is true, it is probably because firms buy influence from both parties or other authorities capable to influence the process (Dávid-Barrett and Fazekas 2019). ​ ​

Controlling for changes in spending preferences as a result of differences in policy priorities (CVMti), little variation is expected in contract winning tenders as a result of the change in government.

To identify the companies for which the value of contracts won was affected by the change in government, I use the Arellano-Bond system GMM transformation of the following dynamic panel regression model:

KTH ROYAL INSTITUTE OF TECHNOLOGY 36 Daniel Pustan ​ ​

CVti = α + β1 CVt-1i + β2 CVt-2i + β3 CVMti + β4 MMi + β5 Cabt + Ui + Ɛit ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​

The abbreviations of the model stand for the following concepts:

CVti : the contract value won by company in period t ​ ​ α : ​ the constant term for the sample

CVt-1i : the contract value won by company i in past period (t-1) ​ ​ CVt-2i : the contract value won by company i in past period (t-2) ​ ​ CVMti: the contract value spent in main market (CPV level 3) ​ ​ MMi : the sectoral dummy for the main market (CPV level 3) ​ ​ Cabt : dummy representing the cabinet in period t ​ ​ Ui : the company fixed effect ​ ​ ​ Ɛit : the error term ​ ​ ​

I ran the two versions of the above model - fourth root and natural log contract value - on a subsample of companies winning contracts in at least two different periods (removing 17,819 observations). The reason for using fourth root is the ability to rescale the data while keeping the observations with zero contract value during the observed period. The two models reveal favored companies at different stages of maturity. The fourth root version explore the case of both well established and incumbent companies. The natural log version analyzes only the more experienced companies in the procurement market as it removes the companies with zero contract value observations during the analyzed period (Dávid-Barrett and Fazekas 2019). ​ ​

During the first nine months, PDL and PSD governed together in cabinet Boc I. However, as important as it may be, given that the period of governing together took place during the first two half-year periods and the model uses two lags, these observations are not used. They are automatically dropped because of perfect collinearity as consequence of the fact that all the observations are zero after the first two periods.

Based on the regressions company-specific error term, companies having ties to government 1 are labeled “surprise losers” and those with government 2 are called “surprise winners”.

KTH ROYAL INSTITUTE OF TECHNOLOGY Government favoritism in public procurement: evidence from Romania ​ ​37 ​

Specifically, companies that have an above period average regression error under one government and below period average regression error under the other government are labeled as “surprise losers” and “surprise winners”. Companies with stable residual patterns (above or below period average in both governments) are considered “stable companies”.

The same terminology as Dávid-Barrett and Fazekas (2019) is used in this paper: “surprise ​ ​ loser” and “surprise winner”. These terms are defined in relationship to the second government period and because their use might give rise to misinterpretation, clarification may be necessary.

Thus, the reader is cautioned not to extend the negative and positive connotations of these terms to the entire period of study. “Surprise losers” while possibly discriminated during the second government period, might have benefited from preferential treatment during the first government period. Therefore, it should be borne in mind that “surprise losers” are the companies suspicious of having ties with government 1 while the “surprise winners” are the firms suspicious of links with government 2.

The regressions set-up controlled for the following: (i) government spending patterns resulting from different policy preferences or priorities measured by contract value spent in main market, (ii) market and technological specifics measured by CPV division (2-digit CPV) and (iii) regulatory as well as government specificities (e.g. parties composing the government) measured by cabinet.

Once identified the companies with a suspicious winning pattern, I analyze whether they won under conditions with a higher potential for corruption. In order to do so, the surprise winners and losers are cross checked with the Corruption Risk Index (CRI).

The Corruption Risk Index is a relatively recent tool developed by Fazekas, Tóth and King (2016) aiming to measure the degree of corruption in a more objective manner as a reliable and ​ ​ cost efficient alternative to current measurements. The index is based on previous research (Charron et al. 2017, Fazekas and Kocsis 2017, Dávid-Barrett and Fazekas 2019) referring to not ​ ​ ​ ​ ​ ​ less than 30 quantitative indicators and 20 corruption techniques in public procurement were identified. A summary table with all these indicators is presented in annex B (Fazekas, Tóth and King 2013). ​ ​

KTH ROYAL INSTITUTE OF TECHNOLOGY 38 Daniel Pustan ​ ​

The model includes one outcome indicator - single bid - and various input indicators. The elementary corruption risk indicators (red flags) that constitute the CRI are identified based on the literature review: 1. Single bid (outcome indicator) 2. Call for tender publication (submission phase) 3. Procedure type (submission phase) 4. Weight of non-price evaluation criteria (assessment phase) 5. Procedure period (assessment phase) 6. Notification of award period (assessment phase) 7. Contract type (assessment phase)

Because “submission deadline” was not present in the dataset, it is not possible to build the continuous variable indicators “length of advertisement period” and “length of decision period”. However, two similar indicators are used, namely “procedure period” and “notification of award period”. The “procedure period” indicator represents the number of days between the call for tender and contract date. The “notification of award period” computes the number of days between the contract date and notification of award date.

Binary logistic regression is used to determine the suitability of each composite index in the index excluding those indicators that do not prove to be significant and strong predictors of single bidding (Fazekas and Kocsis 2017). The logistic model is: ​ ​

1 Pr (single bidder = 1) = ​ 1+e−zi

Zi = α + β1 Sij + β2 Aik + β3l Dil + β4m Cim + Ɛi ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​

Where:

Zi - the logit of a contract being a single bidder contract ​ ​ α - the constant of the regression ​ S - the matrix of j corruption inputs of the submission phase for the ith contract such as length ​ij ​ of submission period

KTH ROYAL INSTITUTE OF TECHNOLOGY Government favoritism in public procurement: evidence from Romania ​ ​39 ​

A - the matrix of k corruption inputs of the assessment phase for the ith contract such weight ​ik ​ of non-price evaluation criteria

D - the matrix of l corruption inputs of the delivery phase for the ith contract such contract ​il ​ lengthening

C - the matrix of m control variables for the ith contract such as the number of competitors ​im ​ on the market

Ɛi - the error term ​ ​

β1, β2, β3l, β4m - the vectors of coefficients for explanatory and control variables ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​

Afterwards, component weights for each variable are derived. The outcome indicator - with no regression coefficient - receives the weight of 1. Also, each significant corruption input receives the weight of 1 while those variables or variable categories that do not prove significant are discarded (weight zero).

The categories within the categorical variables received component weights ranked according to their regression coefficient. For instance, in the case of two significant categories, the one with the highest coefficient would receive the weight of 1 while the other one receives the weight 0.5.

For continuous variables, threshold were identified using linear regression analysis and the residual distribution graphs to determine discrete jumps in residual values. The control variables represent the most important alternative explanation for the outcome.

KTH ROYAL INSTITUTE OF TECHNOLOGY 40 Daniel Pustan ​ ​

F igure 4 Mean regression residuals by two-percentiles of relative procedure period (contract date - call for tender date), linear prediction

F igure 5 Mean regression residuals by two-percentiles of relative notification of award period ​ (award notice date - contract date), linear prediction

KTH ROYAL INSTITUTE OF TECHNOLOGY Government favoritism in public procurement: evidence from Romania ​ ​41 ​

Both figures display three different period-groups. One which overestimates the probability of single received bid, one which underestimates this probability and, lastly, one that approximate it.

Moreover, each corruption input was introduced step by step starting with those from the earliest phases. The inputs that prove to be insignificant were not introduced in the CRI composite indicator.

The control variables in the regression are: (i) institutional endowments measured by type of issuer (national, county or municipal), (ii) market and technological specifics measured by CPV division (2-digit CPV), (iii) contract size measured by log of contract value in euro, (iv) legislation measured by year and (v) contracting authorities specificities measured by activity type.

Finally, CRI is calculated in two ways. First, as the arithmetic average of the individual risk indicators - red flags (Fazekas, Tóth and King 2016): ​ ​ ​ ​ ​

i CRI = Σ j wj * CIj ​ ​ ​ ​ ​ ​ ​

Secondly, CRI is computed using an alternative technique, namely the Principal Component Analysis (PCA).

The validation of the Corruption Risk Index was done both at macro and micro level. At the macro-level I compared the Corruption Risk Index with three notorious perception surveys: The Corruption Perception Index (CPI) from Transparency International, World Bank’s World Governance Indicators (WGI) and the Global Competitiveness Index (GCI) developed by The World Economic Forum.

This last indicator, GCI, covers a much broader area besides corruption (e.g. efficiency of government) for which reason only the most relevant indicator for this purpose was selected, specifically, favoritism in decisions of government officials (1.07).

The validation of the Corruption Risk Index at micro-level is done by looking at the correlation of the relative contract value and CRI. The relative contract value is calculated as the ratio between the actual contract value and the estimated contract value (Fazekas, Tóth and King 2016). A higher ratio of relative contract value represent a higher risk of corruption. ​

KTH ROYAL INSTITUTE OF TECHNOLOGY 42 Daniel Pustan ​ ​

A second micro-level indicator, namely, winner’s country of origin (Fazekas and Kocsis 2017) ​ ​ was not applicable to this study. The purpose of this indicator is to analyze the correlation of companies’ main office located in a tax haven country related to the CRI. However, no winner of public procurement contracts during the 2009-2015 reported their main office in a country included on the EU non cooperative tax jurisdictions and high risk third country list (European Parliamentary Research Service 2018). ​ ​

KTH ROYAL INSTITUTE OF TECHNOLOGY Government favoritism in public procurement: evidence from Romania ​ ​43 ​

5. Data

The public procurement data contains all public procurement contracts in Romania between 2009 and 2015. The original datasets were published on data.gov.ro in the form of 11 .csv files by the Romanian Agency for Digital Agenda (Mihalache 2013). Alternative data was available from ​ ​ The European procurement journal, Tenders Electronic Daily, which comprised of the contracts above a certain threshold.

Nevertheless, the data published by the Romanian national agency was prefered because it allows a better understanding of procurement market in its entirety. Moreover, the complete dataset is suitable to offset the corruption technique of fragmenting big contracts into smaller contracts. This technique is used to conceal irregularities or simply to benefit of more permissive rules.

Unfortunately, the poor quality of the data required a substantial work in order to reach an adequate quality for scientific research. One obstacle consisted in the fact that the elementary variable, the Company Identification Number (CIN), included multiple inaccurate observations. Keeping these observations in the analysis required to manually correct those instances. In spite of the time consuming process, this effort was carried out with the hope that other researchers may also benefit from it.

Over 5,000 companies were searched by name on the Ministry of Finance and other specialized websites (e.g. listafirme.ro and totalfirme.ro). The entities that were still not found, most of ​ ​ ​ them consisting of self-employed persons, were assigned a unique identification comprising their names without spaces. In addition, all foreign companies’ CIN was verified. Kompass.com ​ website together with the official companies registry websites of their countries of origin were employed for this purpose.

Once the correct CIN was assigned for the winning companies, the names and addresses were extracted from the dataset published by the Ministry of Finance on the website data.gov.ro. This dataset consists of all economic operators and public entities. For the operators without a CIN as well as for foreign companies, the name and address were extracted from a dataset created for this purpose.

KTH ROYAL INSTITUTE OF TECHNOLOGY 44 Daniel Pustan ​ ​

Inconsistencies between the CIN and companies’ name in the dataset were tackled by triangulating with company’s address variable. Whichever between the two - CIN or company name - matched the address was considered correct while the other was changed accordingly.

Companies that changed their name for branding or legal purposes were assigned the new name during the entire observation period. Examples of such companies are Engie Romania SRL, previously GDF Suez Energy and Veolia Energie Prahova SRL which changed from Dalkia Thermo Prahova SRL due to the acquisition by the former.

The same process was conducted for the contracting authorities related variables. First, the validity of their CIN was verified. In the event of a mismatch between the entities’ CIN and the name variable, these were triangulated with the address variable. Subsequently, the name and address were imported from the Ministry of Finance dataset (Ministry of Finance 2018). ​ ​

To ensure the quality and integrity of the database, other correction measures were taken for the existing errors and inconsistencies in all variables. Some examples of the issues encountered are: broken lines, leading and trailing spaces, consecutive spaces, use of diacritics and non alphanumeric characters, etc.

The association of companies was treated similarly to Doroftei (2016) and Doroftei and ​ ​ Dimulescu (2015). Specifically, the contracts awarded to a consortium were assigned to the ​ ​ leader. This is consistent with the thesis that most frequently it is the leading company with political connection that enables the consortium to win the contract. Nevertheless, this approach may generate inflated contract values for some companies.

The regressions were fitted on a subsample of competitive markets. Competitive markets are defined by product group (CPV group - level 3) and regions (NUTS1) with at least three participants (Fazekas, Tóth and King 2016). Non-competitive markets under normal ​ ​ circumstances, such as those for specialized services or defense, were excluded to reduce false positive due to the underlying assumption that the lack of competition signals corruption. (Dávid-Barrett and Fazekas 2019, Fazekas and Kocsis 2017). ​ ​ ​ ​

Albeit the purpose of this paper is to measure governmental favoritism, the sample was not limited only to government related contracts for two reasons. First, in a centralized country like Romania the government has significant influence on the local governing authorities through

KTH ROYAL INSTITUTE OF TECHNOLOGY Government favoritism in public procurement: evidence from Romania ​ ​45 ​ the allocation of funds. Secondly, most of the vote buying performed by local party organizations is carried on behalf of the national party organization.

In any case, the distribution of procurement by the type of contracting authority is as follows: the government and its related agencies is the main contractor with 45% of the total value of acquisitions. It is followed by the local administration authorities with 35% and the county level administration with 11%. The remaining 9% consists of non-governmental organizations (NGO’s) and companies.

Company level analysis was performed on a half-yearly aggregated database comprising the companies that won contracts in at least two periods between 2009 and 2015 period. Half-year period “is optimal for retaining a high level of granularity while also taking into account the erratic character of many public procurement markets” (Dávid-Barrett and Fazekas 2019, p.13). ​ ​

The government periods were defined in relation with the government party, the party providing the political support to the prime minister and holding the majority of the ministerial positions in the cabinet. Apart from this, one year period of transition covering Ungureanu and Ponta I cabinets was allowed as a means to capture the differences between two established parties in power (Dávid-Barrett and Fazekas 2019). ​ ​

T able 3 Main characteristics of the dataset ​ 2009 2010 2011 2012 2013 2014 2015 Total Total number of 115,868 121,170 103,342 106,507 65,413 72,560 67,662 652,522 contracts awarded Total number of 9,890 10,208 9,144 7,933 7,041 6,509 6,255 16,010 winners Total number of 3,983 3,647 3,331 3,189 2,814 2,813 2,775 6,826 contract. authorities Total value of awarded contracts 11,100 10,400 15,200 13,400 13,900 16,300 16,400 96,700 (mil. EUR)

KTH ROYAL INSTITUTE OF TECHNOLOGY 46 Daniel Pustan ​ ​

6. Results

The analysis reveals moderate evidence of partisan favoritism in Romanian public procurement market. Specifically, 24% of the public procurement market in Romania appear to be dominated by favored companies of which 15% surprise losers and 9% surprise winners.3 This is above the 10% market capture in the United Kingdom’s but lower than 50%-60% results in Hungary (Dávid-Barrett and Fazekas 2019). ​ ​

Outcomes similar to Hungary were expected given the comparability between the two countries in many aspects including the geographical proximity, the communist background and the equivalent rankings in the corruption perception index. However, there is a resemblance, specifically in that, like Hungary,the Romanian public procurement market is characterized by systematic favoritism. That is to say that favoritism patterns are identifiable at the level of company type.

Using Arellano-Bond system GMM transformation of the dynamic panel regression, I find moderate evidence of company performance through the analyzed period (Table 4).4 The first ​ model has a moderate explanatory power while the second model has a relatively low explanatory power: 0.69 and 0.38.5

In Dávid-Barrett and Fazekas (2019) using a smaller sample on Hungary (log model), the ​ ​ predicting power drops dramatically by almost 90%. This can be a sign of a non-representative sample for the population or there can be influential observations which are lost in the natural log model. The fact that in my case the sample decreases much less is one of the reasons for the reduced decrease in predicting power for the log with respect to the fourth root model.

3 Results for both “fourth root” and “log” contract value regressions were taken into account 4 Sargan test of overidentification of instruments show that the null-hypothesis could not be rejected 5The linear correlation coefficient between predicted and observed outcome was used ​

KTH ROYAL INSTITUTE OF TECHNOLOGY Government favoritism in public procurement: evidence from Romania ​ ​47 ​

T able 4 System GMM linear dynamic panel regression explaining company market success ​ fourthrootvaloareeur lnvaloareeur fourthrootvaloareeur = L1 0.0274*** (0.00453)

fourthrootvaloareeur = L2 0.0220***

(0.00431) lnvaloareeur = L1 0.0732*** (0.0171)

lnvaloareeur = L2 0.120***

(0.0148) Control variables

fourthrootCVmainmarket Y N

lnCVmainmarket N Y

main market Y Y

cabinet Y Y

Constant 14.32*** 6.814***

(1.421) (2.054)

Observations 192,120 28,331

Number of panelID 16,010 6,592

Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

I identified 3,860 suspicious companies (2,359 surprise losers and 1,501 surprise winners) out of the total of 16,010 companies. These company groups have a CRI trajectory consistent with favoritism. Surprise losers win in the presence of more red flags than the rest of the market during the first government period and surprise winners have a higher CRI than the rest of the market during government 2 period.

There is a significant positive persistence effect. In particular, a one percent increase in contract value increases by 7% after one period and by 12% after two periods. For the fourth root model, significant positive effects with one or two lags are also found in the fourth root model.

KTH ROYAL INSTITUTE OF TECHNOLOGY 48 Daniel Pustan ​ ​

F igure 7 CRI-arithmetic by company group, half-year period ​

Figure 8 displays the overall market share trends for each company group. This shows clear ​ contrasting market outcome trajectories between surprise losers and surprise winners. The difference is lower during the period in which the two parties governed together.

F igure 8 Combined market shares of company types, half-year period ​

KTH ROYAL INSTITUTE OF TECHNOLOGY Government favoritism in public procurement: evidence from Romania ​ ​49 ​

In the following, results for the Corruption Risk Index are presented.

First, not publishing the call for tenders strongly predicts single bidding. The average probability of a single received bid increases by 135%.

Second, contrary to expectations, only the “accelerated negotiation” and “negotiation without call for tender” procedure types display higher risk of corruption signs. The contracts with a single bid in an accelerated negotiation tend to be awarded to winning companies 141% more often as compared to open bid contracts. The “negotiation without call for tender” award 92% more single bid contracts than open procedure does. All the other non-open procedure types predict decreases in single bidding. However, “competitive dialogue”, “limited accelerated tendering” and “limited tendering” are insignificant and according to the theoretical framework will not be included in the composite indicator.

Third, the indicator “weight of non-price criteria” is a powerful predictor but with a weak to moderate impact in explaining single bidding contracts. In the complete model the probability of single bidding decreases by about 13%.

Fourth, an extremely short procedure period (up to 24 days) is a very strong predictor of single bidding with 218% compared to those of over 77 days. The contracts awarded between 24 and 77 days from the call for notice date have a more moderate predicting capacity of single bid with 63%.

Fifth, the notification of award period is associated with a low corruption risk. The contracts where the award notice was published within 8 days after the signing of the contract have a 17% higher risk of corruption while there is a lower risk of corruption roughly by the same percentage for the contracts signed over 27 days as compared to contracts between 8 and 27 days.

Sixth, compared to supplies, the works/construction contract type decreases the risk of corruption by 54% whereas the services increases the risk with 35%.

KTH ROYAL INSTITUTE OF TECHNOLOGY 50 Daniel Pustan ​ ​

T able 5 Logit regression results on contract level, 2009-2015, number of winners ≥ 2 ​ singleb Procedure type Ref. cat: 3, Open tendering 1, Competitive dialogue -0.475 2, Invitation to tender -0.916*** 4, Limited tendering -0.130 5, Limited accelerated tendering -0.0659 6, Negotiation -0.336*** 7, Accelerated negotiation 1.412*** 8, Negotiation without contract notice 0.918*** No call for tenders 1.345*** Weight of nonprice Ref. cat:Lowest price Most economically advantageous tender -0.129*** Procedure period Ref. cat: 3, 77 days - max 1, 0 - 24 days 2.179*** 2, 24 - 77 days 0.631*** 4, missing 0.223*** Notification of award period Ref. cat: 2, 8 - 27 days 1, 0 - 8 days 0.176*** 3, 27 days - max -0.194*** Contract type Ref. cat:Supply 2, Works -0.540*** 3, Services 0.351***

Control variables Type of issuer Y CPV division Y lnContractValue Y Year Y Activity type Y

Constant -1.663*** Observations 362,815 Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

KTH ROYAL INSTITUTE OF TECHNOLOGY Government favoritism in public procurement: evidence from Romania ​ ​51 ​

Table 6 presents the component weights derived based on the regression results.

T able 6 CRI component weights ​

Variable Component weight

Single bid 1

Procedure type Ref. cat: 3, Open tendering 7, Accelerated negotiation 1 8, Negotiation without contract notice 0.75 6, Negotiation 0.5 2, Invitation to tender 0.25

No call for tenders 1

Weight of nonprice Ref. cat:Lowest price Most economically advantageous tender 1

Procedure period Ref. cat: 3, 77 days - max 1, 0 - 24 days 1 2, 24 - 77 days 0.66 4, missing 0.33

Notification of award period Ref. cat: 2, 8 - 27 days 1, 0 - 8 days 1 3, 27 days - max 0.5

Contract type Ref. cat:Supply 3, Services 1 2, Works 0.5

KTH ROYAL INSTITUTE OF TECHNOLOGY 52 Daniel Pustan ​ ​

The molding of the CRI on the definition of grand corruption and the fitted regression models are the main supporters for the validity of the Corruption Risk Index. The macro-level perception surveys only reinforces this validity.

T able 7 Bivariate Pearson correlation of CRI and single bid with perception survey indicators ​ Indicator CRI-arithmetic CRI-PCA single bid years CPI -0.7388 -0.8357 -0.4885 7 WGI (1.07) -0.6862 -0.8321 0.0045 7 GCI -0.4205 -0.3414 -0.3741 7

F igure 9 CRI arithmetic compared to perception-based surveys (CPI, WGI, GCI) ​ ​ ​

F igure 10 CRI-PCA compared to perception-based surveys (CPI, WGI, GCI) ​

KTH ROYAL INSTITUTE OF TECHNOLOGY Government favoritism in public procurement: evidence from Romania ​ ​53 ​

F igure 11 Single bid compared to perception-based surveys (CPI, WGI, GCI) ​

The main reason for the moderate level of correlation stems from the fact that perception-based indicators are more adequate to compare the levels of corruption across countries for longer periods of time (Escresa and Pici 2016 acc. Fazekas and Kocsis 2017) as they capture mostly ​ ​ pitty corruption and they are not sensitive to change.

A more reliable indicator is the relative contract value. However, the estimated contract value variable present various inexactitudes that may bias the analysis. The following are some of the observed inaccuracies: (1) use of commas to separate decimals incorrectly, (2) Unrealistically low estimated contract values (below 20€), (3) the estimated contract value appears for the entire project but the real contract value is divided in various batches and (4) Total contract value is repeated because some companies won contracts together but each company is presented with the total amount as if each of them has won the total amount. The outliers with a relative contract value below five are presented in figure 12. ​ ​

F igure 12 Outliers for relative contract value < 2 ​

KTH ROYAL INSTITUTE OF TECHNOLOGY 54 Daniel Pustan ​ ​

For lack of a better solution, considering the time constraints of this study, the regression was run on a subset of companies with relative contract value lower or equal to 1 (one) in order to minimize the number of erroneous observations captured in the analysis.

T able 8 Linear regression explaining relative contract value and CRI ​ (1) (2) (3) (4) (5) (6) relativeCV relativeCV relativeCV relativeCV relativeCV relativeCV

cri_arithmetic 0.284*** 0.418*** -0.00323 -0.00408 cri_pca 0.373*** 0.677*** -0.00477 -0.00669 singleb 0.147*** 0.124*** -0.000977 -0.00114 Control variables TipActivitateAC N Y N Y N Y AuthorityType N Y N Y N Y Year N Y N Y N Y CPV division N Y N Y N Y lnCV N Y N Y N Y

Constant 0.643*** 0.699*** 0.713*** 0.377*** 0.462*** 0.528*** -0.00121 -0.000791 -0.00044 -0.00841 -0.00868 -0.00823

Observations 289,637 224,711 289,637 212,008 169,687 212,008 R-squared 0.026 0.027 0.072 0.154 0.159 0.16 Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

The results confirm the expectations. One additional red flag (1/7 points higher) increases the price by 4% - 6%. Similar values are obtained for CRI-PCA where the price increases by 5.3 - 9.7%. Single bidding contracts have on average 12.4 - 14.7% higher prices than contracts with at least two bidders.

KTH ROYAL INSTITUTE OF TECHNOLOGY Government favoritism in public procurement: evidence from Romania ​ ​55 ​

7. Robustness analysis In this section robustness checks are presented. The CRI-PCA approach and the alternatives in time periods and the change in control variables confirm the robustness of the model.

First, the CRI weighted using Principal Component Analysis (PCA) display similar results to the CRI calculated as an arithmetic average.

F igure 13 - CRI-PCA patterns by company group, half-year period ​

Secondly, changing the aggregated period for the company analysis from half-year to one year ​ ​ yields similar results. According to this specification, the number of favored companies remains almost identical with that for the half year period. Specifically, 13.22% surprise losers and 11.15% surprise winners are identified. In addition, the CRI trajectories are consistent with systematic favoritism.

KTH ROYAL INSTITUTE OF TECHNOLOGY 56 Daniel Pustan ​ ​

F igure 14 CRI-arithmetic patterns by company group, one year period ​

F igure 15 CRI-PCA patterns by company group, one year ​

KTH ROYAL INSTITUTE OF TECHNOLOGY Government favoritism in public procurement: evidence from Romania ​ ​57 ​

Thirdly, variables representing all the changes to the procurement law during the observation period were introduced replacing the variable cabinet as a control for changes in legislation. Again, the results confirmed the robustness of the base model. A very similar number of favored companies was identified (24.15%) and systematic favoritism was evident in these cases, as well.

F igure 16 CRI-arithmetic patterns by company group, changes in legislation ​

F igure 17 CRI-PCA patterns by company group, changes in legislation ​

KTH ROYAL INSTITUTE OF TECHNOLOGY 58 Daniel Pustan ​ ​

8. Conclusions and suggestions for further research

This paper has examined the political influence at the highest level on public procurement market in Romania during 2009-2015. Drawing on the most recent research and novel measurement tools of corruption the study reveals that 24% of the Romanian procurement market is determined by systematic partisan favoritism.

This number is below the 50% market associated with partisan favoritism in Hungary (Dávid-Barrett and Fazekas 2019). Considering the geographical proximity as well as cultural ​ ​ and institutional similarities, a similar score to Hungary was expected.

There are a few possible explanations why that was not the case. First, for comparison purposes, this paper followed Dávid-Barrett and Fazekas (2019). However, the methodology differed in ​ ​ two aspects. Specifically, Dávid-Barrett and Fazekas (2019) sample was restricted to contracts ​ ​ above the EU-threshold and only purchases made by the government and its agencies.

On the other hand, this study included all contracts during the observation period at all levels of public administration. Not restraining the study only to government level contracts is justified by the high level of centralization both administratively and politically in Romania. A very efficient control over the local administration is carried out through budget allocation.

Secondly, the more vibrant and engaged civil society in Romania may have had a significant impact in mitigating corruption. The analyzed period was especially troubled in which two governments had fallen as a consequence of street protests.

A third explanation is related to the high frequency of party switching by Romanian politicians, famous for their lack of ideological attachment. The migration from one party to another also moves the clientele between parties as the favors will be offered through the party in which the politician is currently found. This situation hinders the possibility to identify groups of clientele specific for parties. A good illustration illustration of mass migration from one party to another occurred in August 2014 with the GEO 457/2014. Then, thousands of mayors were able to migrate to PSD party without ceasing to hold office as the law stipulated (Bărbulescu 2014). ​ ​

KTH ROYAL INSTITUTE OF TECHNOLOGY Government favoritism in public procurement: evidence from Romania ​ ​59 ​

The result is more in line to Doroftei (2016) and Doroftei and Dimulescu (2015) findings. Their ​ ​ ​ ​ study discovered 32% favoritism in construction procurement sector in Romania. It should be noted though that construction sector is amongst the most susceptible of corruption.

In conclusion, the results of this paper indicate some progress made by Romania in the fight against high-level corruption in public procurement. However, it invites to moderate optimism since a decrease in one form of corruption does not necessarily imply a drop in the overall corruption. The method applied in this paper cannot rule out alternative practices such as companies buying influence from different parties.

In any case, these results open the way for debates to combat corruption through good government practices that are based on objective proxy measures. In addition, this study provides empirical support for EU anti-corruption policies. It may help focus the attention on specific and quantifiable forms of corruption entailing more precise recommendations by the European Commission’s in the effort to support the fight against corruption in Romania within the Cooperation and Verification Mechanism.

In order to be able to fully exploit the potential benefit for the society of this field, further research is needed. First, as it was shown objective proxy measures are promising tools in understanding the underlying mechanisms of corruption. Yet they are still in an early phase of development. Reliable indicators not only in extension but also in depth, ranking the importance of corruption risk indicators, should be developed.

Another direction of research is represented by the analysis of corruption at local level as well as the interaction between central and local government related to favoritism. Finally, party switching in Romania is a far too common practice. Studying political favoritism at a more granular level, that of the individual politician, may bring important insights on the drives of political favoritism.

KTH ROYAL INSTITUTE OF TECHNOLOGY 60 Daniel Pustan ​ ​

References

A cemoglu, D. and Robinson, J. (2012). Why nations fail: the origins of power, prosperity, and poverty. 1st ​ ​ ed. New York, NY: Crown Business. A idt, T. S. (2010). Corruption and Sustainable Development. https://doi.org/10.17863/CAM.5249 ​ ​ A more, M. and Bennedsen, M. (2013). The value of local political connections in a low-corruption ​ environment. Journal of Financial Economics, 110(2), pp.387-402. ​ A ndersson, S. and Heywood, P. M. (2009), The Politics of Perception: Use and Abuse of Transparency ​ International's Approach to Measuring Corruption. Political Studies, 57: 746-767. ​ A ndrei, A. (2018). Topul firmelor care lucrează cu banii statului. Contracte cât bugetul Iaşului. Ziarul de ​ ​ Iași. [online] Available at: https://www.ziaruldeiasi.ro/stiri/topul-firmelor-care-lucreaza-cu-banii-statului-contracte-cat-bugetu l-iasului--181634.html [Accessed 19 Dec. 2019]. A prahamian, A. and Pop, G. (2016). Building landmarks, smoothing out markets: An Enhanced ​ Competition Framework in Romania. World Bank Studies. Washington, DC: World Bank. License: ​ Creative Commons Attribution CC BY 3.0 IGO A rvate, P., Barbosa, K. and Fuzitani, E. (2018). Party Expertise, Campaign Donation and Government ​ Contracts: Evidence from an Electoral Quasi-Experiment. INSPER Working paper 377/2018. ​ A ttström, K., Kröber, R., Jähne, J., Bergschneider, J., Thomä, J., Swysen, T., Scholzen, D. and Trautenmüller, T. (2012). Final Report: Review of the Functioning of the CPV Codes / System. ​ ​ [online] Berlin: Ramboll. Available at: https://ec.europa.eu/docsroom/documents/21583/attachments/1/translations [Accessed 9 Dec. 2019]. A uriol, E., Straub, S. and Flochel, T. (2016). Public Procurement and Rent-Seeking: The Case of Paraguay. ​ ​ World Development, 77, pp.395-407. A vocatnet.ro. (2014). Garantia de buna conduita obligatorie pentru contestarea licitatiilor publice. ​ ​ [online] Available at: https://www.avocatnet.ro/articol_38023/Garantia-de-buna-conduita-obligatorie-pentru-contestarea -licitatiilor-publice.html [Accessed 20 Dec. 2019]. B acalu, M. (2018). Uioreanu şi Bene merg la închisoare. Sentinţă finală în dosarul de corupţie aeroport ​ [online] Muzicainstantelor.ro. Available at: https://muzicainstantelor.ro/index.php/2018/01/31/uioreanu-si-bene-merg-la-inchisoare-sentinta-f inala-in-dosarul-de-coruptie-de-la-aeroportul-cluj/ [Accessed 23 Nov. 2019]. B andiera, O., Prat, A. and Valletti, T. (2009). Active and Passive Waste in Government Spending: ​ Evidence from a Policy Experiment. American Economic Review, 99(4), pp.1278-1308. ​ B ărbulescu, O. (2014). Ordonanţa de Urgenţă privind migraţia primarilor a fost adoptată de Guvern. ​ ​ Mediafax. [online] Available at httpswww.mediafax.ropoliticordonanta-de-urgenta-privind-migratia-primarilor-a-fost-adoptata-de-g uvern-13160000 [Accessed 1 Jan. 2020].

KTH ROYAL INSTITUTE OF TECHNOLOGY Government favoritism in public procurement: evidence from Romania ​ ​61 ​

B auhr, M. and Grimes, M. (2013). Indignation or Resignation The Implications of Transparency for ​ Societal Accountability. Governance, 27(2), pp.291-320. ​ B auhr, M., Czibik, Á., Fine Licht, J. and Fazekas, M. (2019). Lights on the shadows of public procurement ​ Transparency as an antidote to corruption. Governance, pp.1-29. ​ B ayley, D. (1966). The Effects of Corruption in a Developing Nation. The Western Political Quarterly, ​ ​ 19(4). B BC (2015). Romania PM Ponta resigns over Bucharest nightclub fire. 4 November, [online] Available at: ​ ​ https://www.bbc.com/news/world-europe-34720183 [Accessed 17 Dec. 2019]. B BC (2019). Romania corruption: PSD chief Liviu Dragnea jailed. [online] Available at: ​ ​ https://www.bbc.com/news/world-europe-48421430 [Accessed 26 Dec. 2019]. B efu, D. (2015). Mama lui Ponta: ”Suntem o mică firmă de familie. Nu avem nimic de ascuns”. România ​ Curată vă prezintă principalii clienți ai firmei familiei lui Ponta, la care au ajuns azi procurorii DNA. România Curată. [online] Available at: http://www.romaniacurata.ro/mama-lui-ponta-suntem-o-mica-firma-de-familie-nu-avem-nimic-de- ascuns-romania-curata-va-prezinta-principalii-clienti-ai-firmei-familiei-lui-ponta-la-care-au-ajuns-a zi-procurorii/ [Accessed 19 Dec. 2019]. B lackburn, K., Bose, N. and Emranul Haque, M. (2010). Endogenous corruption in economic ​ development. Journal of Economic Studies, 37(1), pp.4-25. ​ B oas, T., Hidalgo, F. and Richardson, N. (2014). The Spoils of Victory: Campaign Donations and ​ Government Contracts in Brazil. The Journal of Politics, 76(2), pp.415-429. ​ B ogdan, S. (2018). De ce, în România, ne urâm de moarte pe tema anticorupției. Vice. [online] Available ​ ​ at: https://www.vice.com/ro/article/gy3g33/de-ce-in-romania-ne-uram-de-moarte-pe-tema-anticorupti ei [Accessed 25 Dec. 2019]. B rogaard, J., Denes, M. and Duchin, R. (2015). Political Connections, Incentives and Innovation: ​ Evidence from Contract-Level Data. Working paper. ​ B romberg, D. (2014). Can Vendors Buy Influence? The Relationship Between Campaign Contributions ​ and Government Contracts. International Journal of Public Administration, 37(9), pp.556-567. ​ B rugués, F., Brugués, J., & Giambra, S. (2018). Political connections and misallocation of procurement ​ contracts: evidence from Ecuador. Caracas: CAF. ​ B ryman, A. (2016). Social Research Methods - 5th Edition. 5th ed. Oxford: Oxford University Press. C anayaz, M., Martinez, J. and Ozsoylev, H. (2015). Is the revolving door of Washington a back door to ​ excess corporate returns?. Working paper 1507 ​ C enter for Insights in Survey Research (2018). Public Opinion in Romania. [online] International ​ ​ Republican Institute. Available at: https://www.iri.org/sites/default/files/final_romania_poll_presentation.pdf [Accessed 25 Dec. 2019].

KTH ROYAL INSTITUTE OF TECHNOLOGY 62 Daniel Pustan ​ ​

C entrul pentru Integritate în Business (2013). Mandatele politice. [ebook] Available at: ​ ​ https://www.businessintegrity.ro/sites/default/files/fileuploads/newsletter/91%20-%20Mandatele% 20functiilor%20publice.pdf [Accessed 22 Dec. 2019]. C harron, N., Dahlström, C., Fazekas, M. and Lapuente, V. (2017). Careers, Connections, and Corruption ​ Risks: Investigating the Impact of Bureaucratic Meritocracy on Public Procurement Processes. The ​ Journal of Politics, 79(1), pp.89-104. Ciubotaru, B. (2013). Organizarea şi funcţionarea Guvernului României: legislație, doctrină și practică politică. Iaşi: Lumen. C oviello, D. and Gagliarducci, S. (2017). Tenure in Office and Public Procurement. American Economic ​ ​ ​ Journal: Economic Policy, 9(3), pp.59-105. D avid, Daniel (2015), Psihologia poporului român, Editura Polirom, ISBN 978-973-46-5478-9 D ávid-Barrett, E. and Fazekas, M. (2019). Grand corruption and government change: an analysis of ​ partisan favoritism in public procurement. European Journal on Criminal Policy and Research. ​ d e Maria, B. (2008). Neo colonialism through measurement: a critique of the corruption perception ​ - index. Critical perspectives on international business, 4(2/3), pp.184-202. ​ D i Tella, R. and Schargrodsky, E. (2003). The Role of Wages and Auditing during a Crackdown on ​ Corruption in the City of Buenos Aires. The Journal of Law and Economics, 46(1), pp.269-292. ​ D irectorate-General for Regional and Urban Policy (European Commission) and PWC (2016). Stock-taking of administrative capacity, systems and practices across the EU to ensure the compliance and quality of public procurement involving European Structural and Investment (ESI) Funds. [online] Luxembourg Publications Office of the European Union, pp.180-188. Available at https://op.europa.euenpublication-detail-publication60bf8ebc-0204-11e6-b713-01aa75ed71a1langua ge-en [Accessed 7 Dec. 2019]. D oroftei, I. M. (2016). Measuring Government Favouritism Objectively: The Case of Romanian Public ​ Construction Sector. European Journal on Criminal Policy and Research, 22(3), pp.399-413. ​ D oroftei, M., and Dimulescu, V. (2015). Romanian Public Procurement in the Construction Sector: ​ Corruption Risks and Particularistic Links, Deliverable D8.1.5 ANTICORRP. Bucharest: Romanian ​ Academic Society. E scresa, L. and Picci, L. (2017). A New Cross-National Measure of Corruption. The World Bank Economic ​ ​ Review, 31(1), pp.196–219. E uractiv (2014). Public procurement corruption in Romania. [online] Euractiv.com. Available at: ​ ​ https://euractiv.eu/wp-content/uploads/sites/2/special-report/euractiv_special_report_public_pro curement_corruption_in_romania.pdf [Accessed 2 Dec. 2019]. E uropean Commission (2014). EU Anti-corruption report. Report from the Commission to the Council ​ ​ and the European Parliament. [online] Brussels: COM(2014) 38 final. Available at: https://ec.europa.eu/home-affairs/sites/homeaffairs/files/e-library/documents/policies/organized-c rime-and-human-trafficking/corruption/docs/acr_2014_en.pdf [Accessed 26 Dec. 2019].

KTH ROYAL INSTITUTE OF TECHNOLOGY Government favoritism in public procurement: evidence from Romania ​ ​63 ​

E uropean Commission (2017). On Progress in Romania under the Co-operation and Verification ​ Mechanism. Report from the Commission to the European Parliament and the Council. [online] ​ Brussels: Com(2017) 44 final. Available at: https://ec.europa.eu/info/sites/info/files/com-2017-44_en_1.pdf [Accessed 19 Dec. 2019]. E uropean Commission. (n.d.). Corruption. [online] Available at ​ ​ https://ec.europa.euhome-affairswhat-we-dopoliciesorganized-crime-and-human-traffickingcorrupti on_en [Accessed 27 Dec. 2019]. E uropean Parliamentary Research Service (2018). Listing of tax havens by the EU. [online] European ​ ​ Union. Available at: https://www.europarl.europa.eu/cmsdata/147404/7%20-%2001%20EPRS-Briefing-621872-Listing-t ax-havens-by-the-EU-FINAL.PDF [Accessed 27 Dec. 2019]. F azekas, M. (2017) Red tape, bribery and government favouritism: evidence from Europe. Crime Law Soc ​ ​ Change (2017) 68: 403. F azekas, M. and Kocsis, G. (2017). Uncovering High-Level Corruption: Cross-National Objective ​ Corruption Risk Indicators Using Public Procurement Data. British Journal of Political Science, ​ pp.1-10. F azekas, M., Cingolani, L. and Tóth, B. (2016), A Comprehensive Review of Objective Corruption Proxies ​ in Public Procurement: Risky Actors, Transactions, and Vehicles of Rent Extraction. ​ GTI-WP/2016:03, Budapest: Government Transparency Institute. F azekas, M., Tóth, I. and King, L. (2013). Corruption Manual for Beginners: 'Corruption Techniques' in ​ Public Procurement with Examples from Hungary. SSRN Electronic Journal. ​ F azekas, M., Tóth, I. and King, L. (2016). An Objective Corruption Risk Index Using Public Procurement ​ Data. European Journal on Criminal Policy and Research, 22(3), pp.369-397. ​ F erraz, C. and Finan, F. (2008). Exposing Corrupt Politicians: The Effects of Brazil's Publicly Released ​ Audits on Electoral Outcomes. Quarterly Journal of Economics, 123(2), pp.703-745. ​ F erreira Rubio, D. and Andvig, E. (2019). Serious about sustainability? Get serious about corruption. ​ ​ World Economic Forum. [online] Available at: https://www.weforum.org/agenda/2019/09/serious-about-sustainability-get-serious-about-corrupti on/ [Accessed 29 Dec. 2019]. F isman, D., Fisman, R., Galef, J., Khurana, R. and Wang, Y. (2012). Estimating the Value of Connections ​ to Vice-President Cheney. The B.E. Journal of Economic Analysis & Policy, 12(3). ​ F rance24 (2018). Romania party leader denies corruption in trial over fake jobs. [online] Available at ​ ​ https://www.france24.comen20180321-romania-party-leader-denies-corruption-trial-over-fake-jobs [Accessed 29 Dec. 2019]. G herghina, S., & Volintiru, C. (2017). A new model of clientelism: Political parties, public resources, and ​ private contributors. European Political Science Review, 9(1), 115-137. ​ ​ G olden, M. and Picci, L. (2005). Proposal for a new measure of corruption, illustrated with Italian data. ​ ​ Economics and Politics, 17(1), pp.37-75.

KTH ROYAL INSTITUTE OF TECHNOLOGY 64 Daniel Pustan ​ ​

G oldman, E., Rocholl, J. and So, J. (2013). Politically Connected Boards of Directors and The Allocation of ​ Procurement Contracts. Review of Finance, 17(5), pp.1617-1648. ​ G raycar, A. (2015) Corruption: Classification and analysis. Policy and Society, 34(2), 87-96. ​ ​ G ürakar, E. and Bircan, T. (2019). Political Connections and Public Procurement in Turkey. Crony ​ Capitalism in the Middle East, pp.236-262. ​ H all, B. (2019). Romanian government on brink of collapse. Financial Times. [online] Available at: ​ ​ https://www.ft.com/content/3a22e218-c8d6-11e9-a1f4-3669401ba76f [Accessed 29 Dec. 2019]. H icken, A. (2011). Clientelism. Annual Review of Political Science, 14(1), pp.289-310. ​ ​ H ill, R., Griffiths, W. and Lim, G. (2011). Principles of Econometrics. 4th ed. New York: Wiley. H untington, S. (1968). Political order in changing societies. New Haven, CT: Yale University Press. ​ ​ H yytinen, A., Lundberg, S. and Toivanen, O. (2018). Design of public procurement auctions: evidence ​ from cleaning contracts. The RAND Journal of Economics, 49(2), pp.398-426. ​ I slam, R. (2006). Does more transparency go along with better governance?. Economics and Politics, ​ ​ 18(2), pp.121-167. J ackowicz, K., Kozłowski, Ł. and Mielcarz, P. (2014). Political connections and operational performance of ​ non-financial firms: New evidence from Poland. Emerging Markets Review, 20, pp.109-135. ​ J ain, A. K. (2001), Corruption: A Review. Journal of Economic Surveys, 15: 71-121. ​ ​ J ohnston, M. (1996), The search for definitions: the vitality of politics and the issue of corruption. ​ ​ International Social Science Journal, 48: 321-335. K aufmann, D., Kraay, A. and Mastruzzi, M. (2011). The Worldwide Governance Indicators: Methodology ​ and Analytical Issues. Hague Journal on the Rule of Law, 3(02), pp.220-246. ​ K enny, C. (2009). Measuring Corruption in Infrastructure: Evidence from Transition and Developing ​ Countries. The Journal of Development Studies, 45(3), pp.314-332. ​ K nack, S. (2007). Measuring Corruption: A Critique of Indicators in Eastern Europe and Central Asia. ​ ​ Journal of Public Policy, 27(3), pp.255-291. K urer, O. (1993). Clientelism, Corruption, and the Allocation of Resources. Public Choice, 77(2), ​ ​ pp.259-273. K urer, O. (2005). Corruption: An Alternative Approach to its Definition and Measurement. Political ​ ​ Studies, 53(1), pp.222-239. K urtz, M. and Schrank, A. (2007). Growth and Governance: Models, Measures, and Mechanisms. The ​ ​ Journal of Politics, 69(2), pp.538-554. L ambsdorff, J., (2006). The validity and precision of subjective indicators (CPI). In: Sampford, C., ​ ​ Shacklock, A., Connors, C., Galtung, F. (Eds.), Measuring Corruption. Ashgate, Aldershot, pp. 81–99. L ancaster, T. and Montinola, G. (2001). Comparative political corruption: Issues of operationalization and ​ measurement. Studies in Comparative International Development, 36(3), pp.3-28. ​ L eff, N. H. (1964). Economic Development Through Bureaucratic Corruption. American Behavioral ​ ​ Scientist, 8(3), pp. 8–14.

KTH ROYAL INSTITUTE OF TECHNOLOGY Government favoritism in public procurement: evidence from Romania ​ ​65 ​

L eys, C. (1965). What is The Problem About Corruption?. The Journal of Modern African Studies, 3(2), ​ ​ pp.215-230. L indstedt, C. and Naurin, D. (2010). Transparency is not Enough: Making Transparency Effective in ​ Reducing Corruption. International Political Science Review, 31(3), pp.301-322. ​ L uechinger, S. and Moser, C. (2014). The value of the revolving door: Political appointees and the stock ​ market. Journal of Public Economics, 119, pp.93-107. ​ L ui, F. (1985). An Equilibrium Queuing Model of Bribery. Journal of Political Economy, 93(4), ​ ​ pp.760-781. L ukkarinen, J., Morild, K. and Jönsson, S. (2015). Facts and figures on public procurement in Sweden. ​ ​ The Swedish Competition Authority’s report series 2015:9. [online] Stockholm: The Swedish Competition Authority and the National Agency for Public Procurement. Available at: http://www.konkurrensverket.se/globalassets/publikationer/rapporter/rapport_2015-9_english.pdf [Accessed 27 Dec. 2019]. M arcovici, T. (2019). Firma cumnatului premierului Ponta abonată la contracte cu statul. Cotidianul. ​ ​ [online] Available at: https://www.cotidianul.ro/firma-cumnatului-premierului-ponta-abonata-la-contracte-cu-statul/ [Accessed 19 Dec. 2019]. M arinas, R. and Dragin, S. (2017). Tens of thousands rally in Romania against planned emergency ​ decrees. Reuters. [online] Available at: ​ https://www.reuters.com/article/us-romania-protests/tens-of-thousands-rally-in-romania-against-p lanned-emergency-decrees-idUSKBN15D118 [Accessed 19 Dec. 2019]. M arkandya, A. and Pearce, D. (1988). Natural environments and the social rate of discount. Project ​ ​ Appraisal, 3(1), pp.2-12. M auro, P. (1995). Corruption and Growth. The Quarterly Journal of Economics, 110(3), pp.681-712. ​ ​ M ediafax (2009). Guvernul Boc 2 a fost demis. [online] Available at: ​ ​ https://www.mediafax.ro/politic/guvernul-boc-2-a-fost-demis-video-4985350 [Accessed 27 Dec. 2019]. M éon, P. and Sekkat, K. (2005). Does corruption grease or sand the wheels of growth? Public Choice, ​ ​ 122(1-2), pp.69-97. M ihalache, D., (2013) Achizitii publice 2007-2015 (Contracte). Romanian Digital Agenda Agency. [data] ​ ​ Available at: http://data.gov.ro/dataset/achizitii-publice-2007-2016-contracte6 [Accessed 17 Dec. 2019] M inistry of Finance (2018). Date de identificare plătitori actualizate august 2018. [data] Available at: ​ ​ http://data.gov.ro/dataset/informatii_fiscale_august_2018 [Accessed 17 Dec. 2019] M ironov, M. and Zhuravskaya, E. (2012). Corruption in procurement and shadow campaign financing: ​ Evidence from Russia. MIMEO. ​ M o, P. (2001). Corruption and Economic Growth. Journal of Comparative Economics, 29(1), pp.66-79. ​ ​ M ungiu, A. (2006). Corruption: Diagnosis and Treatment. Journal of Democracy, 17(3), pp.86-99. ​ ​

KTH ROYAL INSTITUTE OF TECHNOLOGY 66 Daniel Pustan ​ ​

M ungiu-Pippidi, Alina (editor) 2015, Government Favouritism in Europe: Anticorruption report 3, Barbara Budrich Publishers N ational Anticorruption Directorate (2017). Raport privind activitatea desfășurată. [online] Bucharest. ​ ​ Available at: https://www.pna.ro/bilant_activitate.xhtml?id=40 [Accessed 19 Dec. 2019]. O ECD (2016). Preventing Corruption in Public Procurement. [online] Secretary-General of the OECD. ​ ​ Available at: http://www.oecd.org/gov/ethics/Corruption-Public-Procurement-Brochure.pdf [Accessed 26 Dec. 2019]. O lken, B. (2006). Corruption and the costs of redistribution: Micro evidence from Indonesia. Journal of ​ ​ Public Economics, 90(4-5), pp.853-870. O lken, B. (2007). Monitoring Corruption: Evidence from a Field Experiment in Indonesia. Journal of ​ ​ Political Economy, 115(2), pp.200-249. O lken, B. (2009). Corruption perceptions vs. corruption reality. Journal of Public Economics, 93(7-8), ​ ​ pp.950-964. O man, C. and C. Arndt (2006), Uses and Abuses of Governance Indicators, Development Centre Studies, ​ ​ OECD Publishing, Paris. O prea, A. (2017). Registrul Comerţului: În România există 1,4 milioane de firme, din care doar 1,2 ​ milioane sunt active. Cele mai multe insolvenţe sau radieri se înregistrează în comerţ, construcţii şi transporturi. News.ro. [online] Available at: ​ https://www.news.ro/economic/registrul-comertului-in-romania-exista-1-4-milioane-de-firme-din-c are-doar-1-2-milioane-sunt-active-cele-mai-multe-insolvente-sau-radieri-se-inregistreaza-in-comert- constructii-si-transporturi-1922400517302017051116986291 [Accessed 27 Dec. 2019]. P olitica Românească (2011). Guvernul Emil Boc (2). [online] Available at: ​ ​ http://www.politicaromaneasca.ro/guvernul_emil_boc_2-2638 [Accessed 22 Dec. 2019]. R einikka, R. and Svensson, J. (2004). Local Capture: Evidence from a Central Government Transfer ​ Program in Uganda. The Quarterly Journal of Economics, 119(2), pp.679-705. ​ R omanian Agency for Digital Agenda (2018). Lansarea noului Sistem Electronic de Achizitii Publice. ​ ​ [online] Available at: https://www.aadr.ro/-lansarea-noului-sistem-electronic-de-achizitii-publice_227_0.html [Accessed 7 Dec. 2019]. R osca, A. (2018). Daniel David despre psihologia poporului român. Ce ne definește, ca nație, și cum ne ​ putem schimba. [Blog] Andreea Rosca Playground. Available at ​ https://www.andreearosca.ro/daniel-david [Accessed 1 Dec. 2019]. ​ ​ R ose-Ackerman, R. (1997). The political economy of corruption. In: Elliott, K.A. (Ed.), Corruption and the ​ ​ Global Economy, Institute for International Economics, Washington, DC, pp. 31–60. R ose-Ackerman S. (2008) Corruption. In: Readings in Public Choice and Constitutional Political ​ ​ Economy. Springer, Boston, MA. R othstein, B. and Teorell, J. (2008), What Is Quality of Government? A Theory of Impartial Government ​ Institutions. Governance, 21: 165-190. ​

KTH ROYAL INSTITUTE OF TECHNOLOGY Government favoritism in public procurement: evidence from Romania ​ ​67 ​

S tephenson, M. (2014). Can Foreign Anti-Bribery Enforcement Statistics Help Us Measure Corruption ​ Levels Objectively?. [Blog] The Global Anticorruption Blog. Available at: ​ https://globalanticorruptionblog.com/2014/04/15/can-foreign-anti-bribery-enforcement-statistics-h elp-us-measure-corruption-levels-objectively/ [Accessed 29 Nov. 2019]. S tokes, S. (2005). Perverse Accountability: A Formal Model of Machine Politics with Evidence from ​ Argentina. American Political Science Review, 99(3), pp.315-325. ​ S vensson, J. (2005). Eight Questions about Corruption. Journal of Economic Perspectives, 19(3), ​ ​ pp.19-42. S chwab, K. (2016). The Global Competitiveness Report 2016–2017. [online] Geneva: the World Economic ​ ​ Forum. Available at: http://www3.weforum.org/docs/GCR2016-2017/05FullReport/TheGlobalCompetitivenessReport201 6-2017_FINAL.pdf [Accessed 29 Dec. 2019]. S øreide, T. (2002) Corruption in public procurement: Causes, consequences and cures, Chr. Michelsen ​ ​ Institute Development Studies and Human Rights T anzi V., Davoodi H. (1998) Corruption, Public Investment, and Growth. In: Shibata H., Ihori T. (eds) The ​ ​ Welfare State, Public Investment, and Growth. Springer, Tokyo T imu, A. and Zega, R. (2017). Romanian Premier Refuses to Quit as Protests Grow Despite U-Turn. ​ ​ Bloomberg. [online] Available at: https://www.bloomberg.com/news/articles/2017-02-04/romanian-government-to-scrap-anti-graft-d ecree-amid-protests [Accessed 17 Dec. 2019]. T itl, V. and Geys, B. (2019). Political donations and the allocation of public procurement contracts. ​ ​ European Economic Review, 111, pp.443-458. T oma, V. (2016). CCR despre abuzul in serviciu: "în mod defectuos" inseamna "incălcare a legii." Revista ​ ​ 22. [online] Available at: https://revista22.ro/actualitate-interna/ccr-despre-abuzul-in-serviciu-n-mod-defectuos-inseamna-in clcare-a-legii [Accessed 20 Dec. 2019]. T ran, M. (2015). Bucharest nightclub fire: PM and government resign after protests. The Guardian. ​ ​ [online] Available at: https://www.occrp.org/en/daily/4567-romania-prime-minister-ponta-resigns-amid-protests-over-ni ghtclub-fire [Accessed 22 Dec. 2019]. W eber Abramo, C. (2008). How Much Do Perceptions of Corruption Really Tell Us?. Economics: The ​ ​ Open-Access, Open-Assessment E-Journal, 2(2008-3), p.1. W erner, S. (1983). New Directions in the Study of Administrative Corruption. Public Administration ​ ​ Review, 43(2), p.146-154. X u, X. (2016). Corruption and economic growth: an absolute obstacle or some efficient grease?. Economic ​ ​ and Political Studies, 4(1), pp.85-100.

KTH ROYAL INSTITUTE OF TECHNOLOGY 68 Daniel Pustan ​ ​

Annexes Annex A

Table 9 Summary of selected studies using objective indicators of corruption ​ Source: Fazekas, Tóth and King 2016, p. 2 ​ ​ paper indicator used Country year sector

Auriol, Straub and Exceptional procedure type Paraguay 2004-200 general Flochel (2016) 7 procurement ​ ​ Bandiera, Prat, Price differentials for standard goods Italy 2000-200 various and Valletti (2009) purchased locally or through a 5 standardized ​ ​ national procurement agency goods (e.g. paper)

Coviello and Number of bidders Italy 2000-200 general Gagliarducci Same firm awarded contracts 5 procurement (2017) recurrently ​ ​ Level of competition

Di Tella and Difference in prices of standardized Brazil 1996-1997 health care Schargrodsky products such as ethyl alcohol (2003) ​ ​ Ferraz and Finan Corruption uncovered by federal Brazil 2003 federal-local (2008) audits of local government finances transfers ​ ​

Golden and Picci Ratio of physical stock of Italy 1997 infrastructure (2005) infrastructure to cumulative spending ​ ​ on infrastructure

Goldman, Rocholl Political office holders' position on USA 1990-2004 general and So (2013) company boards procurement ​ ​ Hyytinen, Number and type of invited firms Sweden 1990-1998 cleaning Lundberg and Use of restricted procedure services Toivanen (2018) ​ ​ Olken (2006) Difference between the quantity of Indonesia 1998-1999 welfare ​ ​ in-kind benefits (rice) received spending according to official records and reported survey evidence

Olken (2007) Differences between the officially Indonesia 2003-200 infrastructure ​ ​ reported and independently audited 4 (roads) prices and quantities of road construction projects

Klasnja (2015) Single bidder procedures Romania 2008-2012 general Non-open procedure types procurement

Reinikka and Difference between block grants Uganda 1991-1995 education Svensson (2004) received by schools according to ​ ​ official records and user survey

KTH ROYAL INSTITUTE OF TECHNOLOGY Government favoritism in public procurement: evidence from Romania ​ ​69 ​

Annex B

T able 10 Corruption techniques and indicators ​ Source: Fazekas, Tóth and King 2013, p. 63 ​ ​ No. Name Direct indicator Indirect indicator 1 Defining - - unnecessary needs 2 Defining needs to A) Prevalence of avoiding centralised benefit a procurement particular supplier 3 Tinkering with A) Proportion of non-open D) Contract value according to Public thresholds and procedures Procurement Law/total procurement exceptions contract value B) Average corruption risk score of procedures followed C) Frequency of actual contract value above estimated contract value 4 Tailoring A) Length of eligibility criteria - eligibility criteria 5 Abusing formal A) Length of eligibility criteria - and B) Proportion of excluded bids administrative requirements 6 Tailoring A) Length of evaluation criteria - evaluation criteria B) Weight of non-price criteria 7 Using long term A) Combined value of framework - complex contracts contracts and PPPs / total contract value B) Average contract duration 8 Tinkering with A) Proportion of tenders with - submission period accelerated submission periods B) Proportion of tenders with extremely short submission periods C) Average contract value per weekday available for submission 9 Selective - A) Proportion of tenders with extremely information short submission periods provision B) Proportion of procedures with call for tenders modified within all procedures 10 Avoiding A) Proportion of tenders without call - publication of call for tenders in the official journal

KTH ROYAL INSTITUTE OF TECHNOLOGY 70 Daniel Pustan ​ ​

for tenders 11 Strategically A) Proportion of procedures with call - modifying call for for tenders modified within all tenders procedures 12 Excessively pricey A) Price of documentation/estimated - documents, contract value difficult access to documents 13 Deliberate errors A) Prevalence of extremely erroneous - in document contract award announcements publication B) Hiding or erroneously reporting the final contract value 14 Strategically A) Proportion of annulled procedures B) Decrease in the number of bids annulling re-launched subsequently received in subsequent rounds procedures 15 Repeated A) Repeated court rulings against the - violations of issuer within the same procedure public procurement rules 16 Unfair scoring - A) Average contract value per weekday available for decision B) Length of evaluation criteria C) Weight of non-price criteria 17 Abusing unit A) Proportion of contracts using unit - prices in the prices contract 18 Modifying A) Proportion of modified contracts - contracts strategically B) Difference between the awarded and final contract value C) Difference between originally planned and final completion period 19 Abusing add-on A) Proportion of add-on contracts - contracts B) Proportion of contracts exhausting the planned reserves 20 Performance - - violating contract

KTH ROYAL INSTITUTE OF TECHNOLOGY Government favoritism in public procurement: evidence from Romania ​ ​71 ​

Annex C T able 11 Summary of selected studies on firm’s political connections ​ Paper Country Sample Political power Proxy for political period level connection

Public procurement Mironov and Russia 6 years regional elections illegal donations to parties Zhuravskayay (2012) ​ ​ Goldman, Rocholl and So USA 1990-2004 House and Senate revolving door (2013) (Parliamentary) ​ ​ Bromberg (2014) USA 2001-2006 Federal donation ​ ​ government / House of Representative Straub (2014) Paraguay 2004-2011 government / *member or sympathisant parliamentary Boas, Hidalgo and Brasil 2004-2010 Chamber of donation Richardson (2014) Deputies ​ ​ Luechinger and Moser USA 1993-2003 Department of revolving door (2014) defense ​ ​ Brogaard, Denes and USA 2000-2012 federal donation Duchin (2015) government ​ ​ Doroftei(2016) Romania 2007-2013 donation + connections ​ ​ Doroftei and Dimulescu Romania 2007-2013 donation + connections (2015) ​ ​ Auriol, Straub and Flochel Paraguay 2004-2007 (2016) ​ ​ Titl and Geys (2019) Czech 2007-2014 13 regional donation ​ ​ Republic governments Arvate, Barbosa and Brasil 2007-2010 state legislatures donation Fuzitani (2018) ​ ​ Brugués F., Brugués J. and Ecuador 2006-2018 bureaucrat (incl. revolving door Giambra S. (2018) mayors and local ​ ​ counselors)

Johnson and Roer (2018) USA 2007-2016 House "relocation" and "selection" Appropriations influence subcommittee Ferris, Houston and USA 2006-2013 donation Javakhadze (2019) Gürakar and Bircan (2019) Turkey 2004-2011 Parliament and revolving door ​ ​ local governmental bodies

KTH ROYAL INSTITUTE OF TECHNOLOGY 72 Daniel Pustan ​ ​

Stock market Roberts (1990) USA 1983-09-01 congressional donation seniority

Fisman (2001) Indonesia 1995-1997 President Suharto Suharto Dependency Index Faccio (2006) 47 1991 see the proxy for revolving door countries political connection

Jayachandran (2006) USA 2001-05 donation Cooper, Gulen and USA 1979-2004 House and Senate donation Ovtchinnikov (2010) Fisman et al. (2012) USA 2000-01-03 - revolving door ​ ​ 2001-04-30

Amore and Bennedsen Denmark 2002-2008 local municipality family ties (2013) ​ ​ Gropper, Jahera Jr. and USA 1989-2010 Congress banks' headquartered in the Park (2013) same state with the committee chair Jackowicz, Kozłowski and Poland 2001-2011 central revolving door Mielcarz (2014) government / ​ ​ national parliament / local authorities Coulomb and Sangnier France 2007-01-01 - presidential Sarkozy Index (media) (2014) 2007-05-06 Luechinger and Moser USA 1993-2003 Department of revolving door (2014) defense ​ ​ Akey (2015) USA 1998-2010 Senators and donation Representatives Canayaz, Martinez and USA 1990-2012 government revolving door Ozsoylev (2015) positions ​ ​

Financing Khawaja and Mian (2005) Pakistan 1996-2002 national elections if firm's director participates in an election Fraser, Zhang and Malaysia 1990-1999 government political patronage + Derashid (2006) informal ties Faccio, Masulis and 35 1997-2002 revolving door+friendship McConnell (2006) countries

Claessens, Feijend and federal deputy donations - contribution to Laeven (2008) Brasil 1998 & 2002 candidates campaigns Bliss and Gul (2012) Malaysia 2001-2004 government informal ties

KTH ROYAL INSTITUTE OF TECHNOLOGY Government favoritism in public procurement: evidence from Romania ​ ​73 ​

Blau, Brough and Thomas USA 2008 federal revolving door + lobbying (2013) government expenditures

Yeh, Shu and Chiu (2013) Taiwan 1998-2006 revolving door Infante and Piazza (2014) Italy 2005-09 - national revolving door 2009-03 parliament and all local councils

Houston, Maslar, and USA 2004-2012 White House, donations Pukthuanthong (2018) Senate and Congress

Disaster relief Nikolova and Marinov Bulgaria 2004-2005 (2017)

Annex D

T able 12 Descriptive statistics ​

year mean sd min max p50 sum

2009 97276.44 1777334 0 2.79e+08 3751.26 1.19e+10

2010 87914.86 1308800 0 1.78e+08 3610.09 1.09e+10

2011 157587.2 2387378 0 2.61e+08 6346.42 1.66e+10

2012 141756.1 2609191 0 3.07e+08 5766.765 1.53e+10

2013 215950.2 3063701 0 2.91e+08 5611.643 1.44e+10

2014 239164.9 2658020 0 2.30e+08 6515.018 1.76e+10

2015 247728.3 3747858 0 7.76e+08 7246.912 1.73e+10

Total 155311.2 2463122 0 7.76e+08 5093.455 1.04e+11

KTH ROYAL INSTITUTE OF TECHNOLOGY