INFORMAL PAYMENTS IN ROMANIAN HEALTH CARE SYSTEM. A SAMPLE SELECTION CORRECTION

This paper draws on economic theory, sociology and political science approaches to explain informal payments in the Romanian health care system. It estimates the likelihood of paying a bribe (informal payment) using a reduced health care demand equation in a probit model with sample selection correction. Social capital, as having a relationship with doctors, and the perception of the health care system, as corrupt, are found to infl uence the probability of making an informal payment. The likelihood of making an informal payment in the Romanian health care system is modelled using a maximum-likelihood probit estimation with sample selection correction. In the selection equation, reduced health care demand, self-perceived health status and being afraid of diseases are used as exclusion restrictions for identifying the parameters of the econometric model.


INTRODUCTION
Atenţie is the Romanian word defi ning an informal payment (in the form of money, gifts or services) in addition to the cost of the services paid to the physicians, doctors and nurses in the Romanian public health care system 2 . Some fi gures, calculated from the Barometer of public opinion [2004][2005], indicated that the payment of "atenţie" was widespread. About 30% of the population older than 18 years old has been involved in informal payments to receive medical care in 2005. In 2017, the fi gure was 18% according to Eurobarometer 470 3 . Additionally, based on fi gures from Lewis (2002) and the World Health Organization (WHO) data set, Romanians spend around 450 mills US PPP$ in "atenţie". The out of pocket payments for health care in Romania is at one of the highest levels paid in Europe.
This study aims to assess the profi le of the informal payer and to investigate the mechanisms of paying a bribe in the health care system. The paper's contribution is twofold. First, it sheds light on the perception of corruption in the health care system and the social capital as determinants of the probability of making an informal payment. Second, it circumvents potential biases in the estimation of the likelihood of making an informal payment, given its nature as a binary variable, by using a probit model with sample selection correction.
The phenomenon of informal payments in the health care system is widespread in Central and Eastern Europe. Studies conducted by Kornai (2000) and Lewis (2002) suggest that gratitude payments are aff ecting the provision of health care in post-communist countries like Hungary and Poland. Gaal (1998) indicates a positive correlation between informal payments in health care, and the publicly controlled health care systems. Belli (2002) provides further information regarding the allocation of payment types across diff erent health care facilities. Thus, in the Czech Republic, Hungary, Poland and Romania formal payments 4 are mostly associated with primary and outpatient specialist care, whereas informal payments are more associated with surgery and inpatient care. Informal out of pocket payments are important since they maintained the Romanian health care system functional in the context of under-funded budgets and low wages paid to the medical staff (See Fig.1 and 2 from the Annex).
The paper is divided in seven sections. In the next section, a theoretical background is presented. It discusses three approaches to informal payments in health care systems: an economic approach, a sociological approach and a political science approach. The third section provides a description of the Romanian health care system illustrating some of the theoretical concepts discussed. The fourth, fi fth and sixth section are devoted to develop the model, describing the data and discussing the results, respectively. Finally, the seventh section concludes.

THEORETICAL BACKGROUND
Informal payments in health care are defi ned as "payments to individual and institutional providers, in kind or in cash, which are made outside offi cial channels". This encompasses "«envelope» payments to physicians and «contributions» to hospitals as well as the value of medical supplies purchased by patients and drugs obtained from private pharmacies but intended to be part of government-fi nanced health care services" (Lewis, 2007). Since medical staff is using their positions for private gains and patients are expected to provide an extra payment for something they were entitled to receive, informal payments can be seen as an act of corruption. Gupta et al. (2000) show that across countries there is a strong correlation between corruption and service provision, countries with less corruption and higher predictability of corruption have better quality health care and more effi cient provision of public services.
The phenomenon of informal payments in health care systems is analysed from three perspectives. First, an economic approach emphasizes the role of incentives and market failures (Ensor, 2004); as well as the role of government failures (Balanova and McKee, 2002). Second, a sociological approach emphasizes the roles of institutions and the culture of gratitude payments placing the phenomenon into a context of informal economies (Kornai, 2000). The role of the communist past plays the main reason for generating a second parallel economy in which fostering relationships with providers of goods and services is the key to access scarce resources (Ledeneva, 1998). Third, a political science approach emphasizes the lack of accountability and control (Lewis, 2002(Lewis, , 2007 as well as the lack of government legitimacy (Rose-Ackerman, 2001).

THE ECONOMIC APPROACH
Economic theory can give insights fi rst with respect to problems inherent in the nature of health care as a good and second with respect to incentives and failures that derive from its nature and from the interplay of demand and supply in the market. Preker and Harding (2000) emphasize the fact that health care activities produce externalities and they are not pure public goods as they have some elements of exclusion, and rivalry making them private goods and hence marketable 5 .
Economics also emphasizes that health care by nature is a good that has problems like: a) asymmetry of information, which makes patients reliant on doctors; b) adverse selection, making the market reluctant to insure the most vulnerable; and c) moral hazard of over-consumption and over-production.
Given the nature of health care as a good, social insurance systems have been organized to pool the risk of sickness over the population, diminishing the costs paid by the most vulnerable and addressing the adverse selection problem.
The asymmetry of information between the patients and the health care provider and between the doctor and the health care administrators, gives the doctor a key power position in the delivery of health care that can be used to extract rents (Ensor, 2004). In this line of analysis, monopolies can occur due to: (a) legal restrictions on competition; (b) access to subsidized capital and revenues, creating an uneven "playing fi eld"; (c) below cost distribution of goods and services to achieve equity goals; or (d) production of public goods or goods where markets are not viable (Preker and Harding, 2000).
The monopolies of health care occurred during communism in most Central and Eastern European countries. These systems are funded directly from the government budgets and publicly provided.
The nationalization of the health care sector led to an increased demand (Gaal and McKee, 2004). This increased demand was not matched by increased funding, creating shortages in supply (Gaal and McKee, 2004).
The low funding levels meant scarce human resources and scarcity of equipment. This was coupled with constraints on the number of bed days and the fact that funding was given based on the levels of past demand (Ensor, 2004). The gap between the available resources and those required can lead to several responses (Ensor, 2004): a) health care workers can reduce quality of health care and treat patients below the required standard; b) health care workers reduce the numbers of patients treated by increasing the number of days of treatment and create queues; c) health care workers reduce both the treatment and the numbers of patients treated; d) health care workers can reduce the number of hours worked, and moonlight since they are not appropriately remunerated in the public system (Gaal and McKee, 2004). Patients can reduce the gap between the budget and the cost of health care by paying for medical supplies, drugs necessary for their treatment and buying their food (in case of inpatient care). Informal payment can also increase the salaries of the medical staff up to their reservation wage ensuring that they do not change their occupation and that they have incentives to provide health care (Ensor, 2004).
Health care workers can also exploit their power and market position to extract rents from patients. Market power derives from a position of monopoly and the principal-agent relationship existing between practitioner and patient (Ensor, 2004). The health care worker can delay treatment until informal payment is done, create a bottleneck in the provision of the service and afterwards rationing to those that pay (Ensor, 2004) Thus health care workers can demand a contribution which is higher than the average cost of the treatment. Its level is higher if the probability of being caught and punished is low.
Health care workers can also off er additional services or higher quality of treatment for which they require informal payments. From a diff erent perspective, they can also capitalize on information asymmetry to maximize income by providing unnecessary services (Gaal and McKee, 2004).
To sum up, on the demand side, informal payments can serve for multiple purposes: jumping the queue, shortening the duration of waiting for the treatment, and receipt of more care. To these are added increasing the quality of the medical service, obtaining drugs or ensuring the receipt of care in all instances, even when the medical staff has poor motivation, moreover, insurance against future care needs, access to care from specialists with good reputation (Lewis, 2007). Whereas, on the supply side, informal payments could be seen as a way to fi nance the costs of care, a way to increase performance or as rent-seeking by adding an extra price for making use of a scarce service on the market (Kornai, 2000). They could be used for private purposes or they could be used to fi ll in the lack of medical supplies required for the functioning of the service (Lewis, 2007).

THE SOCIOLOGICAL APPROACH
From a sociological perspective, traditions of gratitude inherited from communism generate informal payments (Kornai, 2000) as well as the tradition of "the use of personal networks and informal contacts to obtain goods and services in short supply and fi nd a way around formal procedures" and bureaucratic rigidities (Ledeneva, 1998). During communism, making favours in exchange for other services and goods created an informal second economy in which networks were built (Ledeneva, 1998). From this perspective, fostering a relationship with a health care worker through informal payments can act as insurance in case of future health care needs. According to Richard Rose (1998) "social capital is defi ned as the stock of formal and informal networks that individuals use to produce or allocate goods and services". Social capital is therefore about recurring relationships between individuals. An alternative approach (Rose, 1998) "treats social capital as a social psychological or as cultural beliefs and norms in Inglehart's view (1997, 188) «a culture of trust and tolerance in which extensive networks of voluntary associations emerge»". Therefore, in this view social capital is seen as a predisposition to cooperation and trust which is essential for both political and economic cooperation.
The communist legacy stands for lack of government legitimacy and the perception of rules as irrational. Rose-Ackerman (2001) draws attention on the fact that there is a communist legacy leading to low levels of trust in institutions and a habit of relying on inter-personal relations not on public institutions and laws. This can explain the perception of the health system as being corrupt and the habit of relying on interpersonal relations with medical providers.

THE POLITICAL SCIENCE APPROACH
The political science perspective emphasizes the lack of accountability of the health care providers, the inability of the government to enforce the rule of law and lack of private alternatives (Lewis, 2002(Lewis, , 2007Rose-Ackerman, 2001;Kornai, 2000). The government did not want to spend more on health care and it overlooked informal payments, as they were a way to fund the defi cits. From a political science approach the perception of the health care as corrupt can infl uence the probability to make an informal payment.

OUTCOMES OF MECHANISMS
Informal payments have consequences both on effi ciency and equity grounds. From an effi ciency point of view, the allocation of health care is likely to be distorted away from the optimum given that patients with higher resources are served before those unable to provide informal payments (Ensor, 2004). From an equity point of view, informal payments can be barriers in receiving and accessing care for poor patients and they create inequity in the distribution of health care. Studies prove that poor people give a larger share of their income in informal payments compared to rich people (Lewis, 2002(Lewis, , 2007 and that informal payments are highly regressive (Szlender and Cuyler, 2006).
The main mechanisms discussed in this section are further illustrated in the case of the Romanian health care system.

THE HISTORICAL DEVELOPMENT OF THE ROMANIAN HEALTH CARE SYSTEM AND THE FACTORS FAVOURING "ATENŢIE"
The Romanian health care system includes a public system based on social insurance and a private system funded by private insurance or out of pocket pay for the consultation or service provided. Doctors that work in the public system work in the private system as well increasing their wages. The Romanian health care system has been largely underfunded leading to poor quality of health services provision, ineffi ciency, infl exible norms and inadequate health care equipment and facilities (European Observatory on Health Care Systems, 2000). In 2014 5.7% of the GDP was spent on health care according to WHO estimates compared to Germany and Austria where around 11% of GDP is spent on health care. At EUR 814 per capita (adjusted for diff erences in purchasing power), Romania spends under a third of the EU average, the lowest level of all Member States on health care (OECD, 2017). According to OECD (2017) out-of-pocket spending comprises one fi fth of total health care expenditure, and includes direct and informal payments, aff ordability is the main reason for reported unmet health care needs.
According to the OECD (2017) although strengthening primary care has been on the policy agenda since 1990, primary and community health care services are still underprovided, and under-used, and there continues to be inappropriate use of inpatient and specialised outpatient care, including care in hospital emergency departments. Moreover, few resources were allocated for primary care and the majority was targeted at funding hospitals causing scarcity of medicine and equipment for general practitioners. Consequently, many Romanians had little confi dence in primary care and often sought care directly from the specialist and hospitals facilitating informal payments (Vladescu and Radulescu, 2001).
As the majority of Eastern European health care systems, the Romanian health care system was organized according to the centrally-coordinated Semashko health care system, which was state fi nanced (Bara, van den Heuvel, and Maarse, 2002) through the State Law on Health Organization passed in 1949. During the communist regime, the health system maintained the principles of universal coverage and free access, as well as health care at the point of delivery.
Primary health care in Romania had been provided mainly by dispensaries, which had been part of the hospital system and had served as primary health care centres for the population living in the area. Health care workers had a civil servants status (Bara, van den Heuvel and Maarse, 2002), and this had a major infl uence on their economic behaviour. This meant that doctors lacked monetary incentives to continue attending the patient other than their moral ethics, as wages were not defi ned by performance, but using fi xed rates 6 .
The organization and fi nancing of the health care system has begun a process of reform from 1997 onwards. Law 145/1997 introduced social insurance funding based on mandatory defi ned contributions separating service provision from fi nancing by creating the National Health Insurance House and its local branches 7 . In 1997, Romania introduced a modifi ed Bismarck model, which became eff ective since 1999 meaning the system is funded by social health care insurance depending on income (Stanciu, 2013). The law that governs the actual health care system is Law 95/2006, which has incurred numerous modifi cations throughout the time. According to Paun (2017), the health care system underwent a transition during the last three decades from a centralized system to a contract based system in which the health care providers make contracts with health insurance houses, which off er services to the population.
Health funding was supplemented with state budget complementary funds. The insurance covered a medical services package, which unfortunately was not clearly defi ned. The reform also established a new way of payment for the health care providers: since 1997 primary care physicians are paid by capitation 8 (Law 145/1997). This meant that they had more incentives to spend their time attracting new patients than treating those already registered with them, while fee-for-service arrangements were used for ambulatory specialised care (former polyclinics), and global budgets and salaries for hospitals (inpatient care) 9 . Patients were able to choose their general practitioner.
Health care was free for any medical condition, free annual check-ups, and drugs were subsidized (at least 50%). Despite the provisions, patients had problems in accessing health care, they complained of long queues for appointments and shortages or limited availability of medicines and investigations (Schneider, 2007). Payment by capitation did not diminish the informal payments since the doctors' wages remained low and the health system continued to be under-funded. The salaries of the doctors were up to 80% of the average salary until 2005. In 2018 due to increased migration of the medical staff , the salaries of doctors were increased up to 70%, which meant they could earn more than the average wage in order to curtail brain drain but despite eff orts, they remain low compared to other potential earnings abroad.

MODELING THE LIKELIHOOD OF MAKING AN INFORMAL PAYMENT IN THE ROMANIAN HEALTH CARE SYSTEM
Recent literature looks at informal payments through the lens of institutional theory (Williams and Horodnic, 2017). Most papers use a simple regression to model the willingness of the probability of paying a bribe using micro-data (Wiliams and Horodnic, 2017; Stepurko et al., 2015;Guerrero and Rodríguez-Oreggia, 2005).
The aim of this study is to calculate the propensity of making an informal payment in the Romanian health care system. We develop a model that takes into account the perception of corruption in the health care system and the role of social capital (having friends or relatives in the health care system). Due to the binary nature of the informal payment variable (atenție) and a potential incidental truncation (Wooldridge, 2010, 802;Greene, 2012, 872), we estimate a binary response model (probit) with sample selection correction (van de Ven and van Praag, 1981). The potential incidental truncation or sample selection bias refers to the fact that we can only observe the behaviour, regarding informal payments in the health care system, of those individuals who actually use the health care system. Thus, we estimate a model with two equations. On the one hand, we estimate the probability of using the health care system. This equation is called the selection equation. On the other hand, we estimate an equation of the likelihood of paying an informal payment only for those individuals in the sample who actually use the health care system. This equation is called outcome or structural equation.
The binary response model (probit) with sample selection correction was fi rst proposed by van de Ven and van Praag (1981). To correct for sample selection with a binary dependent variable in the structural equation, van de Ven and van Praag (1981) apply a methodology based on sample selection correction procedure, the so-called Heckit (Heckman selection model) after Heckman (1976) and Heckman (1979).
We estimate the propensity of making an informal payment in the Romanian health care system using the van de Ven and van Praag (1981) approach. The outcome (or structural) equation calculates the likelihood of making a gratitude payment based on demographic and labour related variables (including income) and the perception of corruption in the health care system and social capital (counting on reliable doctors within their network). The selection equation assesses the probability of using the health care system based on the self-perception of health that is also used as exclusion restriction in the structural equation. 10 The selection equation (likelihood of health care utilization) could be considered an estimation of the demand for health care, which is derived from the basic demand for health (Grossman, 1999). These kind of models use proxy variables of the consumption of health care (e.g. number of visits to the doctor) and are estimated through Poisson models or ordered probit models (Grossman, 1999;Vork, 2000). In this study, we use the van de Ven and van Praag (1981) approach to estimate a sort of reduced form of health care demand. We model the health care demand in a binary dependent variable setting (using utilization of health care system as dependent variable in the selection equation). Since the likelihood of gratitude payment in the health care system is only observed for those that use the system, the binary response model (probit) with sample selection correction is applied. The binary response model (probit) with sample selection correction allows us to model the decision of making an informal payment in the health care system since this decision is only observed for those individuals who attended the health system in the period of reference. The selection process is not a random one, but a process that can be characterized and measured. It is possible to calculate the likelihood of going to the health care facilities based on observed characteristics of the population in the sample.
For the empirical strategy, we apply the latent utility approach. We assume that every individual ( i ) in the sample optimizes his/her decision in order to maximize the utility: . The decision we are interested in is making an informal payment in the health care system ("atenție"): . is a binary variable which indicates whether the individual ( i ) makes an informal payment ( = 1) or does not Since it is not feasible to identify both parameters ( 1 and 0 ), we take the diff erence and make an implicit transformation for the likelihood of making an informal payment based on the latent utility ( * ) approach and the result of Eq. Where is a vector containing and the exclusion restriction variables, in this case self-perception of health. 11 To estimate the probability of making a gratitude payment with a sample selection correction, we assume a bivariate probit model formulation with an error term , independent of and with a bivariate normal distribution. Particularly, .e | (X_i,Z_i )~ N(0,Ω). Ω. is the 2 2 matrix with ones down its diagonal and off -diagonal element , (Wooldridge, 2010, 595). Under these assumptions, we write the model of the probability of making an informal payment in the health care system like: To provide consistent, asymptotically effi cient estimates for all the parameters, Eq. 5 is calculated with a maximum-likelihood probit model with sample selection correction. If 0 , the standard probit techniques applied to Eq. 5 yield biased results. We test indirectly the presence of selectivity by testing the null hypothesis The POBS's sampling was representative for the Romanian population over 18 years old, with an error of +/-3% in 2005. The sample was built as a stratifi ed sample probabilistic with three-stages sampling. The stratifi cation criteria were: 18 cultural areas grouped in historical provinces, residential areas (urban-rural), 4 types of sizes, 3 degrees of development of rural villages and towns. To select respondents, the last electoral lists have been used. Data were checked for consistency comparing its results with the National Statistics data and the results from 1992 and 2002 censuses. No weighting procedure was applied.
Among the several years of data available, only the 2005 survey provides enough variables to estimate the probability of making an informal payment in the Romanian health care system (atenţie). Among its questions, POBS 2005 includes a variable identifying whether people, that have used the health system, have paid an atenţie 13 . The question regarding health care demand, which is our dependent variable for the selection equation is framed as: "In the last fi ve years did you go to the doctor for a consultation, treatment or surgery?". The question regarding informal payments which is the dependent variable in our outcome equation is framed as: "If you went to the doctor, did you give anything else (money, gifts, services) than the legal fees?".
The POBS 2005 contains variables related to the perception of corruption in the health care system: "In your opinion how spread do you think corruption is among doctors?". Further, it includes a variable measuring social capital which is framed as: "Do you have relationships, or acquaintances on which you can count in case you get sick for a consultation, treatment or surgery?".
As a proxy of the health status, we use an indicator of subjective health status that is framed as: "How satisfi ed are you with your own health? This question is translated into a binary variable, where 1 means satisfi ed. The data does not contain any other variables measuring health status. However, since people who are afraid of getting sick could be more likely to use the health care system, we proxy this by using a variable regarding personal fears. This variable measures: "Which are the issues that you are most afraid of currently?, which have options such as: 1. prices; 2. war; 3. disease; 4. unemployment; 5. crime; 6. social unrest; 11. the future of my children; 12. terrorist attack; 95. something else; and, 97. I am not afraid of anything." These categories have been grouped in afraid of getting sick, afraid of other issues and not afraid of anything.
The POBS 2005 data contains information regarding demographic variables like: gender, age, education, experience with migration, household size, area where the person lives and residential size area. Additionally, it includes important variables concerning the labour market status, for instance: employment status, present or last occupation, and last monthly household income 14 . Since the variable measuring current occupation combines information regarding sector of employment and job type, we decided to construct a new variable measuring risky occupations. Bluecollar workers and people working in the agriculture are considered risky jobs as they imply hard physical work.
POBS 2005 data sheds some light about the extent to which Romanian citizens are involved in informal payments to their health providers. About 33% of the citizens were involved in these activities during the year 2004, whereas the percentage correspondent to 2005 is 41%. These fi gures are consistent with the fi gures obtained by a study of the World Bank (2001) which observes 42% of the households having encountered bribery in the last 12 months. According to latest Eurobarometer 470, this fi gure was 18% of the people who went to see a doctor in 2017.
Since our model takes into account the possibility of selection in using health care, Table 1 from the Annex introduces the variables used in the analysis, whereas Table 2 presents some statistics of those variables. Means and standard deviations are provided to compare individuals that have used the health system in the last 5 years to those who did not use the system. Additionally, last column of Table 2 presents the results of tests on the equality of proportions comparing the samples of those who use the health care system with those individuals who do not use it. The p-value indicates whether the null hypothesis of the equality of the proportions should be rejected or not. We observe from Table 2, that 41% of the people who used the health care system paid atenție. Around 32% of people who used the health care system know a health care system worker among friends or acquaintances they can rely on, compared to 21% among those who did not use the health care system. Both users and non-users of the health care system perceive the system as corrupted. Around 57% think doctors are corrupt to a high degree or almost all are corrupt.
Non-users of the health care system score better in subjective health. Around 69% are satisfi ed with their health compared to 43% of users of the system. Users of the health care system are more afraid of diseases compared to non-users (38% compared to 23%, respectively). Among users, there are more retired people (30%) than in the non-users (16%).
The average of household income per capita seems higher for users of the health care system compared to non-users. The proportions of individuals in the highest income deciles among users of the health care system are greater than those in the nonusers' sample. Concerning the own perception that the income is not adequate to fulfi l the basic needs, is quite similar between the two groups around 38%. There are more women among the users than among the non-users (57% against 39%, respectively). The average age of the users is around 51 years, compared to 45 years non-users. Regarding the distribution concerning residential area, we observe that 26% of the health care users live in big cities, while only 16% of the nonusers live in big cities.

RESULTS OF THE MODEL
The estimation of Eq. 5 is presented in Table 3. We present the results of three alternative specifi cations of the probability of making an informal payment in the Romanian health care system. First, column [A] presents our preferred specifi cation: binary response probit maximum-likelihood estimation with sample selection correction. We estimate the coeffi cients and the average marginal eff ects of the control variables. 15 The second, column [B] presents the binary response maximum-likelihood estimation with sample selection correction. Finally, column [C] presents the Heckman two-step estimator, in this case a linear probability model.  b We use heckprob command in STATA software to produce the fi rst two columns, and heckman command in STATA software to produce the last two columns. We observe that the results are quite stable across the three specifi cations. Coeffi cients are very similar. 16 In the specifi cations involving maximum-likelihood estimation, [A] and [B], the sample selection correction proved to be eff ective. In both cases, the null hypothesis of no correlation of the error terms of the selection and outcome equation was rejected ( ). In the Heckman twostep specifi cation, column [C], the hypothesis of independence of the error terms of the selection and outcome equation can be tested directly by using the coeffi cient of the so-called Mills' ratio. In this case, it is signifi cantly diff erent from zero at any p-value higher than 0.079. Our preference for the column [A] specifi cation is because it takes into account the binary nature of the dependent variable (making an informal payment). The structural/outcome equation assumes a probit specifi cation. In consequence, the estimated predicted values of the dependent variable, probability of making an informal payment, is bound between zero and one. That is not necessarily the case in the specifi cation [C]. Nevertheless, the three specifi cations have similar sample average predictions, very close to the 41% sample value. The diff erences between specifi cation [A] and [B] are minimal, but specifi cation [A] minimizes both the Akaike information criterion (AIC) and the Bayesian information criterion (BIC).

Standard errors are provided in brackets. c It is a Wald test that all coeffi cients in
The perception of health care providers as corrupt and making informal payments are positively correlated. However, it is not clear whether people who paid bribes perceive more the system as corrupt or the perception of the system as corrupt drives individuals to make informal payments.
In regard to social capital, individuals who have relationships with health care providers are more likely to make informal payments. Thus, the hypothesis of informal payments as a form of social insurance against future health needs is validated.
We fi nd no evidence of income, whether subjectively, via the perception variable, or in any deciles infl uencing the probability of making an informal payment. The likelihood of making an informal payment is not correlated with income deciles. This fi nding is more evident in the analysis of Figure 1. No variability is observed regarding the probability of making an informal payment across income deciles. We cannot infer about the communist legacy. Should the communist past infl uence informal payments, one would expect that older cohorts of the population would be more likely to make an informal payment. According to our results, younger cohorts are more likely to make informal payments. Figure 2 allows us to observe that the probability of making an informal payment decreases with age. Individuals who have migration experience are more likely to make an informal payment. In 2005, migrants in the EU of Romanian origin were mostly circulatory migrants because the country was an accession country to the EU, and most migrant continue to be circulatory migrants as their permanent residence remains in Romania but they commute for work in the EU. It could be that because their main job is abroad they are not insured with the Romanian health care system, meaning they could only receive health care in case of emergency and thus they could be more prone to make an informal payment.
People with vocational studies and high-school education are less likely to make informal payments controlling for all the other variables.
We cannot infer any eff ects of working status or living in a big town on the likelihood of making an informal payment in the health care system. We could not relate these results with those of Lewis (2007) who fi nd that in rural areas, where the health providers are not many, the patients have higher chances to pay atenţie because of the monopolistic behaviour of providers. Table 4 reports the coeffi cients of the three diff erent estimations discussed earlier: the binary response probit maximum-likelihood estimation with sample selection correction [A], the binary response maximum-likelihood estimation with sample selection correction [B], and the Heckman two-step estimator [C]. The results in the three specifi cations are very similar since they are based on a probit estimation. According to the results reported in Table 4, the probability to use the health care system is lower for those individuals who are satisfi ed with their health status. Health care utilization is higher among those who are afraid of getting sick, compared to those who are not afraid of anything. Therefore, self-perception of health status and fears play an important role in health care utilization.
Perception of not enough income decreases signifi cantly the probability of making an informal payment. Controlling for other variables, individuals who perceive their income is not enough for the basic necessities are less likely to go to the doctor. This fi nding could confi rm that poor people might have access problems. Nevertheless, none of the coeffi cients of the variables indicating income deciles is statistically signifi cant. Women are more likely to use the health care system than men are. This result is consistent with previous fi ndings that showed that women are more likely to go to the doctor for non-threatening injuries or diseases. With regard to age, the so-called J curve 17 is confi rmed. Younger cohorts are less likely to use the health care system with respect to older cohorts.
We fi nd no evidence of working, occupation or migration experience aff ecting the likelihood of going to the doctor.
University graduates, and people that are more educated, are more likely to use health care. People living in a big city of more than 200000 inhabitants are more likely to go to the doctor compared to individuals living in villages or small towns. This is an indication that our proxy for health care supply is performing well, as small towns are likely to lack proper and advanced hospitals.

CONCLUSIONS
Suggesting policies to improve the current situation of the Romanian Health Care System (under-funding), must take into account that corruption is part of the public services in Romania. As our analysis has shown, making informal payments is seen as a way to ensure the access to public health care.
The payment of atenţii is widespread, to the extent that more than 40% of the population older than 18 years old has been involved in informal payments to receive medical attention. Additionally, based on fi gures from Lewis (2002) and the World Health Organization (WHO) data set, Romanians spend around 450 mills US PPP$ in atenţii. This creates inequities and ineffi ciencies in the distribution of health care.
The phenomenon of informal payments in health care systems can be studied from many angles. In this study, we have discussed three perspectives. First, we discussed an economic approach emphasising the role of incentives and market failures (Ensor, 2004); as well as the role of government failures (Balanova and McKee, 2002). Second, we mentioned a sociological approach emphasising the roles of institutions and the culture of gratitude payments placing the phenomenon into a context of informal economies (Kornai, 2000). The role of the communist past is the main reason generating a second parallel economy in which fostering relationships with providers of goods and services is the key to access scarce resources (Ledeneva, 1998). Third, a political science approach emphasizes the lack of accountability and control (Lewis, 2002(Lewis, , 2007 as well as the lack of government legitimacy (Rose-Ackerman, 2001).
As we have shown, the Romanian health care system's evolution generates incentives and creates room for informal payments. Informal out-of-pocket payments maintain the Romanian Health Care System functional in the context of under-funded budgets and scarce resources.
A binary response maximum-likelihood probit model with sample selection correction was estimated. The outcome regression calculated the likelihood of paying an atenţie based on demographic and labour related variables (including income) and the perception of corruption and social capital.
The model illustrates that the perception of health care providers as corrupt and making informal payments are positively correlated. However, it remains unclear if people who made informal payments perceive more the system as corrupt or the perception of the system as corrupt drives individuals to make informal payments. This could be a source of potential endogeneity. Regarding social capital, individuals who have relationships with health care providers are more likely to make informal payments. Thus, our hypothesis of informal payments as a form of social insurance against future health needs is validated. Moreover, we cannot infer about the communist legacy discussed in the context of the sociological perspective. Should the communist past infl uence informal payments, one would expect that older cohorts of the population would be more likely to make an informal payment. According to our results, younger cohorts are more likely to make informal payments. Figure 2 allows us to observe that the probability of making an informal payment decreases with age. The analysis also showed that controlling for income, and the perception of having inadequate income to cover basic needs does not play a role in making an informal payment.
The selection equation assessed the probability of going to the doctor based on the self-perception of health and personal fears as exclusion restrictions. It found that people who evaluate positively their state of health are less likely to use the system. The results indicated that people who are afraid of getting sick, women are more likely to go to the doctor. With regards to age, the so-called J curve 18 is confi rmed. Younger cohorts are less likely to use the health care system with respect to older cohorts. Controlling for income, people who perceive their income is not enough to satisfy their basic needs are less likely to go to the doctor, indicating that health care needs are not met for the poor. This fi nding could confi rm that poor people might have access problems. Nevertheless, none of the coeffi cients of the variables indicating income deciles is statistically signifi cant. We fi nd no evidence of working, occupation or migration experience aff ecting the likelihood of going to the doctor.
University graduates and people that are more educated are more likely to use health care. People living in a big city of more than 200,000 inhabitants are more likely to go to the doctor compared to individuals living in villages or small towns. This is an indication that our proxy for health care supply is performing well, as small town are likely to lack proper and advanced hospitals.
Our analysis reveals the fact that generating trust among Romanians as well as increasing health funding could decrease informal payments. Accountability of the medical service could decrease the need for atenţie accompanied by good wages and premiums for performance complemented by a well funding system of hospitals and ambulatory care. However, this investment has to be complemented by mass media campaigns providing information regarding the rights of the patients.
As we could see from the empirical analysis this could have an impact in reducing the payment of atenţie, and reducing the need of having contacts in the system. The increase of professionals, especially in rural areas, could decrease the needs for paying an atenţie to get a service in those circumstances. Reforms are needed to improve the health indicators of the country and help to generate solid foundations to build sustainable economic growth with a healthier and productive population. 2 Previous studies have distinguished conceptually between gratitude payments, which are voluntary, and envelope payments that are required by the staff . Moreover, it could be argued that gratitude payments are a form of insurance against future care needs (Lewis, 2007). In this paper that distinction, whether voluntary or required, is not made as it is diffi cult to assess the nature of the transaction. Therefore, terms like informal payments, gratitude payments will be used interchangeably in this paper.
3 Data is available from https://www.gesis.org/index.php?id=11248. 4 Formal payments are legal fees paid for the health care costs. 5 In microeconomics, economics of health, health care appears as a derived demand from the demand of health and is an input in the production function of health. The standard model to derive health care demand assumes that the individual can choose between consuming a generic good and increasing his/ her health status in order to maximize utility. Grossman (1999) suggests a model in which the basic model takes a step further and medical care is seen as an investment good. The basic assumption is that individuals invest in human capital to improve outcomes in both market and non-market sectors and maximize their outcome in healthy days. Most of the microeconomics circumvents the problem of investigating directly health care by making assumptions about the health preferences and assuming that health is a normal good. 6 The central planning bureau set the prices for medical services; therefore, supply and demand had little impact on doctors' wages that were in a position to increase their standard of living through gratitude payments (Kornai, 2000). 7 Even though the Law was passed in 1997, the enforcement of the legislation was done in 1999 when the local branches of the National House for Health Insurance became autonomous public institutions.