WPS6101 Policy Research Working Paper 6101 Overdraft Facility Policy and Firm Performance An Empirical Analysis in Eastern European Union Industrial Firms Leopoldo Laborda Castillo Jose Luis Guasch The World Bank Financial and Private Sector Development Finance and Private Sector Department June 2012 Policy Research Working Paper 6101 Abstract This article evaluates the effect of the overdraft facility for individual sectors), production workers trained, (or line of credit) policy by comparing a large sample of expenditures on research and development, and export overdraft facilitated firms and matched non-overdraft activity. In order to avoid the selectivity problem, facilitated firms from Eastern Europe at the sector level. propensity score matching methodologies are adopted. The sample firms are compared with respect to rates of The results suggest that a certain level of overdraft different performance indicators including: technical facility provided to firms would be needed to stimulate efficiency (a Data Envelopment Analysis approach investment in research and development, which will is applied to estimate the technical efficiency level eventually result in increased growth in productivity. This paper is a product of the Finance and Private Sector Department,Financial and Private Sector Development. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The author may be contacted at jguasch@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Overdraft Facility Policy and Firm Performance: An Empirical Analysis in Eastern European Union Industrial firms1 Leopoldo Laborda Castillo* World Bank (USA) and University of Alcalá (Spain) Jose Luis Guasch** World Bank (USA) and University of California, San Diego (USA) Keywords Overdraft Facility Policy, Propensity Score Matching, Technical Efficiency, Data Envelopment Analysis (DEA), Bootstrapping. JEL Classifications C40, H43, H8, L25, L26, L53 * STC at The World Bank and Associate Professor at the University of Alcalá: c/ Trinidad nº 1, Colegio de Trinitarios, Alcalá de Henares, 28801, Madrid (Spain). Phone: +34 91 8820399. e-mail: llabordacastillo@gmail.com ** Former Senior Adviser at the World Bank and Professor of Economics at University of California, San Diego (USA) e-mail: jguasch@worldbank.org 1 This paper is forthcoming in the Journal of Business Economics and Management. The authors are thankful to Esperanza Lasagabaster and two anonymous referees for very useful comments. 1 Introduction The overdraft facility2 policy from private financial institutions provides financial support to increase firms‘ accessibility to private financing sources. However, according the IMF (2005) the overdraft facility policy has often been criticized for its negative effects by impairing the development of an innovative private financial sector and for making firm highly dependent on government support policy measures. Even though there have been some qualitative remarks on the effectiveness of the policy, an evaluation of the overdraft facility policy has not been conducted systematically in terms of methodology and data, especially in Eastern European Union Industrial firms. This study aims to fill the gap in the existing literatures. In this work, we evaluate the effect of the overdraft facility policy in terms of human capital (production workers trained), technological capital (expenditures of R&D), competitiveness (sales export level) and the firm‘s technical efficiency in the Eastern European Union Industrial firms, by comparing overdraft facilitated firms and non- overdraft facilitated firms. The amount of funds allocated to the overdraft facility policy is huge and the number of targeted firms is large. Therefore, we need to investigate its effectiveness and to provide background information for further evolution of the policy. This work utilizes an original dataset that covers manufacturing firms in 11 Eastern European countries (Bulgaria, Czech Republic, Estonia, Hungary, Latvia, Lithuania, Moldova, Poland, Romania, Slovak Republic, and Slovenia). Moreover, we analyze the effectiveness of the overdraft facility in four industrial sectors (Beverages, Food, Garments, and Metals and machinery). Methodologically, reliable policy evaluation should solve the ‗selectivity‘ problem. According to Muthen and Jöreskog (1983), selectivity problems can occur whenever one tries to estimate population parameters from a nonrandom sample. The sample may be nonrandom because only firms with certain characteristics are selected into the sample (sample selection), or because firms participate voluntarily in the sample (self-selection). Selective samples can also occur because firms fall out of the sample for various reasons, despite an initial random sample (attrition). According to Jaffe (2002), if we cannot control the selectivity problem, we might over- or under- estimate the true effects of the policy. In order to deal with the selectivity issue, we adopt propensity score matching estimators, which have recently been applied to firm-level studies (Yasar and Rejesus 2005; Loof and Heshmati 2005; Oh et al. 2009). Propensity score matching was chosen because: (i) it reduces (although does not eliminate) selection biases, (ii) it reduces the limitation from matching on many observable variables for finite data, and (iii) it is best suited to the structure of the available data. For example, because baseline data was not collected at the outset of the program and detailed secondary source data were not available to construct an appropriate baseline, evaluation techniques such as difference-in-difference, which would have removed non-observable biases between the treatment and control groups, could not be utilized. The difference-in-difference compares the difference between the treatment and control groups before (first difference) and 2 An overdraft facility is defined as a flexible account that allows firms to draw upon in the event their account balance becomes negative. 2 after the program (second difference)3. See Ravallion (2009) for a discussion of commonly used evaluation techniques4. This study is organized as follows. The following section introduces the economic environment of the Eastern European Union Industrial firms and the existing overdraft facility policy schemes. Section 3 briefly explains the methodologies of Data Envelopment Analysis (DEA) Frontiers (with Bootstrap procedure) and propensity score matching. Section 4 describes the data and it presents the key descriptive statistics. The empirical results are discussed in Sect. 5. The final section summarizes and concludes this study. 2 Background 2.1 The Eastern European Union Companies The EU enlargement to 27 countries in 2004 and 2007 constitutes a historical benchmark in the forming of the European space. In contrast to previous enlargements, the entering of Eastern European countries has peculiar characteristics due to the large number of nations entering the EU and due to the heterogeneity in its parameters and levels of development (Hay, 2003: 13). Although the Eastern European countries have the common trait of their recent history linked to the Soviet Union and the fall of the Berlin Wall, their entry into the European club required the compliance of three conditions: democratic institutions, a market economy capable of insertion in a competitive world and the capacity to take on the Union‘s goals. These conditions were an important impulse to leave that common trait behind and, to a certain extent, have allowed differentiating the strengths and weaknesses of each country. On the other hand, one must emphasize that this important enlargement in terms of numbers did not maintain a proportional relation with its economic relevance. Comparing the UE15 (2004) and taking into account data of that year, the contribution in population of the Eastern European nations is of about 100 million inhabitants, which translated into a 28% increase in the EU population whereas in added terms, the increase in GDP was of nearly 7%, emphasizing Poland‘s contribution of 42.7% of that value (Alcalá, 2004: 144). In terms of rent per capita, the set of the countries of Eastern Europe was placed in 2004 in 4,380 Euros (current exchange rates) and 9,100 (PPP) Euros, respectively, which supposed 20% and 40% of the per capita income in the Europe of the fifteen. By countries the issue was somewhat different: in the case of Rumania the number in PPP was of 24% whereas in Hungary the per capita income was almost a 70% of the EU15. All this put into evidence the retard of these countries, but mainly the effort (with a very high degree of heterogeneity) needed to obtain the much desired real convergence of the EU27. Another trait to consider is the characteristic of its productive structure: in the countries of the enlargement, more than 20% of the population was occupied in the agricultural sector whereas in the EU15 that percentile was of 4.8%. Among the countries one can highlight are Romania (40%), Bulgaria (28%) and Poland (26%). As far as foreign trade is concerned, the entailment of the 12 countries with the EU15 (exports and imports) approximately reached 60% of its trade balance whereas for the EU15 that value remains below 2.5% (Alcalá, 2004). 3 Randomization, was not possible either since the treatment group was not randomly selected. 4 Ravallion, M. (1999: ―The Mystery of the Vanishing Benefits: Ms. Speedy Analyst's Introduction to Evaluation‖. Policy Research Working Paper 2153, World Bank, Development Economics Research Group, Washington, D.C. 3 In any case, the enlargement is a reality and as all integration processes, it involves benefits and costs to bear in mind. Among the benefits one can mention: the inclusion of new rules in the operation of the society and the economy that allow greater opportunities for initiatives and creativity, as well as the suppression of tariff barriers and the exploitation of scale economies that allow a suitable resource allocation. But also the risks related to the appearance of structural deficits are present as a result of the commercial deficit, budgetary deficit of both national governments and the European Union, and deficit in the social cohesion process of the EU27 as a result of a greater number of population and regions with incomes far below the average. At the micro level, one of the most worrisome problems—and perhaps the one that is of greater interest for our study—is the way in which the productive companies of these countries will respond (the majority of them are of recent creation and are modestly consolidated in terms of management background). In this sense, it is interesting to analyze the behavior of companies from the point of their technical efficiency, as well as of their contribution to the value chain, which will allow shedding light on some conclusions on policy priorities, not only among the countries but among the productive sectors5. The primary aim of this article is to examine the technical efficiency and its determinants in companies from the countries included in the enlargement of the EU and to know the main conditioning factors of the growth of productivity6. In this context, a variety of features distinguish the business environment in sectors from those typically observed in the OECD. We will begin by mentioning the most remarkable and generally acknowledged among them, as identified by Tybout (1992): a) Market size (e.g. population and GDP). Although some economies are quite large, most are not. Hence, with the exception of countries such as Poland or Romania, the size of the domestic market for manufactured products is relatively limited. This means that the strategy of companies must be oriented towards the external sector and thus, must necessarily bear in mind the macroeconomic aspects that are translated through channels such as direct foreign investment, exchange rates and immigration. b) Access to manufactured inputs (e.g. merchandise imports and exports). The set of choices of domestically produced intermediate inputs and capital equipment is also often limited. c) Human capital (e.g. school enrollment, primary). Low rates of secondary education (less than 90%) and an insufficiency of technicians and scientists in countries like Hungary of Lithuania also affect the mix of goods manufactured and the factor proportions used to produce them. 5 A study on the experience of the start of the internal European market in 1993 (that had as one of its main objectives the increase in company efficiency) shows that during the 1987-1996 period, there has not been an increase in efficiency in the productive sectors of the Union. This paper argues that a possible cause of this stagnation is the possibility that the efficiency has a cyclic character and, therefore, the recession suffered in the beginning of the 90‘s could explain this situation (Esteban, Gallizo and Hernández, 2002). 6 In this research paper, the data of Malta and Cyprus have not been taken into consideration since they are considered as Mediterranean countries. On the contrary, data from Moldova (not in the EU) with the sole objective of homogenizing the sample of the block of countries considered as Eastern European. 4 d) Infrastructure (e.g. road density). Roads, ports, airports, communication facilities, power, and safe water access also tend to be relatively limited in all countries, although they especially affect countries like Bulgaria or Moldova. e) Financial markets (e.g. domestic credit to private sector). Credit markets are also relatively thin in countries like Romania or Moldova. f) Volatility (e.g. inflation, consumer prices). Macroeconomic and relative price volatility is typically more extreme in countries like Romania or Moldova than in other economies like the Czech Republic or Lithuania7. g) Governance (e.g. procedures to enforce a contract). Red tape is also relatively high in countries like Poland or Romania, for this reason the protection of contract enforcement can be problematic8. The information on the economic characteristics of the countries analyzed was taken from different World Bank publications. A synthesis is shown in Table 1. 7 ―All developing regions do worse than the industrialized countries‖, Hausmann and Gavin (1996). 8 ―Anti-trust policy is also often weak, as are environmental standards‖ Brunetti (1997). 5 Table 1 Socio-Economic and Development Indicators by country (2005 year). Country/ Total GDP per Industry, Merchandis Merchandise School Road Domestic Inflation, Procedures industry population1 capita, value added3 e exports4 imports4 enrollment, density6 credit private consumer to enforce a PPP2 primary5 sector7 prices8 contract1 Bulgaria 7781000 8737.161 4022807808 9932382032 14453494993 94.98052791 40.48639206 36.31128868 6.346133143 34 Czech Republic 10206923 19094.62 22018136064 68986459239 69967456060 .. 165.3908879 32.61472877 2.833643175 22 Estonia 1348999 15107.31 1911966336 5932476581 8335935795 93.90955231 134.0858693 40.01893348 3.048071707 25 Hungary 10107095 16306.48 15011529728 55566718661 60538321226 89.09593776 .. 45.82225836 6.780023712 21 Latvia 2312791 11885.7 2287386624 4009170454 7095505106 .. 111.6262642 50.77709798 6.191878487 24 Lithuania 3435585 12967.48 4650113536 9301566446 12386991757 88.62493488 126.5650925 28.80164646 1.195219124 24 Moldova 3925170 2012.422 290329920 986255000 1773742000 92.37393608 38.73745056 21.23095682 12.5283047 37 Poland 38182200 13091.86 53077282816 75046614430 89696102001 97.51612288 .. 28.14845 3.576547231 41 Romania 21684884 8977.659 15057262592 23553215092 32690617860 90.9535242 86.44217391 15.6740974 11.87686787 43 Slovak 5382449 14989.53 8776060928 27744734233 29861579419 .. 89.3970894 30.54339454 7.548500882 27 Republic Slovenia 1997000 22132.83 7092050944 16361365166 17758527740 95.54712377 190.9126156 48.77266733 3.589026796 25 1 Number. 2 Constant 2005 international $. 3 Constant 2000 US$. 4 Current US$. 5 % net. 6 Km of road per 100 sq. km of land area. 7 % of GDP. 8 Annual %. Source: DDP Quick Query database of WDI & GDF, World Bank. 6 As already mentioned in the introduction, the socio-economic and development indicators for the analyzed countries reflect economies that have gone from a planned system to a mixed system, which has culminated with its incorporation into the European Union (EU). These magnitudes show the necessity of convergence with respect to the set of countries that make up the present EU, both from the nominal and the real points of view: 1) Firstly, in aspects related to infrastructure endowment (highway density), bureaucracy (number of procedures to sign a contract), education (incorporation of the population to primary education), with respect to the rest of the countries that form the EU. 2) In second place, in aspects related to the efficiency of its productive structure, in order to promote economic growth and development magnitudes (growth of the per capita gross domestic product). This second aspect is precisely the object of analysis of this paper, although the interaction between both factors is an aspect that is widely discussed in the literature on economic growth and development. In fact, another line of work with the countries of the East is related to the analysis of the convergence among these countries and the rest of the European Union (Esteban, Gallizo and Hernández, 2002). 2.2 The Overdraft Facility Policy According Lerner (2002), firms, especially young and high-tech firms, face great difficulties in accessing the loan market due to asymmetric information which arises from the lack of public information on standardized financial statements and further growth potential. In this context Stiglitz and Weiss (1981) showed that in equilibrium a loan market might be characterized by credit rationing. In a perfect economic system, if prices do their duty, credit rationing should not exist. However, there exists residual imperfect information even after the evaluation of loan applications. Due to imperfect information and followed adverse selection and moral hazard, market failure occurs in the firms loan market and banks are more likely to demand collateral in loan contracts (Cowling 1998), and it is often difficult for firms to borrow funds, even at higher interest rates9. There exist some recently published studies aimed at analyzing and evaluating overdraft facility policy schemes in various countries. In the case of the UK, Cowling and Mitchell (1997), the overdraft facility policy scheme has successfully addressed a real capital constraint for the majority of firms which applied to the scheme10. In the case of Malaysia, Boocock and Shariff (2005) evaluated the overdraft facility policy scheme in terms of finance additionality, i.e. an increase in accessibility to loans, and economic additionality, i.e. utilizing the funds to benefit their own firms and to generate positive spillovers. These authors concluded that Malaysian overdraft facility policy schemes in general failed to satisfy financial additionality, though it showed some positive outcomes in relation to economic additionality. Kang (2005) and IMF (2005) shared the views that the nonselective financial support of firms was one of the key sources of the sluggish firm restructuring process after the financial crisis. 9 Adverse selection reflects a situation in which lenders do not have sufficient information to ascertain the risk level of a loan applicant. With moral hazard, borrowers react to increases in interest rates by switching to yet riskier projects. 10 In addition, the paper also showed that firms which received loans to finance their working capital had a higher default rate compared to the firms which invested in buildings. 7 They argued that a too generous and non-selective guarantee provision worsened the firm market environment in two ways: the first was the effect of crowding out the private financial sector. The second and long-term negative effect was to make firms become more dependent on public support11. In the case of Korea, Lee (2006) criticized that firms with low creditability will gradually receive the overdraft facility by excessive supply. It will eventually increase the credit default ratio and hence financial difficulties for government in the end. Using a data from a cross-section of Bulgarian firms Gatti and Love (2008) estimate the impact of access to credit, as proxied by indicators of whether firms have access to a credit line or overdraft facility, on productivity. To overcome potential omitted variable bias of Ordinary Least Squares (OLS) estimates. These authors find credit to be positively and strongly associated with TFP. Augier, Dovis and Gasiorek (2010) focuse on the role of the business environment in understanding differences in the performance of Moroccan firms. The evidence on the relationship between credit and productivity is strongly indicative of credit resources misallocation in Morocco. Kasseeah (2011) studies the link between firm performance and firm characteristics of small and medium-sized enterprises (SMEs) in Mauritius. This author observes that the growth of SMEs in Mauritius depends mainly on access to finance and firm size. Access to finance is captured through access to overdraft facilities, line of credit and self reported measures of access to finance. Butler and Cornaggia (2011) study the relation between access to finance and productivity. These authors exploit an exogenous shift in demand for a product to expose how producers adapt their productivity in the presence of varying levels of access to finance. These authors find that production increases the most over the sample period in areas with relatively strong access to finance. Using a panel of Chinese manufacturing firms over the period 2001-2007, Chen and Guariglia (2011) find that, especially for illiquid firms, productivity is strongly constrained by the availability of internal finance. Furthermore, these authors find higher sensitivities of productivity to cash flow for private exporters, but lower sensitivities for foreign exporters. Finally, Clarkea, Cullb and Kisunkoc (2012) study how country and firm characteristics affected firms‘ financial constraints and their likelihood of survival during the early phase of the recent global financial crisis in Eastern Europe and Central Asia (ECA). These authors find that financial constraints during the crisis were less severe in countries with well-established foreign banks, and that changes in the severity of financial constraints were more pronounced for large firms than others during the crisis. Controlling for other relevant characteristics, firms were more likely to survive the crisis if they had access to external credit. 11 The survival of uncompetitive firms with the help of government support might result in a decrease in market share and the profits of competitive firms. Thus, uncompetitive firms might replace competitive ones with overdraft facility policy. 8 3 Methodology Not much quantitative research has been conducted referring to the effectiveness of the overdraft facility policy, mainly due to the unavailability of consistent data and an appropriate methodological framework. Although there is no consensus among researchers regarding the way to establish the process to evaluate the influence of environmental variables on service efficiency levels, in this paper we have attempted to detect the repercussion of certain exogenous factors on the said efficiency levels by using a two-stage process made up of the following steps: 1) Obtaining the Charnes, Cooper and Rhodes (CCR) efficiency index. In order to calculate efficiency, the behavior of each unit observed is optimized, thus determining the efficient production frontier by means of linear segments based on the Decision Making Units (DMUs) that operate with the best practices. This corresponds to the set of units considered efficient in Pareto‘s terms. Therefore, the only requirement established is that each DMU should belong to the frontier envelopment (Cooper et al. 2001: 3). 2) Using re-sampling methods and bootstrapping techniques. Though there is currently information available on statistical inference with a reasonable level of certainty, as a result of advancement in the development of bootstrap techniques (Simar and Wilson 2000), we have opted for the traditional application of determinist mathematical models, and as a complement to them, we have used the re-sampling methods and bootstrapping techniques, in accordance with the contribution by Simar and Wilson (1998)12. The rationale behind bootstrapping is to simulate a true sampling distribution by mimicking the data-generating process, and the results throughout this paper were obtained from 2,000 bootstrap iterations. 3) Assessment of overdraft facility policy impact using Propensity Score Matching (PSM). The PSM was selected as the most suitable methodology to conduct the overdraft facility policy impact evaluation. The objective of the matching is to construct treatment and control groups that are as similar as possible, so that the control group would resemble what would have happened to the beneficiaries in the absence of overdraft facility policy (Rosenbaum and Rubin, 1983). The PSM estimation strategy requires multiple steps: (i) estimation of the propensity score, in terms of probability to have overdraft facility given observable variables, (ii) the review of overlap and common support conditions (i.e. which firms are matched) and (iii) assessing matching quality. 3.1 Data Envelopment Analysis (DEA) Frontiers and Bootstrap Procedure According to Leibenstein (1966), a company can be categorized as technically efficient if it is able to produce maximum output given available resources. It has been acknowledged in the literature that a gap normally exists between a firm‘s actual and potential levels of economic performance13. Efficiency will be defined herein as the activity which produces maximum 12 Simar and Wilson noted that the DEA efficiency estimates are biased and serially correlated, which invalidates conventional inferences from the two-stage approaches. In this context, the authors proposed a procedure, based on a double bootstrap, that enables consistent inference within models explaining efficiency scores simultaneously producing standard errors and confidence intervals for these efficiency scores. 13 To better illustrate the relation existent between added growth of the productivity an the evidence at the company and sector levels, see Foster, Haltiwagner and Krizan (1998). 9 production given a certain set of resources, or in other words, the action which consumes the least possible volume of resources in order to achieve a certain volume of production. According Farrell‘s contribution (1957), this paper focuses on technical efficiency, which measures total production volume produced with allocated productive resources. In this context, Data envelopment analysis (DEA) will be used to estimate the above components 14. The essence of this technique lies in the definition of an efficient frontier as a point of reference for evaluating the variations observed in the performance of different production units (countries in this case)15. The DEA method does not create assumptions about functional forms, being a non-parametric performance evaluation approach; however, it does allow different inputs and outputs to be incorporated, something that is essential to the aims of the present study and the principal justification for its employment. One of the major advantages of dea is that it can be used to work with multiple inputs and outputs that have different units, while it does not create a need to consider full employment of production factors or require explicit functional forms. As for the drawbacks, the method is sensitive to measurement errors (the reference points are highly productive) and does not permit easy application of statistical tests, and relative rather than absolute inefficiencies are identified. To minime these limitations we use Boostrap Techniques to analyze the sensitivity of efficiency scores relative to the sampling variations of the estimated frontier. The original DEA estimator proposed by Charnes, Cooper and Rhodes (1978), referred to as the CCR formulation, allows the efficiency of any Decision Making Unit (DMU) to be measured from the maximization of a ratio of weighted outputs with respect to weighted inputs, subject to the restriction that similar ratios for the rest of the DMUs are less than or equal to the unit. More precisely, the linear calculation program would be: min ï?± 0 Subject to: n y j 1 rj ï?¬ j  y ro ; r  1,..., s n ï?± o xio   xij ï?¬  0; i  1,..., m j 1 ï?¬j  0 14 Coelli, T. J., Rao, P., O‗Donnell, C. J., and Battese, G. E. (2005) give a detailed presentation of this technique and an excellent review of the concepts of productivity and efficiency. 15 In other words, the starting assumption when applying this technique in the case before us is that all the countries in the sample ought to be able to function at an optimal level of efficiency, determined by the efficient countries included therein. In the specialist literature, these efficient countries are called ―peer countries‖ and are the ones that determine the efficiency frontier, so that the distance to the efficient frontier represents the measure of efficiency or the lack thereof. 10 ï?¬ This program calculates a virtual unit as a linear combination, where j represents all the weighted units evaluated, and obtains an identical or greater number of outputs with a smaller number of inputs than the unit being evaluated. If it is not possible to find a virtual unit that obtains the same outputs with a smaller number of inputs, the unit is efficient and is situated on the frontier. ï?± represents the factor that weighs all the inputs, and takes values between 0 and 1. Efficient DMUs will have ï?±  1 , which means that it is not possible to reduce the number of inputs used to produce an identical level of outputs. The measurement of technical efficiency calculated by the Banker, Charnes and Cooper (1984) formulation makes it possible to find out whether there is proper use of resources in relation with the production of goods or services of the DMU analyzed. As for scale efficiency, it is equal to the quotient of BCC efficiency and CCR efficiency, and provides a measurement of the distance from the analyzed DMU to a virtual DMU that operates with the most productive scale size (MPSS). For this purpose, these authors propose the existence of a single difference between the envelopment of the BCC and the CCR formulations: the inclusion of the restriction of convexity  ï?¬ jk  1 n j 1 (relating to the DMU k): . Finally, the procedure applied in this study follows Simar and Wilson (2007). It consists of the following steps: (a) standard DEA efficiency point estimates are calculated, (b) Truncated maximum likelihood estimation is used to regress the efficiency scores against a set of explanatory variables, (c) these estimates are then integrated into a bootstrap procedure that is similar to the smoothed bootstrap procedure of Simar and Wilson (2000). This bootstrap procedure allows correcting for bias, and (d) the bias corrected scores produced by the preceding bootstrap are used in a parametric bootstrap on the truncated maximum likelihood, thus creating standard errors for the parameters of the regression. Confidence intervals are then constructed for the regression parameters as well as for the efficiency scores. 3.2 Propensity Score Matching (PSM) Since Rubin (1974), the effect of a program is defined as a created value added by participating in a program. In this regard, the effect of a program can be defined as ‗what would have happened to those who, in fact, did receive treatment, if they had not received treatment?‘. Hence, mere comparison between supported firm groups and non-supported firm groups cannot identify the exact additional effect of the support program, since their characteristics before participation in the supporting program were different already, which is also generally referred as the selection bias. The concept of effect of a program defined by the additional value added should, thus, be based on the appropriate construction of a counterfactual. Modern evaluation methods are focused on estimating this counterfactual (Blundell and Costa Dias, 2000). Given the counterfactual problems, the most appropriate measure of the effectiveness of government support might be a comparison of the performances of two firms with the same characteristics, assuming that one received support and the other did not. However, it is hard to find appropriate comparison groups, which can represent the non-supported firms in evaluating the program. In this study we apply the PSM methodology, which allows us to construct a comparison group by matching twin firms based on the propensity score in the population of unsupported firm 11 groups. With this approach, we expect to solve the selection bias problem and to compare the factual and the counterfactual to estimate the sole outcome from the program. PSM methodology was first introduced by Rosenbaum and Rubin (1983). The concept of PSM requires fulfillment of the conditional independence assumption (CIA). It means that conditioned on the observable characteristics (X variables) of possible participants, the decision for participation in the program should be independent of the outcome measures. CIA in this respect can be written as following: Y0 , Y1  ï?ž T X ï?›1ï?? where ï?ž denotes independence, Y1 means the outcome of the supported firm and Y0 denotes the outcome of the unsupported firm. T is an indicator variable denoting participation in the program. Another required assumption is that the probability to participate in the program for program group and comparison group should lie in the same domain, which is called the common support assumption. If these assumptions are satisfied and when there exist a sufficient number of observable variables related to the characteristics of participants in a program, it is theoretically possible to obtain an unbiased estimation of the effect of a program. Propensity score indicates a conditional probability of applicants to participate in a program when observable characteristics of applicants are given. In other words, Propensity score  P X   PrT  1 X  ï?›2ï?? Rosenbaum and Rubin (1983) proved that under CIA and the common support assumption, with propensity score defined as ï?›2ï?? , all biases due to observable variables can be removed by conditioning solely on the propensity score: Y0 , Y1  ï?ž T P X  ï?›3ï?? Based on ï?›3ï?? , for a population of units denoted by i, we can define the policy effect, which is defined as the difference between real and counterfactual outcomes, as the average effect of treatment on the treated (ATT) as follows: ATT  Eï?»Y 1i  Y0i Ti  1ï?½  Eï?»E Y1i  Y0i Ti  1, P X i ï?½  CIA EP  X i  ï?»E Y1i Ti  1, P X i   E Y0i  Y0i Ti  0, P X i Ti ï?½ ï?›4ï?? where the outer expectation is taken over the distribution of P X i  in the population of supported firms, Ti  1 . ATT is the difference in average outcomes of the supported and unsupported firms, where the unsupported firm group is formed matching units based on the propensity score. However, estimation of the propensity score only, is not enough to estimate the ATT of interest using Eq. ï?›4ï?? . This is because the probability of observing two units with exactly the same value 12 of the propensity score is in principle zero, since P X  is a continuous variable. Various methods have been proposed in the literature to overcome this problem, and three of the most widely used are nearest neighbor matching, radius matching, and kernel matching. Based on the literatures on the selection and implementation of matching estimators, we decided to analyze the effect of credit guarantee with kernel matching which showed good precision when the unsupported-to-supported ratio is high (Frolich 2004) and can avoid the criticism expressed by Abadie and Imbens (2006), which pointed out the problem of nonsmoothness in calculating standard errors with bootstrapping for the case of nearest neighbor matching or radius matching. 4 Data and Variables 4.1 The Data The statistical source used for this analysis is the World Bank‘s Enterprise Surveys (ES). The ES collect data from key manufacturing and service sectors in every region of the world. The Surveys use standardized survey instruments and a uniform sampling methodology to minimize measurement error and to yield data that are comparable across the world‘s economies. To generate internationally comparable data, the questions in the Core questionnaire are asked in all countries and for all industries where the survey is implemented. In addition to this Core instrument, the Manufacturing Module and Services Module questions are asked to establishments in the manufacturing and services sectors, respectively. The Core instrument is comprised of eleven sections: the first eight sections contain qualitative questions, asking for the manager‘s opinion on the business environment and for his motivation for business decisions16. The last three sections of the questionnaire deal with facts and figures specific to the transactions businesses make in order to operate17. The Manufacturing Module contains questions about capacity (use of production capacity and hours of operation). The sampling methodology of the World Bank‘s Enterprise Survey generates sample sizes appropriate for achieving two main objectives: (1) a sample representative of the whole economy that substantiates assertions about the whole economy, not only about the manufacturing sector. In addition to selected manufacturing industries, the overall sample should include services industries and other relevant sectors of the economy, and (2) a sample that is large enough in size for selected industries to conduct statistically robust analyses with levels of precision at a minimum 7.5% for 90% confidence intervals about18: (a) estimates of population proportions (percentages), at the industry level, and (b) estimates of the average mean of log of sales at the industry level. 16 These sections deals with the characteristics of the business and the investment climate in which it operates including: Control Information, General information (ownership, start-up), Infrastructure and Services (power, water, transport, and communication technologies), Sales and Supplies (imports, exports, supply and demand conditions), Degree of Competition (price and supply changes, competitors), Land (land ownership, land access issues), Crime (extent and losses due to crime), Business-Government Relations (quality of public services, consistency of policy, regulatory compliance costs), and Investment Climate Constraints (evaluation of general obstacles). 17 More specifically, these sections contain questions on production costs, investment flows, balance sheet information and workforce statistics. These sections include: Finance (sources of finance, terms of finance, financial services), Labor (worker skills training, skill availability, employment, education levels of workers) and Productivity (Numbers and figures needed to estimate productivity). 18 A 7.5% precision of an estimate in a 90% confidence interval means that we can guarantee that the population parameter is within the 7.5% range of the observed sample parameter, except in 10% of the cases. 13 Table 2 Statistical distribution: firms according to country and sector (year 2005). Country/industry Beverages Food Garments Metals and machinery Bulgaria 10 2 5 9 Czech Republic 7 3 3 29 Estonia 4 1 5 8 Hungary 7 28 41 153 Latvia 2 4 5 Lithuania 4 3 8 Moldova 40 4 14 20 Poland 15 46 97 178 Romania 56 36 75 86 Slovak Republic 1 1 12 Slovenia 1 1 2 23 Total 146 122 250 531 Source: Authors‘ calculation from the Investment Climate Survey Databank, World Bank. The previous table 2 comprises the sectors analyzed in this research paper respectively and includes secondary activity (industry) sectors. 4.2 Variable Selection In order to estimate technical efficiency with DEA, there exists considerable disagreement in empirical literature on the definition of outputs (in the context of multiple-outputs firms) of an industrial institution. Nevertheless, according to Coelli et al. (2005), a commonly-used classification of inputs involves five categories: capital (K), labour (L), energy (E), material inputs (M), and purchase services (S). The use of data according to these categories in productivity measurement is sometimes referred to as KLEMS approach19. To ensure the validity of the DEA model specification, an isotonicity test (according to Golany and Roll, 1989) was conducted. This isotonicity test involves the calculation of all inter- correlations between inputs and outputs for identifying whether increasing amounts of inputs lead to greater outputs. Finally, according to Golany and Roll (1989), the number of DMUs should be at least twice of the total number of input and output factors considered when applying the DEA model. In this study the number of DMUs was 146 (for Beverages sector), 531 (for Metals and Machinery sector), 250 (Garments sector) and 122 (for Food sector), at least ten times of the selected four factors. According to these facts, in this study, the proposed DEA model has good construct validity. On the other hand, Smith (2000) argues that, in order to obtain correct and robust estimates of program effects using PSM, one need to use a broad set of observed firm characteristics in the estimation of the propensity score. The propensity score is in general estimated using probit or logistic regression, from a set of possibly abundant observable variables which are supposed to 19 In our sample, Capital (K) and Purchased Services (S) are not available. 14 affect both the probability that the firm enters the program and the future outcome of the firm, based on economic theory and previous empirical findings (Caliendo 2006). In this research, we utilized the age of the firm and dummies variables for the existence of R&D expenditures, export activities and production workers trained. Others dummies variables are the size of the firm (large, medium and small), the country income level (high/ upper-middle), and the existence of domestic owners. The industrial sector was divided into four categories based on the World Bank‘s Enterprise Surveys (ES). As the survey does not contain information related to the performance or productivity of firms, we calculated the technical efficiency of each firm using one output (sales) and five inputs (capital stock , labor cost, materials, energy cost and others costs). We evaluated the effect of overdraft facility policy by observing various aspects of a firm‘s performance. The outcome variables considered are: technical efficiency, R&D expenditures, export activities and production workers trained. The definition of variables including firm characteristics and outcomes are explained in Table 3. Table 3 The definition of variables including firm characteristics and outcomes. Variables Definition Production Sales Used as the measure of output for the production function estimation. For all Function countries, sales figures in local currency are converted into USD using PPP (output and conversion factor to the official exchange rate ratio inputs) Capital Book value of all fixed assets1. For all countries, capital stock figures in local Stock currency are converted into USD using PPP conversion factor to the official exchange rate ratio. Labor cost Total expenditures on personnel2. For all countries, labor cost figures in local currency are converted into USD using PPP conversion factor to the official exchange rate ratio. Materials Total costs of intermediate and raw materials used in production (excluding electricity, fuel, and water). For all countries, materials figures in local currency are converted into USD using PPP conversion factor to the official exchange rate ratio. Energy Total annual costs of electricity, fuel, and water. Cost Other Total annual costs of communications services, transport for goods (not Costs including fuel), and rental of land/buildings, equipment, furniture, etc. Outcomes Technical CCR efficiency index (Bias-Corrected) by Industry (year 2005). efficiency Dummy Dummy variable that takes value 1 whem firm is exporter and 0 whem firm is exporter non-exporter Dummy Dummy variable that takes value 1 if the firm has permanent full-time production productions workers who have been trained (by the firm) workers trained Dummy Dummy variable that takes value 1 if the firm carried out internal R+D during Expenditur 2006 (performed within this establishment). es of R&D Treatment Dummy Dummy variable that take value 1 if the firm has an overdraft facility and control Overdraft facility 15 Matching Age Difference between the year that the firm started operations and current year. Dummy Dummy variable that takes value 1 if the firm has >99 permanent workers Large firm Dummy Dummy variable that takes value 1 if the firm has 20-99 permanent workers Medim firm Dummy Dummy variable that takes value 1 if the firm has <20 permanent workers small firm Dummy Dummy variable that takes value 1 if the country has high/ upper-middle high/ income according with World Bank (2005) upper- middle income country Dummy Dummy variable that takes value 1 if the firm has private domestic owners domestic and 0 if the firma has Foreign owners owners 1 Total annual expenditure for purchases of: machinery and equipment (including vehicles), and land and buildings. 2 Including wages, salaries, bonuses and social payments. Source: Adapted from based in OECD (2001) Economic Studies No. 33, 2001/II, and from World Bank (2007). Most of the variables listed above are related to the firm‘s risk and growth potentials. The size (sales and employment) of a firm is often taken to be a good proxy for the firm‘s risk (Cowling and Westhead 1996). The amount of fixed capital can be related to the availability of collateral, which can act as a sorting device to decide which firms are to receive loans (Bester 1985). R&D and technical efficiency of firms are related to the future growth potentials as well as current profitability. The country income level (high/ upper-middle), and the existence of domestic owners are also used to control the firm‘s risk, value, and growth potential in general. The production workers trained will represent improvement in skill or the quality of employees. Variables indicating R&D and exporter status are introduced to observe if overdraft facility is used to enhance future productivity and expanded production facility. The level of technical efficiency might be an ultimate goal of such a government support policy. 5 Estimation Results 5.1 Estimation of Technical Efficiency To evaluate the efficiency of manufacturing companies, five inputs20 (labor cost, materials and energy cost) and one output (sales) have been incorporated into the model. A synthesis of production function variables and the descriptive statistics for the indicators considered for the sample is presented in table 4. 20 Coelli, et al. (2005), a commonly-used classification of inputs involve five categories: capital (K), labor (L), energy (E), material inputs (M), and purchased services (S). The use of data according to these categories in productivity measurement is sometimes referred to as KLEMS approach. In this study, Capital (K) and Purchased Services (S) are not available. 16 Table 4 Means Difference for Supported and Non-supported Groups Industry Beverages Food Garments Metals and machinery Firms Supported Non- Supported Difference Supported Non- Supported Difference Supported Non- Supported Difference Supported Non- Supported Difference Variable Obs. Mean Obs. Mean T-Test Obs. Mean Obs. Mean T-Test Obs. Mean Obs. Mean T-Test Obs. Mean Obs. Mean T-Test Age 41 25.12 94 16.03 -1.8097** 28 18.54 88 15.14 -0.9826 55 16.33 188 14.95 -0.5961 133 19.84 382 17.99 -1.0110 Variable Obs. % Obs. % T-Test Obs. % Obs. % T-Test Obs. % Obs. % T-Test Obs. % Obs. % T-Test Dummy Large 13 31.71 21 22.34 -1.1501 5 17.86 15 17.05 -0.0982 20 36.36 29 15.43 -3.4739*** 43 32.33 74 19.32 -3.1099*** firm Dummy Medim 15 36.59 32 34.04 -0.2832 8 28.57 36 40.91 1.1687 11 20.00 59 31.38 1.6420* 40 30.08 118 30.81 0.1580 firm Dummy small 13 31.71 41 43.62 1.2974* 15 53.57 37 42.05 -1.0642 24 43.64 100 53.19 1.2457 50 37.59 191 49.87 2.4541*** firm Dummy high/ 10 24.39 28 29.79 0.6374 20 71.43 55 62.50 -0.8561 30 54.55 120 63.83 1.2449 101 75.94 304 79.37 0.8292 upper-middle income country Dummy 5 12.20 21 22.34 1.3741* 10 35.71 23 26.14 -0.9740 16 29.09 76 40.43 1.5254* 25 18.80 112 29.24 2.3583*** domestic owners Note: * significant at 10%; **significant at 5%; and *** significant at 1%. Dummy variables take on the value of 1 if the firm has the stated characteristic and 0 otherwise. Source: Authors‘ calculations 17 The results of the analysis of efficiency with monetary inputs are shown in Table 5. These results reveal the sensitivity of the efficiency measures with respect to sampling variation. The bias- corrected efficiency (BC) reveals that differences in measurement efficiency are of a different magnitude than when the original efficiency scores (N) are considered. For all of the manufacturing sectors, the efficiency declines slightly. Table 5 Descriptive statistics of the output and inputs used by sector and country (year 2005). Variables/industry Beverages Food Garments Metals and machinery Mean Sales (Y) 2476.644 895.9026 592.492 1030.763 Materials Cost (X1) 1282.234 453.563 242.7717 497.1371 Energy Cost (X2) 136.3903 50.45932 35.12418 53.18183 Personal Cost (X3) 209.1959 80.92757 151.6997 149.9907 Median Sales (Y) 196.418 262.3651 118.1627 82.81187 Materials Cost (X1) 85.13541 116.6796 46.78612 37.21292 Energy Cost (X2) 9.772637 12.94952 6.90287 5.241257 Personal Cost (X3) 22.49096 23.00957 27.17094 14.67552 Standard deviation Sales (Y) 9214.583 1604.977 1536.927 3170.787 Materials Cost (X1) 5205.379 856.9301 664.3403 1578.464 Energy Cost (X2) 595.0032 87.42706 104.0559 154.6308 Personal Cost (X3) 641.1238 154.4191 427.9902 445.9945 Maximum Sales (Y) 88827.03 8673.232 13709.87 30116.26 Materials Cost (X1) 52686.09 5364.427 5982.488 15777.23 Energy Cost (X2) 6747.266 587.511 875.7089 1700.407 Personal Cost (X3) 5275.788 773.6096 3373.633 4774.785 Minimum Sales (Y) 2.154254 0.509635 0.2548175 0.3011479 Materials Cost (X1) 0.4787231 0.2162088 0.1235479 0.1389914 Energy Cost (X2) 0.2393615 0.0386087 0.0154435 0.0154435 Personal Cost (X3) 0.4015306 0.1081044 0.0540522 0.0540522 Source: Authors‘ calculation from the Investment Climate Survey Databank, World Bank Having found that there exists wide variation in technical efficiency among the sample firms it is important and useful to understand the impact of the overdraft facility policy. 5.2 Propensity Score Matching Estimation. To better construct the match, table 6 below shows tests for mean differences between supported and non-supported groups for the impact and other key variables. Identification of the matching technique to create the propensity score requires that the impact variables are independent of the treatment conditional on the propensity score. Few statistically significant differences between participants and non-participants were observed, with the exception of the large and younger firms being more prevalent among those receiving overdraft facilities. 18 Table 6 Descriptive statistics of the CCR efficiency index (Normal and Bias-Corrected) by Industry and country (year 2005). Country/industry Beverages Food Garments Metals and machinery CCR efficiency Normal Bias Corrected Normal Bias Corrected Normal Bias Corrected Normal Bias Corrected index Bulgaria 0.781 0.658 0.977 0.866 0.811 0.682 0.780 0.718 Czech Republic 0.611 0.554 0.536 0.504 0.554 0.518 0.687 0.656 Estonia 0.547 0.492 0.503 0.430 0.609 0.566 0.777 0.722 Hungary 0.752 0.653 0.763 0.681 0.719 0.628 0.700 0.656 Latvia 0.784 0.627 0.790 0.627 0.769 0.693 Lithuania 0.751 0.645 0.606 0.564 0.870 0.801 Moldova 0.665 0.598 0.525 0.485 0.560 0.521 0.664 0.637 Poland 0.624 0.571 0.678 0.617 0.663 0.620 0.726 0.685 Romania 0.708 0.636 0.759 0.682 0.706 0.629 0.733 0.682 Slovak Republic 0.647 0.608 0.552 0.519 0.768 0.726 Slovenia 0.702 0.616 0.673 0.621 0.927 0.781 0.912 0.814 Mean 0.693 0.605 0.673 0.611 0.682 0.605 0.762 0.708 Source: DDP Quick Query database of WDI & GDF, World Bank. 5.2.1 Estimation of the propensity score The propensity score calculated to reflect the probability of a firm receiving overdraft facility, predicts that firms that are younger are more likely to receive benefits. The score, presented in Table 7, is predicted based upon a logit model controlling for factors such as age, age squared, and dummy variables (size firm, country income level, and domestic owners). Each additional year of existence (age) lowers the likelihood of receiving overdraft facilities. Country income level and owners characteristics do not seem to influence the likelihood of receiving overdraft facilities. 19 Table 7 Logistic Regression to Estimate Propensity Score for overdraft facility policy. Dependent Dummy for HTAP Subsidy, =1 if treated by the program Variable Industry Beverages Food Garments Metals and machinery Independent Coefficient Standard Z P>z Coefficient Standard Z P>z Coefficient Standard Z P>z Coefficient Standard Z P>z variables error error error error Age -0.01245 0.024976 -0.5 0.618 0.12828** 0.062587 2.05 0.04 -0.01708 0.026247 -0.65 0.515 -0.00263 0.01533 -0.17 0.864 Age squared 0.000164 0.000178 0.92 0.359 -0.00135 0.000882 -1.53 0.126 0.000147 0.000269 0.54 0.586 1.79E-05 0.000155 0.12 0.908 Dummy Large firm 0.447817 0.557797 0.8 0.422 1.305646*** 0.45932 2.84 0.004 0.522057* 0.279631 1.87 0.062 Dummy Medim firm 0.327875 0.45841 0.72 0.474 -0.43002 0.679557 -0.63 0.527 Dummy small firm 0.194275 0.659108 0.29 0.768 0.326923 0.41772 0.78 0.434 -0.1526 0.254381 -0.6 0.549 Dummy high/ upper-middle income country -0.42417 0.469876 -0.9 0.367 0.046765 0.562043 0.08 0.934 -0.05657 0.366503 -0.15 0.877 -0.08481 0.248774 -0.34 0.733 Dummy domestic owners -0.44578 0.565572 -0.79 0.431 0.182613 0.579717 0.32 0.753 -0.28575 0.371397 -0.77 0.442 -0.34143 0.279259 -1.22 0.221 Constant -0.82554* 0.427795 -1.93 0.054 -2.64108*** 0.875654 -3.02 0.003 -1.38919*** 0.439206 -3.16 0.002 -0.93797*** 0.315 -2.98 0.003 Number of obs. 135 116 243 515 LR chi2 6.55 9.29 12.20 12.07 Prob > chi2 0.2128 0.1577 0.0576 0.0604 Log likelihood 0.0351 0.0725 0.0469 0.0205 Pseudo R2 -79.974592 -59.461831 -123.85812 -288.14436 Note: * significant at 10%; **significant at 5%; and *** significant at 1%. Dummy variables take on the value of 1 if the firm has the stated characteristic and 0 otherwise. Source; Authors estimation 20 5.2.2 Review of overlap and common support conditions The distribution of the propensity score for the treatment and control groups is presented in Figure 1 to assess whether there are any differences in each group‘s probability to receive overdraft facility. The distribution for the control group firms is more right skewed, i.e. conditioned towards those firms that are more likely to receive treatment. Observations with a propensity score below 0.10 were dropped, as they had a very low likelihood of being treated. The regions of common support show the observations with propensity scores that were matched are those where the treatment and control groups had similar propensity score values. Figure 1 Distribution of Propensity Scores by sector Beverages Food Garments Metal and machinery 21 In order to compute ATT accurately, one should match the supported and unsupported firms precisely on the basis of the propensity score21. Table 8 shows the results of the Average Treatment Effect on the Treated (ATT) for key outcomes. That is, it shows the difference between the average outcomes for the firms with overdraft facility policy and the average outcomes of those same firms had they not had overdraft facility policy. 21 In practice, it is never possible to match the scores precisely and therefore in this study, three alternative matching methods of nearest neighbor matching, radius, and kernel matching methods were used and compared. The radius matching estimator was conducted with different radius. However, since the results from these four estimators were robust and essentially gave the same implications, here we only presented the results based on kernel matching methods. It should be noted that all the analyses were based on the implementation of common support, so that the distribution of supported and unsupported firms were located in the same domain. However, only a few observations were discarded, and given the large samples, the number of excluded observations is relatively small. Standard errors for support effects for all cases were calculated by the bootstrapping method. 22 Table 8 Average Impact for Firms Receiving overdraft facility. Industry Beverages Food Garments Metals and machinery Impact Obs ATT SE z P>z Obs. ATT SE z P>z Obs. ATT SE z P>z Obs. ATT SE z P>z Variable . Technical 134 107 243 515 efficiency -0.02649 0.025574 -1.04 0.3 0.03408 0.034031 1 0.317 -0.0262 0.018725 -1.4 0.162 -0.00291 0.0122 -0.24 0.811 Dummy 134 107 243 515 exporter -0.10264** 0.051905 -1.98 0.048 0.087419 0.066266 1.32 0.187 -0.03621 0.067501 -0.54 0.592 0.033311 0.047338 0.7 0.482 Dummy 129 103 236 496 production workers trained -0.02158 0.099918 -0.22 0.829 0.01824 0.127035 0.14 0.886 -0.03743 0.072809 -0.51 0.607 0.047712 0.04969 0.96 0.337 Dummy 103 86 207 449 Expenditures of R&D 0.018612 0.089425 0.21 0.835 -0.07904 0.08403 -0.94 0.347 -0.02203 0.063987 -0.34 0.731 0.119596*** 0.0475 2.52 0.012 Note: * significant at 10%; **significant at 5%; and *** significant at 1%. Dummy variables take on the value of 1 if the firm has the stated characteristic and 0 otherwise. Impact variables: Technical efficiency variable is continuous, Exporter, Production workers trained and Expenditures of R&D and are not. Matching variables: Age, Age squared, dummy variables for size (Large, Medium and Small), dummy variable for high/ upper-middle income country and dummy variable for domestic owners. Weighting matrix: inverse variante Number of matches (m) = 4 Number of matches, robust std. err. (h) = 4 Source: Authors‘ estimates 23 The results indicate that the sole statistically significant impact attributable to overdraft facility policy is that these firms have a higher incidence of doing R&D activities. The sign of the coefficients representing the impact of overdraft facility on expenditures of R&D activities are positive but onlu statistically significant in Metals and machinery sector. 5.2.3 Assessing matching quality The quality of the propensity score match is conditional on the validity of the conditional independence assumption. That is, the treatment and control groups should be as similar as possible and have no significant differences in terms of observables and unobservable characteristics. Table 9 shows that following the match observable statistically significant differences between the treatment and control group exist. However, to identify a causal relationship, matching requires that the treatment and control groups be similar in unmeasured characteristics. It is uncertain how likely this assumption is because treated firms had the initiative to apply to overdraft facility while the control did not, and there may be other unobserved differences between the treatment and the control group (e.g., the degree of entrepreneurship and quality of management teams). 24 Table 9 Means Difference for Supported and Non-supported Groups after Matching Industry Beverages Food Garments Metals and machinery Firms Supported Non- Supported Difference Supported Non- Supported Difference Supported Non- Supported Difference Supported Non- Supported Difference Variable Obs. Mean Obs. Mean T-Test Obs. Mean Obs. Mean T-Test Obs. Mean Obs. Mean T-Test Obs. Mean Obs. Mean T-Test Age 40 21.25 94 16.03 -1.1901 27 19 80 13.59 -2.4066*** 55 16.33 188 1.02 -0.5961 133 19.84 382 17.99 -1.0110 Variable Obs. % Obs. % T-Test Obs. % Obs. % T-Test Obs. % Obs. % T-Test Obs. % Obs. % T-Test Dummy Large 12 30.00 21 22.34 -0.9378 5 18.52 14 17.50 -0.1186 20 36.36 29 15.43 -3.4739*** 43 32.33 74 19.37 -3.0944*** firm Dummy Medim 15 37.50 32 34.04 -0.3811 7 25.93 30 37.50 1.0892 11 20.00 59 31.38 1.6420* 40 30.08 118 30.89 0.1752 firm Dummy small 13 32.50 41 43.62 1.1981 15 55.56 36 45.00 -0.9446 24 43.64 100 53.19 1.2457 50 37.59 190 49.74 2.4271*** firm Dummy high/ 9 22.50 28 29.79 0.8593 20 74.07 51 63.75 -0.9769 30 54.55 120 63.83 1.2449 101 75.94 303 79.32 0.8153 upper-middle income country Dummy 5 12.50 21 22.34 1.3168* 10 37.04 22 27.50 -0.9309 16 29.09 76 40.43 1.5254* 25 18.80 111 29.06 2.3193** domestic owners Note: * significant at 10%; **significant at 5%; and *** significant at 1%. Dummy variables take on the value of 1 if the firm has the stated characteristic and 0 otherwise. Source: Authors‘ calculations 25 6 Summary and Discussion of Results This research evaluates the impact of overdraft facility policy, for recipient firms by controlling the selection bias with the up-to-date propensity score matching technique. The results are based on an investigation of the differences in the performance indicators from 2005, for the firms receiving overdraft facility policy. In 2005, growth in the eight new European Union (EU) member States from Central and Eastern European (EU-8) economies had become more broadly based, driven by robust consumption and investment expenditure and strengthening external demand. On the other hand, most Southern and Eastern European economies had preserved a strong growth momentum in 2005. Finally, macroeconomic policy in most of these countries had remained restrictive.22 23 For all of these factors, 2005 have been chosen to establish performance baseline and establish the impact of overdraft facility policy. In this context, the empirical analysis in Eastern European Union industrial firms showed that the overdraft facility policy would be needed to stimulate investment in R&D, which will eventually result in growth in productivity. We could not find sufficient evidence that the overdraft facility has affected positively technical efficiency, production workers trained, and sales export activity. Since this research was based on relatively short period of time not accompanied with significant changes in overdraft facility schemes, the comparative analyses regarding the changes on the parameters were not possible. The designing of these parameters to reach the optimal level could remain as an interesting further research topic. On the other hand, comparing with today situation, we have conducted our empirical analysis in a very different economic scenario. In this context, a longitudinal study with new data could provide a greater richness to the results, capturing the effects that persist over time. Extensions of future research could be aimed at addressing the limitations discussed. 22 The OECD (2001b) provides a framework allowing policy-makers to identify strong and weak points in their country‘s business environment by comparing their performance and business environment to that of other OECD countries. The OECD report concludes that four micro-drivers (human capital, information and communications technology, innovation and entrepreneurship) are key drivers of productivity and economic growth performance in knowledge-based economies 23 The most notable micro-policies highlighted in the OECD report are: (1) increasing access to venture capital corresponds to the driver, ―fostering firm creation and entrepreneurship‖, (2) Enhancing public-private partnerships for innovation relates to the driver ―harnessing the potential of innovation and technology diffusion‖, (3) Promoting Information and Communications Technology (ICT) diffusion to business is the driver ―seizing the benefits of ICT‖, and (4) Developing highly-skilled workers for future industry needs corresponds to the driver ―enhancing human capital and realizing its potential 26 References Abadie, A.; Imbens, G. 2006. Large sample properties of matching estimators for average treatment effects, Econometrica, 74, 235–267. Alcalá, F. 2004. La ampliación de la Eurozona, Información Comercial Española, ICE, 817, 143- 160. Augier, P.; Dovis, M.; Gasiorek, M. 2010. The Business Environment and Moroccan Firm Productivity, CARIS working paper, 2010-4. Blundell, R.; Costa Dias, M. 2000. Evaluation methods for non-experimental data, Fiscal Studies, 21(4), 427–468. Boocock, G.; Shariff, M. 2005. Measuring the effectiveness of credit guarantee schemes: Evidence from Malaysia, International Small Business Journal, 23, 427–454. Brunetti, A. 1997. Political Variables in Cross-country Growth Analysis, Journal of Economic Surveys, Vol. 11, No. 2, 163-190. Butler, A.; Cornaggia, J. 2011. Does access to external finance improve productivity? Evidence from a natural experiment, Journal of Financial Economics, Vol. 99, Issue 1, 184–203 Caliendo, M. 2006. Microeconometric evaluation of labour market policies. Berlin: Springer. Clarke, G.: Cull, G.; Kisunko, G. 2012. External finance and firm survival in the aftermath of the crisis: Evidence from Eastern Europe and Central Asia, Journal of Comparative Economics, Available online. http://dx.doi.org/10.1016/j.jce.2012.01.003 Coelli, T.; Prasada Rao, D.S.; O‘Donnell, C.; Battese, E. 2005. An Introduction to Efficiency and Productivity Analysis, Second Edition. New York: Springer. Cooper, W.; Park, K.; Yu, G. 2001. IDEA (Imprecise Data Envelopment Analysis) with CMDs (Column Maximum Decision Making Units), Journal of the Operational Research Society, Vol. 52, No. 2, 176-181. Cowling, M. 1998. Regional determinants of small firm loans under the U.K. loan guarantee scheme, Small Business Economics, 11, 155–167. Cowling, M.; Mitchell, P. 1997. The evolution of U.K. self-employment: A study of government policy and the role of the macroeconomy, The Manchester School, 65, 427–442. Charnes, A.; Cooper, W.; Rhodes, E. 1978. Measuring the Efficiency of Decision Making Units, European Journal of Operational Research, Vol. 2, No. 6, 429-444. Chen, M.; Guariglia, A. 2011. Financial Constraints and Firm Productivity in China: Do Liquidity and Export Behavior Make a Difference?, Available at SSRN: http://ssrn.com/abstract=1863864. Esteban, L.; Gallizo, J.L.; Hernández, J.M. 2002. Eficiencia Técnica y Convergencia en la Industria Manufacturera de la Unión Europea, Estudios de Economía Aplicada, Vol. 20, No. 2: 381- 401. 27 Farrell, M.J. 1957. The Measurement of Productive Efficiency, Journal of the Royal Statistical Society, Series A (General), Vol. 120, Part 3, 253-290. Foster, L.; Haltiwagner, J.; Krizan, C. 1998. Aggregate Productivity Growth: Lessons from Microeconomic Evidence, NBER Working Paper, No. 6803. Frolich, M. 2004. Finite-sample properties of propensityscore matching and weighting estimators, The Review of Economics and Statistics, 86, 77–90. Gatti, R.; Love, I. 2008. Does access to credit improve productivity? Evidence from Bulgaria, Economics of Transition, 16: 445–465. Hausmann, R.; Gavin, M. 1996. Securing Stability and Growth in a Shock Prone Region: The Policy Challenge for Latin America, Inter-American Development Bank, Research Department Working Papers, No. 315. Hay, J. 2003. Desafíos: la Unión Europea ante su ampliación, Madrid: Siddhart Metha Ediciones. IMF 2005. Republic of Korea: 2004 article IV consultationstaff report; staff statement; and public information notice on the executive board discussion, IMF Country Report 05/49, International Monetary Fund, Washington, DC. Jaffe, A. 2002. Building programme evaluation into the design of public research-support programmes, Oxford Review of Economic Policy, 18(1), 22–34. Kang, D. 2005. Corporate distress and restructuring policy of Korean small and medium sized enterprises: Role of credit guarantee scheme. (Paper presented at the 2005 KDI-KAEA conference, Korea Development Institute, Seoul, Korea). Kasseeah, H. 2011. Financial Constraints, Gender and Firm Performance: Evidence from Mauritius, Middle Eastern Finance and Economics, Issue 13, 60-88. Lee, K. 2006. An evaluation of loan guarantee system for S&M firms in Korea, The Korea Journal of Public Finance, 20, 203–229. Leibenstein, H. 1966. Allocative Efficiency vs. ―X-Efficiency‖, American Economic Review, Vol. 56, No. 3, 392-415. Lerner, J. 2002. When bureaucrats meet entrepreneurs: The design of effective ―public venture capital‖ programmes, The Economic Journal, 112, F73–F84. Loof, H.; Heshmati, A. 2005. The Impact of public funding on private R&D investment: New evidence from a firm level innovation study, CESIS Electronic Working Paper Series No. 6., Centre of Excellence for Science and Innovation Studies, Sweden. Muthen, B.; Jöreskog, K.1983. Selectivity Problems in Quasi-Experimental Studies, Evaluation Review, 7(2):139–174. OECD 2001a. Measuring Productivity, Economic Studies No. 33, 2001/II. Paris. 28 OECD 2001b. Organisation for Economic Co-operation and Development, The New Economy: Beyond the Hype, Paris, 27-68. Oh, I.; Lee, J.; Heshmati, A.; Choi, J. 2009. Evaluation of credit guarantee policy using propensity score Matching, Small Business Economy, 33:335–351. Rosenbaum, P.; Rubin, D. 1983. The central role of the propensity score in observational studies for causal effects, Biometrica, 70, 41–55. Rubin, D. 1974. Estimating causal effects of treatments in randomized and non-randomized studies, Journal of Educational Psychology, 66, 688-701. Simar, L.; Wilson, P. 1998. Sensitivity Analysis of Efficiency Scores: How to Bootstrap in Nonparametric Frontier Models, Management Science, Vol. 44 (January 1998): 49-61. Simar, L.; Wilson, P. 2000. A general methodology for bootstrapping in non-parametric frontier models, Journal of Applied Statistics, Taylor and Francis Journals, vol. 27(6), 779-802. Simar, L.; Wilson, P. 2007. Estimation and inference in two-stage, semi-parametric models of productive efficiency, Journal of Econometrics 136 (2007) 31–64. Smith, J. 2000. A critical survey of empirical methods for evaluating employment and training programs, Swiss Journal of Economics and Statistics, 136(3), 247–268. Stiglitz, J.; Weiss, A. 1981. Credit rationing in markets with imperfect information, The American Economic Review, 71, 393–410. Tybout, J. 1992. Linking Trade and Productivity: New Research Directions, The World Bank Economic Review, Vol. 6, No. 2, 189-211. Yasar, M.; Rejesus, R. 2005. Exporting status and firm performance: Evidence from a matched sample, Economics Letters, 88, 397–402. World Bank 2007. Enterprise Survey: Understanding the Sampling Methodology. New York. 29