SMEs and Export Performance in South East Europe Staffordshire University Business School Centre for Applied Business Research (CABR) Working Paper No. 002/2012 The Small and Medium Enterprise Sector and Export Performance: Empirical Evidence from South Eastern Europe Petrit Gashi, Faculty of Economics, University of Prishtina, St. Ramiz Sadiku, nn.10000 Prishtina, Kosova Iraj Hashi Staffordshire University, Stoke-on-Trent, United Kingdom Geoff Pugh Staffordshire University Business School, Leek Road, Stoke-on-Trent, ST4 2DF, UK. March 2012 Final_1: September 10, 2008 1 SMEs and Export Performance in South East Europe The Small and Medium Enterprise Sector and Export Performance: Empirical Evidence from South Eastern Europe Abstract The export performance of small and medium-sized enterprises (SMEs) is analysed using an eclectic theoretical framework. Using the not-much-explored large data base of the World Bank/EBRD Business Environment and Enterprise Performance Surveys (BEEPS), we investigate the influences of various internal and external factors on SME export performance in South Eastern Europe (SEE). Our econometric estimates highlight firm size, ownership, sector of activity, the availability of external finance, affiliation with business organisations, the education of the workforce and, to a lesser extent, technology-related factors as major influences. The BEEPS data base contains a large number of missing observations for many variables – a shortcoming not highlighted by previous authors using this data base. We use recently developed multiple imputation techniques (MI) to avoid sample bias arising from missing values, an otherwise endemic problem associated with survey data. JEL classifications: F23, M16, O16 Key words: export behaviour, SMEs, human and technology related factors, multiple imputation. Final_1: September 10, 2008 2 SMEs and Export Performance in South East Europe 1. INTRODUCTION It is now well established that small and medium sized enterprises (SMEs) play a vital role in the process of transition to a market economy. As the large firm sector, the prevalent form of organisation under central planning, underwent restructuring and decline, thousands of new SMEs took advantage of liberalised entry conditions and entered the market. They responded rapidly to systemic shocks, produced goods and services demanded by the population and, in the process, contributed to the generation of new jobs and incomes.1 While the contribution of SMEs to domestic output and employment has been studied by a variety of authors for many transition economies, their role in cross-border trade and their contribution to exports has not been studied widely. The aim of this paper is to develop the research in this area by investigating the factors influencing the export propensity of SMEs and by providing empirical evidence for countries of South Eastern Europe (SEE). We have chosen to focus on these countries because of the relative similarities in their development trajectories and also because of the paucity of the literature in this area. The bulk of previous studies have relied on small sample surveys conducted by governments, SMEsupport institutions or by the authors themselves. This paper employs large datasets drawn from the Business Environment and Enterprise Performance Surveys (BEEPS) conducted jointly by the World Bank and EBRD, which have remained relatively underutilised (Carlin et al., 2001b; Vagliasindi, 2001; Vagliasindi, 2006; Svejnar and Commander, 2007; Gorodnichenko et al., 2008; and Transition Report 2005 are some of the studies based on BEEPS). An important constraint on our analysis is the absence of well-developed theory on the behaviour of SMEs (Brock and Evans, 1989) and, in particular, on SMEs and international trade. Moreover, papers linking the small firm sector with international trade in the transition context are especially scarce. For this reason, in Section 2 we adopt an eclectic approach drawing not only on a largely empirical SME literature but also on a range of economic theories. Section 3 presents our empirical strategy, the datasets, and applies Multiple Imputation (MI) to deal with missing data. Section 4 reports and discusses the econometric results, and Section 5 reports the robustness checks. The final section concludes. 1 See, for example, Bartlett and Prasnikar (1995); Futo, et al. (1997); Scase (1998); McMillan and Woodruff (2002); Hoshi, et al. (2003); Iakovleva (2005); and Estrin, et al. (2006) among many other contributions. Final_1: September 10, 2008 3 SMEs and Export Performance in South East Europe 2. DETERMINANTS OF EXPORT PERFORMANCE: AN ECLECTIC THEORETICAL FRAMEWORK In line with previous research, we use export intensity (foreign sales as a percentage of total sales) to measure the degree of firms’ involvement in foreign markets (Bonaccorsi, 1992; Calof, 1993; Wakelin, 1998; Becchetti and Rossi, 2000; Wagner, 2001; Gorodnichenko et al., 2008; and others). Similarly, we assess the firm’s export behaviour in two ways: namely, by measuring the likelihood of exporting as well as the level of firm’s involvement in export markets (Kumar and Siddharthan, 1994; Wakelin, 1998; Sterlacchini, 1999; and others). To guide the specification of an empirical model to explain the export performance of firms, we draw upon the SME literature and a range of economic theories to argue for three types of independent variables: human-related factors; technology-related factors; and other firm characteristics (such as firm size, ownership structure, type of sector, etc.). We discuss each of these in turn. Human capital related factors Human capital is at the core of the New Growth Theory, which argues that the creation and diffusion of knowledge is the primary engine of economic growth (Grossman and Helpman, 1994). At the micro level, human capital is considered to contribute to sustainable competitive advantage of firms (Bryan, 2006). Accordingly, we hypothesise that human capital factors affect firms’ export performance indirectly through increases in productivity. Chevalier et al. (2004) argue that greater levels of education or skill acquisition signal or enhance productivity. In addition, according to Bryan (2006), training helps to sustain higher levels of productivity. In our model of export behaviour we measure the impact of human capital accumulation through several proxies: [i] the education of the workforce; [ii] the presence of highly skilled workers within the firm, which includes also the managerial staff and other professionals; [iii] on-the-job-training; [iv] the general manager’s tenure; and [v] the general manager’s education. A number of studies (Keeble et al., 1991; Wood, 1991; Dex and Scheibl, 2001, 2002; Power and Reid, 2005; etc.) argue that SMEs are more inclined to have flexible organisational arrangements than are larger firms, because of their limited scope of operations, wellunderstood relationships within the firm, relatively simple organisational structures, ease of accessing networks of firms, etc. Conversely, Meijaard et al. (2005) argue that organisational structures within SMEs are much more complex than is argued by transaction costs and Final_1: September 10, 2008 4 SMEs and Export Performance in South East Europe agency costs theories. We investigate whether or not organisational flexibility translates into better export performance by a dummy variable indicating if a firm underwent any organisational transformation (from minor reallocations to adoption of completely new organisational arrangements) in the previous three-year period. Technology-related factors The New Growth Theory also highlights the role of innovation and technological change in improving the competitiveness of firms and economies. There are several ways of measuring or accounting for innovation and technical change. Firstly, following Carlin et al. (2001a), and relying on the Endogenous Growth Theory, we use investment in capital goods as a proxy for embodied technological change, and expect it to have a positive impact on the export performance of the firms under consideration.2 The investment-export relationship may also be explained by the accelerator model. According to this approach, investment responds positively to changes in demand conditions, including new export opportunities. This, of course, raises the empirical complication of potential endogeneity. Secondly, R&D expenditure can be used as an indicator of the innovation process (an input measure of innovation) to investigate its effect on the export performance of firms.3 Thirdly, the introduction of new or upgraded technology and new or upgraded products can be used as another indicator of the innovation process (another input measure), expected to have positively affected the firm’s export performance. Finally, a firm’s level of technology relative to its main rivals may also be used as an indication of technological progress, with positive impact on export performance.4 The last two indicators reflect process or product innovation and the level of technological sophistication and are expected to translate into greater export capabilities. Likewise, firms with higher observed levels of technology are 2 The 2002 data base provides information on the average investment-sales ratio in the previous three years (measuring investment expenditure on new buildings, machinery and equipment in the previous three years as a percentage of annual sales over the same period). For the 2005 data base, however, BEEPS does not provide the investment-sales ratio but only the expenditure on new buildings, machinery and equipment (i.e. investment) in the previous year (2004). Given that the total sales figure is not published in the 2005 BEEPS, it is not possible to calculate a comparable investment-sales ratio for the 2005 dataset, and also for the pooled and panel datasets. Therefore these three specifications are estimated without the investment-sales ratio variable. 3 The 2002 data base includes information on R&D intensity - R&D expenditure in the previous three years as a percentage of total sales in the previous three year. However, as with the investment intensity, the 2005 data base provides information on R&D expenditure in the previous year (2004) and not R&D intensity. The 2005 estimation is subject to the same problem as that discussed for investment expenditure. 4 This information, too, is contained in the 2002 survey only. Final_1: September 10, 2008 5 SMEs and Export Performance in South East Europe expected to be more export oriented. As in the case of the investment-sales ratio variable, one may raise concerns about potential endogeneity affecting the latter variable. Other firm characteristics The literature largely supports the export proficiency of larger firms relative to smaller firms on the grounds of resource availability and lower transaction costs (Brock and Evans, 1989; Kim et al. 1997; Acs et al. 1997; Wakelin, 1998; Bleaney and Wakelin, 2002; etc.). However, a number of studies support the idea that smaller firms perform better in export markets due to their inherent flexibility (Mills, 1984; Mills and Schumann, 1985; etc). Accordingly, ambiguous results may be obtained for firm size variables, which are expressed as dummies for small and medium sized firms. The Industrial Economics literature has established the impact of ownership structure on firm performance. Demsetz (1997, p. 429), for example, argues that wealth and its distribution among different stakeholders matters to society’s productivity. Furthermore, the transition economics literature has established the superior performance of foreign owned firms in many countries. Here, focusing on export activities, we investigate the impact of ownership structure on the export performance of firms – ownership structure identifying (i) whether the firm is state owned or privately owned, and (ii) whether a firm is foreign owned or domestic owned. The percentage share of private capital and the percentage share of foreign capital in a company are used to measure these two aspects of ownership. There is some evidence showing that the performance gap between foreign and domestic companies is not solely a result of foreign ownership, but also of firm-specific assets and firm characteristics such as size, industry, parent country, etc. (Moosa, 2002; Bellak, 2004). We draw also upon the so-called Base-Multiplier Hypothesis (Fujita et al. 1999), according to which manufacturing export is the ‘economic base’ of an economy. In turn, ‘non-base activities’ are derived from the base and grow and shrink depending on the base’s performance. From this theory, we hypothesise that production activities (which consist of manufacturing together with mining and quarrying) should show a greater tendency towards exporting. To capture these effects we use the percent of company sales generated by production activities. Final_1: September 10, 2008 6 SMEs and Export Performance in South East Europe Institutional Economics has highlighted the impact of formal and informal institutions on economic performance at both firm and country levels. Two variables reflecting the development of the institutional framework (both formal institutions and the supporting legal framework) are used in this study. On the one hand, we investigate the impact of membership in business associations on SME export performance. For successful export activities, systematic collection of information is required, since it can act as a catalyst to reduce the uncertainties of the international environment (Leonidou and Adams-Florou, 1999). In addition, due to their resource constraints, SMEs appear to be more dependent than large firms on services, information and contacts generated through associations (Bennett, 1998). On the other hand, we assess the impact of the level of financial development on SMEs exporting activity. Many country-level studies have demonstrated that financial development positively affects the export performance of firms reliant on external funding (Beck, 2001 and 2003; Manova, 2006; and Becker and Greenberg, 2005) especially in industries where scale factors and sunk costs play an important role. SMEs have even greater need for credit relative to large firms due to their limited capital resources. Moreover, SMEs face greater difficulties in obtaining external finance (due to information asymmetries and/or institutional factors) which may be reflected in their overall performance including international activities (Beck et al. 2004 and 2006; Hutchinson and Xavier, 2006). Here, we use the percentage of a firm’s working capital and fixed investment financed by credit from commercial banks as a measure of the availability of external finance, and test for the link between external financing and the export behaviour of SMEs in the SEE region. Two indicators of the firm’s experience are also included in our models, expected to be conducive to export activity: the number of years since the firm was established; and the number of years since the company started exporting. We rely on Learning Theory – rooted in the behavioural theory of the firm – which argues that development of knowledge may have an impact on perceptions of opportunities offered by greater internationalisation (Clercq et al., 2005). Regarding the effect of exporting experience, “learning-by-exporting” suggests a positive association between current export performance and the firm’s past export experience. In addition, we allow for a non-linear relationship between business experience and export growth, hence increasing or decreasing returns to experience. We anticipate a Ushaped relationship between firm’s age and exports due to the pattern of firm survivorship (Everett and Watson, 1998); namely, that the rate of failure among younger firms is higher than among experienced ones, due to the greater variability in their cost functions while Final_1: September 10, 2008 7 SMEs and Export Performance in South East Europe learning. Everett and Watson concentrate on firms’ experience in the domestic market. However, this effect may be more pronounced in foreign markets, where cost variability is likely to be higher to the extent that foreign markets are unfamiliar and entrepreneurs face lack of information and different systems as well as different languages and cultures. On the other hand, New Trade Theory has focused on the effect of productivity and sunk costs on the firm’s decision to enter/exit export markets. Export experience has been used in several studies to explain patterns of firms’ entry and exit strategies in the presence of sunk costs. In this literature, an inverse U-shaped relationship has been identified between export experience and export performance (Roberts and Tybout, 1997; Bernard and Jensen, 2004; and Bernard and Wagner, 1998). Two more firm-related aspects that we investigate are the level of capacity utilisation (facilities and manpower) and market share. First, a greater utilisation of resources is expected to influence export performance positively as it is an indication of improved efficiency and competitiveness. On the other hand, greater exports should increase capacity utilisation, thereby raising the issue of endogeneity. Second, we assume that firms with a greater share of the domestic market (more than 5 percent) would have the incentive to try to expand their activity across borders to take advantage of greater demand in foreign markets. Accordingly, we anticipate that the likelihood of exporting would be higher for firms that have a 5 percent or larger share of the domestic market. Finally, we control for potential differences in the exporting behaviour of firms in different sub-regions of SEE. We identify two sub-regions based on their level of economic and institutional development: the countries of the Western Balkans (i.e. Albania, Bosnia, Macedonia, and Serbia and Montenegro); and the countries that are already EU members or candidate countries (Bulgaria, Croatia, and Romania). Although countries within these two sub-regions may have had many similarities in terms of social and political environment, core economic activities and their socialist past, their initial economic conditions and transition paths have been different and the experience of joining the EU or becoming a candidate country has tended to increase the development gaps between them.5 In addition, these countries have had different success with policy and institutional reforms. The development Although Macedonia has now achieved the status of a ‘candidate country’, this was not the case in the period under consideration. Were it not for political reasons, Croatia would have joined the EU some time ago. 5 Final_1: September 10, 2008 8 SMEs and Export Performance in South East Europe path may have led to companies in Bulgaria, Croatia and Romania having different approaches to exporting compared to their counterparts in the Western Balkans region. 3. EMPIRICAL STRATEGY 3.1 Methodology As the BEEPS database contains information on exporters and non-exporters, the dependent variable (y) – percentage share of export sales in total sales – is zero in a significant number of cases (i.e. for non-exporters), and the observations for exporters are roughly continuous over the positive range of values. This type of data is addressed by the generalised tobit model (Wooldridge, 2003, p.565). In our specific case, there is an additional complication as the dependent variable is a proportion, therefore bounded by zero and one. However, diagnostic checks endorse the tobit model as a valid estimator for our data.6 The model for cross-section data has the following form: y x i ε i yi i 0 ε i ~ N(0, σ 2 ) for exporters for non - exporters And, for panel data: x α i ε it y *it it 0 α i ~ N(0, σ ε2 ) for exporters wher e i 1, 2, ..., N and, t 2002 and 2005 for non - exporters ε it ~ N(0, σ ε,2 t ) x is a 1k vector containing the k variables of interest discussed in Section 2; and β is the 6 Maddala (1977, 162-63) and Wooldridge (2002, pp.518-19) discuss the use of tobit models to estimate models where the dependent variable is generated by, in effect, a dual decision making process: in our case, firms’ decisions as to whether or not to export and, if so, how much to export. The advantage of tobit estimation is that zero observations are incorporated into the model as the outcome of a decision-making process. Accordingly, zero observations potentially yield useful information. Moreover, truncation at one is unlikely to affect our estimates in a substantial manner: in our pooled sample, for example, only 2.45 percent of firms generate 100 percent of their sales from exports (four percent when the upper limit is set at 95 percent). Nonetheless, we implemented two robustness checks to address residual concerns on this issue. We replicated our preferred model using our pooled sample: firstly, we implemented tobit estimation with censoring at both zero and one; secondly, we implemented the generalized linear model recommended by Baum (2008, p.301) for modelling ‘proportions data in which zeros and ones may appear as well as intermediate values’. In neither case were the estimates substantially different from those reported below. Finally, we note that in Tobin’s (1956) original presentation of what came to be known as the tobit model, his dependent variable is a proportion. Final_1: September 10, 2008 9 SMEs and Export Performance in South East Europe corresponding k1 vector of coefficients to be estimated. Although a 2-year panel sample is not sufficient to identify any dynamics in the data, it is sufficient to estimate a random effects tobit model, which accounts for unobserved effects that are constant over time but vary between firms by means of the firm-specific error term i. We follow Wooldridge (2002, pp. 521-524; and 2003, pp.567-569) who distinguishes between two types of marginal effects: the ‘conditional’ marginal effects, which account for changes in the expected (or predicted) value of exports (y) for the subpopulation of firms for which exporting activity is observed (y>0); and the ’unconditional’ marginal effects that account, in addition, for the effect of changing values of the independent variables on the probability that exporting will take place at all (i.e., will change from zero to positive and, hence, be observed). For dummy variables, both conditional and unconditional marginal effects are calculated as the discrete change in the expected value of the dependent variable as the dummy variable changes from zero to one. 3.2 The data 7 The data used in this investigation are a sub-sample of BEEPS, an extensive survey targeting the business environment and the performance of enterprises in Central and Eastern Europe (CEE) and the Commonwealth of the Independent States (CIS). The survey has been conducted three times: 1999 - 4,104 firms (excluding farms) in 25 post-communist countries (Serbia and Montenegro was not covered in this year); 2002 - 6,667 firms in 27 countries; and 2005 - 9,655 enterprises in CEE, CIS, and Turkey.8 We derive four sets of data from the 2002 and 2005 BEEP surveys to estimate the econometric models presented earlier; namely, two for the individual years: a pooled dataset; and a panel component of SEE companies surveyed in both years. The use of these different datasets provides an internal check on the robustness of the results. Our SME datasets consist respectively of 1,246 observations for 2002; 1,563 for 2005; 2,498 for the pooled data; and, 588 for the panel sample. It encompasses seven SEE countries: Albania; Bosnia; Bulgaria; Croatia; Macedonia; Romania; and Serbia and Montenegro. The definition of our variables and their summary statistics are provided in Tables 1 and 2. 7 Only a short description of the content of the BEEPS dataset is provided here as the details can be found on: http://info.worldbank.org/governance/beeps and http://www.ebrd.com/pubs/econo/beeps.htm (both accessed on February, 2008). See also the EBRD 2005 Transition Report. 8 We do not present results estimated from the 1999 survey partly for reasons of space and partly because many variables covered in the 2002 and 2005 surveys were not included in the 1999 survey. The 1999 estimates are generally consistent with those presented here. Final_1: September 10, 2008 10 SMEs and Export Performance in South East Europe Table 1. Description of variables used in the econometric specifications Dependent variable expint Export intensity – the percentage of total sales generated by exportsa Independent variables ftwor_edu Education of the workforce – the share of the workforce with some university or higher education training Dummy for firms who have conducted on-the-job-training skilled The percentage of a firm’s current skilled full-time workers (including also the managerial staff and other professionals) in the total workforce org_str Dummy for firms which underwent changes in organisational structures gross_inv Spending on new buildings, machinery and equipment since 1998 as a percentage of the firm’s sales over the same period (for 2002 only) inv_rd Spending on research and development (including wages and salaries of R&D personnel, materials, R&D related education and training costs) since 1998 as a percentage of the firm’s sales over the same period (for 2002 only) prli_tech Dummy for firms who established/upgraded products or introduced new technology between surveys med Dummy for medium sized firms private Ownership structure – the percentage share of private capital in the company foreign Ownership structure – the percentage share of foreign capital in the company entact Type of activity – the share of sales generated by production activities credit Access to external finance – the proportion of the firm’s working capital and new fixed investments financed by external sources bus_assoc Dummy for membership in business associations age Business experience – years since establishment exp_age Export experience – years since starting with exports cap_util Current capacity utilisation of facilities/manpower see4 Dummy for the Western Balkans sub-region (Albania, Bosnia, Macedonia, and Serbia and Montenegro - distinguished from Bulgaria, Romania and Croatia, which are already members or candidate countries of the EU) reinvest_prof For 2002: the percentage (expressed in levels/intervals) of profits reinvested in 2001; for 2005: the percentage of profits in 2003 reinvested in the firm in 2004 The variable is expressed in percentage terms (from 0-100 percent). The question in BEEPS asks: ‘What percentage of your firm’s sales is exported directly?’ a The summary statistics in Table 2 show that the average share of export sales in total sales is almost identical for all years in which the survey was conducted; namely, approximately 10 percent. In all sub-samples small firms comprise almost three quarters of the total, and are mainly private, domestically owned. They engaged in trade and services, with the share of production activities in 2002 and 2005 being just over 25 percent. We omit further discussion of the descriptive statistics for reasons of space. Final_1: September 10, 2008 11 SMEs and Export Performance in South East Europe Table 2. Summary statistics for the variables used in the parsimonious specification Variable Datasetsa expint ftwor_edu training skilled org_str gross_inv inv_rd prli_tech med private foreign entact credit bus_assoc age exp_age cap_util Final_1: September 10, 2008 Panel 2002 2005 Panel 2002 2005 Panel 2002 2005 Panel 2002 2005 Panel 2002 2005 2002 2002 Panel 2002 2005 Panel 2002 2005 Panel 2002 2005 Panel 2002 2005 Panel 2002 2005 Panel 2002 2005 Panel 2002 2005 Panel 2002 2005 Panel Panel 2002 2005 Fractions 1 42.95 38.95 40.39 45.08 51.16 35.32 32.07 28.44 31.18 22.88 22.95 23.93 60.13 53.21 54.83 - 0 57.05 61.05 59.61 54.92 48.84 64.68 67.93 71.56 68.82 77.12 77.05 76.07 39.87 46.79 45.17 - Mean 10.33 9.91 9.89 25.85 26.67 24.60 78.15 78.25 78.06 6.13 1.73 83.57 80.68 91.40 9.23 13.10 8.37 25.56 25.25 27.16 22.70 16.30 24.00 16.24 12.73 15.32 2.44 80.19 81.63 81.26 Std. dev. Min Max % missing 24.60 23.80 24.13 27.64 29.32 27.40 26.17 26.07 26.62 7.03 5.02 35.99 38.13 26.81 26.58 31.30 25.84 40.75 40.40 42.05 45.17 38.82 45.40 16.83 14.80 16.15 7.90 20.53 19.38 21.01 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 3 4 0 10 15 10 100 100 100 100 100 100 100 100 100 58 70 100 100 100 100 100 100 100 100 100 200 200 200 138 138 180 104 100 100 100 0.33 0.96 0.19 0.49 1.60 1.80 3.76 2.57 9.34 0.51 2.32 1.02 0.33 1.20 0.45 11.80 11.80 0.65 1.12 0.96 0 0 0 0 0 0 1.30 0 0 0.43 0 0 0.98 2.09 2.11 0.14 0 0 0 0 0 22.79 1.36 4.57 1.28 12 SMEs and Export Performance in South East Europe see4 reinvest_prof a Panel 2002 2005 2002 2005 38.56 52.81 51.70 77.12 - 61.44 47.19 48.30 22.88 - 48.18 42.32 0 100 0 0 0 5.54 5.25 Not all information are available for all variables; hence specifications differ (see Table 3). 3.3 HANDLING MISSING DATA The rate of missing values in the 2002 and 2005 samples is fairly low for most variables (see Table 2). The highest rate of missing values is for export experience in the panel data (23 percent). Furthermore, in the 2002 dataset the investment to sales ratio and R&D intensity variables have about 12 percent of values missing. Given the prominence of these variables in our discussion (Section 2 above), to drop these variables would be to risk omitted variables bias in our estimates. Neither is it a satisfactory option to drop observations with missing values of these variables. For this could be justified only on the implausible assumption of values missing completely at random (MCAR); otherwise, the consequence is again inefficient and biased coefficient estimates (Schafer and Graham, 2002), arising from differences between the distribution of the missing observations and the distribution of the observed items. Because it is not satisfactory to drop either variables or observations, we imputed the missing values (the percentage of imputed values for each variable corresponds to the percentage of missing values detailed in Table 2). To this end, we applied multiple imputation (MI) as the technique most favoured in the statistical literature on analysing survey data with missing values, because it yields valid estimates from imputed datasets of the standard errors in addition to approximately unbiased estimates of all parameters.9 Accordingly, in all four datasets used for estimation – i.e., 2002, 2005, pooled, and panel - all the missing values are imputed, regardless of the number of missing values for individual variables (see Table 2 above). Consequently, the sample sizes have increased substantially in relation to the non-imputed samples: the 2002 dataset by 28 percent; the 2005 dataset by 17 percent; the pooled 2002 and 2005 dataset by 16 percent; and the two-year panel 9 We omit further discussion of MI for reasons of space. The theoretical underpinnings of MI were provided by Rubin (1978 and 1987) and later developed by Van Buuren et al. (1999), Allison (2000), Schafer and Olsen (1998), Schafer (1999), Schafer and Graham (2002), Kenward and Carpenter (2007), and Harel and Zhou (2007). For practical implementation of MI, we use the programmes written for STATA (see Royston, 2005a, 2005b, 2007; and, Carlin et al., 2008). The syntax written to implement MI for this paper is available on request. Final_1: September 10, 2008 13 SMEs and Export Performance in South East Europe (longitudinal) dataset by 41 percent.10 This large increase of the size of the dataset is reflected in the results. Although none of the coefficients change direction, in some cases they differ in terms of statistical significance and even in their magnitude. The differences in the results point to bias caused by a relatively large proportion of missing observations in the dataset; hence, to the importance of imputation techniques to address this problem. 4. RESULTS AND DISCUSSION The outcomes are in the majority of cases significant and consistent across specifications and samples, and the magnitude and direction of coefficients are mostly as anticipated. The marginal effects and their respective p-values are presented in Table 3.11 We comment in detail only on the unconditional marginal effects, because these refer to the whole population of firms (i.e. both those that may begin exporting and those that may export more), and are therefore the effects most relevant for policy discussion. For reasons of space, we discuss mainly those estimates that appear to be robust; namely, those variables whose importance is supported by coefficients with a consistent sign and at least two of which are statistically significant, except in the case of investment-sales ratio and R&D intensity for the 2002 dataset. The latter exceptions reflect the importance of these variables in our theoretical discussion. Not all the variables discussed in Section 2 have been included in the model. For 2002 we have reduced the full model to a more parsimonious one by dropping two variables related to the CEO impact on firm’s export performance (i.e. level of education and tenure) as well as the market share variable. The primary reasons for dropping these variables are that they are highly insignificant and that they do not appear in the 2005 survey. The joint statistical insignificance of the deleted variables is confirmed by a Wald test. In addition, the estimated coefficients on the other variables are robust to testing down; dropping these variables did not make any difference to the results estimated from the 2002 sample, either in terms of size or direction of effect. The latter fact is an indication of the robustness of the results, an issue to be discussed in Section 5. 10 Rubin (1987, p. 2) suggests m repeated imputations in a range of 2 to 10. However, recent research shows that in some cases a larger m is required for reliable estimation and inference (Kenward and Carpenter, 2007, p. 208), especially in cases when the proportion of missing data is high. Because the percent of missing data for some of our variables is relatively large, we apply m=100. 11 Different specifications are reported, because the three rounds of the BEEP Surveys did not contain the same set of questions. Final_1: September 10, 2008 14 SMEs and Export Performance in South East Europe The number of specification tests one can perform after multiple imputation is currently rather limited (indeed, imputation techniques in general, and multiple imputation specifically, is a developing field). Fortunately, to assess the validity of tobit estimation there is the diagnostic check suggested by Greene (2003, p. 768) and Wooldridge (2002, p. 534). As this check requires, we find that the probit and tobit coefficient estimates are consistent after appropriate transformations.12 This diagnostic check suggests that the tobit model provides consistent and unbiased estimates for the sample of seven SEE countries. The discussion of these estimates follows. Human capital related factors The human capital measures affect positively firms’ propensity to export. The higher education indicator (ftwor_edu) shows that the greater the percentage of employees with higher education the higher the expected percentage share of exports in firm’s sales. The unconditional marginal effect for the pooled sample indicates, ceteris paribus, that a one percentage increase in the pool of employees with higher education will increase the percentage share of exports in a firm’s turnover, on average, by 0.058 percent. The literature provides abundant evidence on the relationship between education (measured by mean years of schooling) and export performance (for instance Carlin et al., 2001a). Generally, the relationship is explained through the contribution of education to quality-adjusted productivity. This finding for human capital is consistent with the results suggesting that export competitiveness is associated with continuous training (training). Keeping other variables constant, the unconditional marginal effects suggest that SMEs providing on-the-job training generate, on average, up to three percent more of their total sales from exports than do SMEs that do not provide formal training. Changes in the organisational structure (org_str) do not have a consistently significant effect on export performance. Yet the consistent sign and two estimated coefficients displaying significance at, respectively, just below and just above the 10 percent level are consistent with emphasis on the effective organisation of labour as the key to firm performance (Meijaard et al., 2005). 12 These results are not reported, but are available on request. Henceforth, the same form of words may be applied to all empirical results referred to but not reported in detail. Final_1: September 10, 2008 15 SMEs and Export Performance in South East Europe Table 3. Unconditional marginal effects after tobit regression for the determinants of SME export performance in SEE countries Dependent variable: Export intensity – the percentage of total sales generated by exports Panel samplea Pooled sample 2002 sample 2005 sample % of workforce with higher education (ftwor_edu) 0.023 (0.190) 0.058*** (0.000) 0.045** (0.013) 0.071*** (0.000) Dummy: 1-if formal training provided (training) 0.714 (0.459) 2.546*** (0.002) 3.350*** (0.003) 1.673* (0.102) Highly skilled workers as a % of total (skilled) -0.007 (0.683) -0.012 (0.377) -0.008 (0.686) -0.011 (0.536) Dummy: 1-changes in the organisational structure in previous 3 years (org_str) 0.251 (0.763) 1.233* (0.098) 1.617 (0.130) 0.934 (0.338) Investment-sales ratio (gross_inv) - - -0.040 (0.614) - R&D intensity (inv_rd) - - 0.082 (0.416) - 0.140 1.615** 2.111* 0.807 (0.879) (0.052) (0.083) (0.435) 4.315** (0.022) 5.756*** (0.000) 7.560*** (0.000) 4.859*** (0.001) % of private capital in the firm (private) -0.030** (0.039) -0.00004 (0.997) -0.006 (0.651) 0.010 (0.585) % of foreign capital in the firm (foreign) 0.059*** (0.000) 0.068*** (0.000) 0.080*** (0.000) 0.060*** (0.000) % of sales generated in production (entact) 0.002 (0.863) 0.070*** (0.000) 0.056*** (0.000) 0.078*** (0.000) % of firm’s working capital and new fixed inv. financed by external sources (credit) 0.011 (0.224) 0.011 (0.165) 0.012 (0.345) 0.017* (0.075) 1.860** (0.043) 5.502*** (0.000) 4.628*** (0.000) 5.811*** (0.000) -0.353*** (0.001) 0.037 (0.499) -0.093 (0.252) 0.121* (0.095) 0.003** (0.013) -0.0003 (0.629) 0.001 (0.139) -0.001 (0.179) HUMAN RELATED FACTORS TECHNOLOGY RELATED FACTORS Dummy: 1-firm established or upgraded prod. line or intro. new tech. in previous 3 years (prli_tech) CONTROL VARIABLES Dummy: 1-medium firms (med) Dummy: 1- if member of a business association (bus_assoc) Year since establishment (age) age squared (agesq) Final_1: September 10, 2008 16 SMEs and Export Performance in South East Europe Years since it started exporting (exp_age)b 1.248*** (0.000) - - - -0.011*** (0.000) - - - Capacity utilisation of facilities/man power expressed in % (cap_util) -0.011 (0.614) -0.012 (0.516) -0.013 (0.621) 0.014 (0.503) Western Balkan region (see4) -0.433 (0.694) 1.426** (0.045) 0.678 (0.507) 1.895** (0.038) 588 436 152 0 2498 1855 643 0 1246 922 324 0 1563 1156 407 0 exp_age squared (exp_agesq) Number of observations Left censored observations Uncensored observations Right censored observations Note: p-values in brackets. Levels of statistical significance are denoted as follows: *** – 1 percent; ** – 5 percent; and * – 10 percent a There are no observations for Bosnia in the panel estimates. b Although data on export experience were available for 2005, we had to drop them from the regressions, as they were predicting perfectly the zero outcomes (i.e. non-exporters) in estimations. A 2-year panel enabled greater variation in the data, hence we have been able to include the export experience variable in the panel estimations. Firm technology The 2002 results show that the effects of the investment-sales ratio (gross_inv) and R&D intensity (inv_rd) on export performance are insignificant. Yet, the influence of technical progress on export performance is suggested by the dummy variable for whether or not the firm has “established new, upgraded a product line or introduced a new technology” in the recent past (prli_tech). The estimated coefficients are consistently positive and they are significant in both the 2002 and pooled samples. The size of the estimated coefficients suggests that firms reporting recent technical progress export around two percent more of their output than firms that do not. As we discussed above, the estimated relationship between investment and export performance is potentially flawed by endogeneity, caused by reverse causation. However, in the case of the present variable of interest, the way in which it is defined makes endogeneity less likely. From new growth theory, we hypothesise that past technical progress may influence current export intensity. However, we have no such reasons for hypothesising that current export intensity could affect past technical progress. In the 2002 survey, the question put to firms was: “Has your company undertaken any of the following initiatives since 1998?”. And in 2005: “…over the last 36 months?” In both cases, the activities captured by these questions substantially lag current export performance, our dependent variable. Final_1: September 10, 2008 17 SMEs and Export Performance in South East Europe Although implausible, given the way in which our variable of interest is defined, we investigate the matter further to clear up any doubts with respect to the exogeneity of our technology variable of interest (prli_tech). This investigation is reported in Appendix 1 and confirms our contention that the particular definition of our technology variable is such as to minimise the impact of potential endogeneity. Control variables Firm size (med) is positively related to performance in export markets. Compared to small firms, medium firms generate a larger share of their sales in export markets. Across the subsamples the unconditional marginal effects range from just over four percent additional sales generated from export activities by medium companies in 2005 to over seven percent in 2002. This is an indication that when small firms reach a certain size they may commit more fully to exporting. The share of private capital (private) seems to have a statistically insignificant effect on export intensity (except in the panel estimates where the impact is negative and significant). Although it was expected that the private capital share would have a positive effect on firm performance, the empirical evidence is relatively inconclusive on this matter. Indeed, this is consistent with some of the earlier studies, which also failed to observe the positive relationship of private ownership on firms’ behaviour.13 On the other hand, the unconditional marginal effects for the panel data indicate that a one percent increase in the foreign ownership (foreign) of the SMEs increases the percentage of export sales in total sales by 0.059 percent. This means that, for instance, a 10 percent increase in the foreign stake of the SMEs increases the percentage of sales generated by a little more half a percent. The effect is quite consistent across the various datasets. Sector of activity (entact) reveals that companies involved in production export a greater share of their turnover relative to trade and service companies. The pooled unconditional marginal effects indicate, ceteris paribus, that a one percent increase in sales generated by production activities will increase, on average, the percentage share of sales generated by export activities by 0.070 percent. 13 See for example Bevan, et al. (1999) who argue that the absence of a significant positive relationship between private ownership and performance in the early transition period (except for de novo firms) is due to significant inside ownership, a poor corporate governance framework, and the absence of an effective and functioning capital market. Final_1: September 10, 2008 18 SMEs and Export Performance in South East Europe Being a member of a business association (bus_assoc) in the SEE region appears to be of great importance in SME export performance. The panel unconditional marginal effects suggest that members of business associations typically generate two percent more of their total sales from exports than do non-members. In other samples the effect is even greater, suggesting an advantage of five percent or more. In general, the results for experience indicators are in line with predictions from theory. The firm’s general experience (age and agesq) has a significant effect on export performance only in the panel estimates and, partially, in the 2005 specification. In the latter case, only the outcome for age is significant, and that at the 10 percent level, and the direction of the effect is reversed relative to that in the panel estimates. The panel estimates on business experience are in line with the firm survivability argument: namely, that there is a lower probability of firm’s survival at the early stages of their life cycle; yet at some later stage business experience becomes conducive to the companies’ activities, including exporting. Regarding export experience, the association with firms’ export intensity reveals an inverted U-shaped relationship. Each additional year of experience in foreign markets (expage) increases the export intensity of exporting firms. Once an export market has been entered, then the sunk costs of entry have been acquired (e.g. costs of identifying potential target markets, establishing distribution networks, adapting products and services to foreign requirements and tastes, etc.). However, the longer the firm is in an export market (i.e., the more ‘experience’ is gained) the easier it becomes to increase exports to that existing market. This is a ‘learning-by-exporting’ effect, whereby the longer the experience in an export market, the greater the knowledge acquired and the better the firm becomes at exporting to that market. Moreover, experience and the resulting export ‘knowhow’ might even reduce the sunk costs of entering new markets. In sum, the greater the firm’s export experience the greater its export ‘know-how’ and the lower the sunk costs of increasing export activity in both existing and new markets. However, the quadratic effect (expagesq) between export experience and export performance means that after a certain point this advantage declines steadily. 5. ROBUSNESS OF THE RESULTS There are a number of ways in which we check the robustness of our econometric results. First, we assess the robustness of results as we move from the full to a more parsimonious Final_1: September 10, 2008 19 SMEs and Export Performance in South East Europe model; secondly, we apply a number of different specifications to different datasets; and, thirdly, we apply model diagnostic tests and other specification checks. In the first case – applying both the full and a more parsimonious model to the 2002 dataset – we find that the results are thoroughly consistent in terms of direction of the effect. Indeed, there are no noteworthy changes in magnitude. Moreover, no variable moves from statistical significance to insignificance, or vice versa, at conventionally accepted levels, although there are some slight gains in the significance of coefficients (especially for the variables ftwor_edu and prli_tech). In our analysis we utilise a range of datasets and apply different specifications (reflecting nonconformities between the 2002 and 2005 BEEP surveys). Nonetheless, the results are overwhelmingly consistent in terms of the direction of the estimated effects. Moreover, most of the coefficients are consistent across different specifications in terms of statistical significance. There are slight differences in the magnitude of the coefficients, albeit not worth dwelling upon as they do not imply any change in our policy conclusions. Regarding the diagnostics and other statistical checks, the diagnostic check suggested by Greene (2003, p.768) and Wooldridge (2002, p. 534) suggests that the tobit estimates are consistent and unbiased for the whole range of specifications that we apply here. In addition, the joint significance of the coefficients is established by the model Wald tests. 6. CONCLUSIONS This paper investigates the determinants of export performance of SMEs in seven SEE countries, using the World Bank/EBRD Business Environment and Enterprise Performance Survey (BEEPS) carried out in 2002 and 2005. We were concerned, in particular, with the impact on export performance of human capital and technology related factors. In addition, we investigated the effects on export performance of firm-size, type of sector, ownership structure, financial constraints, membership in business associations, experience-related factors, and location in SEE sub-regions. Tobit models were estimated to analyse the relationship between firms’ export performance (measured by the share of total sales generated by exports) and these potential influences. This econometric approach enabled us to differentiate between exporters and non-exporters, while including both in our investigation. Hence, we have analysed firms’ export behaviour by taking into account not only the level of export activity but also the likelihood that firms will export at all. Final_1: September 10, 2008 20 SMEs and Export Performance in South East Europe Well-established theories explaining the behaviour of SMEs in the economy are scarce, especially the literature linking SMEs with international trade. As a result, we adopted an eclectic approach to this investigation. Table 4 shows the strands of economic theory from which our research questions were derived together with the outcomes for each econometric model estimated for SEE. In addition, the final column summarises the effects of MI, by providing information on the imputed results as well as highlighting differences from the nonimputed estimates. A number of checks indicate that the results are overwhelmingly robust. In addition, we also highlight the importance of using MI in SME and other firm-level based research. Missing data are endemic to any survey analysis, hence the implementation of this technique for imputing missing values is important for obtaining valid estimates and inference. In particular, the 25.5 percent average increase in the size of our datasets made a substantial difference to the precision of the estimates (Table 4). Where non-imputed estimates were more precise, this is likely to be due to sample bias induced by non-random “missingness”. In general, non-imputed survey datasets are likely to have missing values that are not random. In this case, econometric estimates will be biased, hence invalid. This suggests that applied economic research will benefit from a more routine application of new data imputation techniques. The independent variables examined in the model for the SEE countries were, in the majority of cases, significant and consistent across specifications and samples, and the signs of the estimated coefficients are overwhelmingly as anticipated. Although the analysis is unable to explore any dynamic relationships, these results take the empirical analysis as far as is permitted by the available data, which is restricted to cross-section samples and a two-year panel. Nonetheless, our results do indicate the importance of various factors influencing firms export performance. Under the ceteris paribus condition, the tobit estimates show that human capital more than technology-related factors seems to be an important source of international competitiveness for SEE companies. Nonetheless, although companies with a larger share of educated and skilled workers export more, investments in introducing or upgrading products and technologies also promote exports. The size variable indicates that the bigger the size of the firm the larger the share of sales generated in export markets. In this context, the results indicate Final_1: September 10, 2008 21 SMEs and Export Performance in South East Europe Table 4. A summary of outcomes and motivations for SME export performance in SEE Dependent variable: Export intensity – percentage of total sales generated by export sales Motivation/Theory Expected sign Human capital resources (education; on-the-jobtraining; skilled employees; etc.) New Growth Theory + Coefficient on the education and on-the- Similar outcomes as after MI, however job-training variables positive and workforce education is significant also in highly statistically significant in the the panel estimates. 2002, 2005 and pooled dataset. Changes in organisation structure positive and significant in the pooled dataset. Technology-related factors (invest.-sales ratio; R&D intensity; introduced new or upgraded products or new technology etc.) New Growth Theory + Investment-sales ratio and R&D intensity highly insignificant. Introduction or replacement of products and old technologies positive and significant in the 2002 and pooled datasets. Similar as after MI; however the dummy for companies that introduced new or established products or new technology is insignificant for the pooled dataset. Firm size Factor endowment and transaction-cost approach +/- Positive and highly significant throughout. Similar to the results after MI. Private or state ownership Industrial economics +/- Overwhelmingly statistically insignificant, except for the negative coefficient in the panel estimates. In all cases highly statistically insignificant. Foreign capital share International economics + Highly significant and positive throughout. Insignificant in the panel estimates. Otherwise similar to those after MI. Type of activity Base-multiplier approach + Highly significant and positive, except in the panel estimates. Highly significant and positive for all datasets. Variables Final_1: September 10, 2008 Outcomes (imputed) 22 Outcomes (non-imputed) SMEs and Export Performance in South East Europe Access to external finance Financial economics + Positive but insignificant except for 2005, which is moderately significant. Insignificant just for 2002. In other cases positive and significant. Membership in bus. associations Institutional economics + Highly significant and positive throughout. Insignificant in the panel estimates. The rest are similar to those after MI. Business experience Industrial economics Quadratic relationship Mainly insignificant. U-shape relationship identified for panel data. Similar to the results after MI. Export experience International economics Quadratic relationship Inverse U-shape relationship identified for panel data. Similar to the results after MI. Capital utilisation Industrial economics Highly statistically insignificant throughout. Similar to the results after MI. Final_1: September 10, 2008 + 23 SMEs and Export Performance in South East Europe that within the group of SMEs the medium sized firms are those who are more intensive in exporting activities relative to small firms, which, as hypothesised in theory, mainly operate in the domestic market. Companies with a foreign capital share have better prospects for exports; the same applies to those companies involved in production activities, which are more active than non-production firms in foreign markets. Moreover, experience in foreign markets plays an important role in successful export engagement, albeit with diminishing returns. Regarding export experience, the results are consistent with theories suggesting that patterns of firms’ entry and exit strategies in foreign markets are affected by the presence of sunk costs. Namely, the greater the export experience the greater the export ‘know how’ and, consequently, the lower the sunk costs of additional export activity, hence the lower the barrier to entry. The results show that the availability of external finance has no statistically significant effect on exporting by SMEs in SEE. Finally, membership in a business association is positively associated with export activity. The results have demonstrated that most of the hypothesised effects are significant and have non-negligible effects. However, taken individually, the magnitude of most of the effects is rather small. This suggests that there is no single policy that will, on its own, transform the export activity of SMEs, or indeed the entire enterprise sector in SEE. Rather, a whole range of well designed and consistently implemented policies will be required to boost the growth of SME exporting and, indeed, to promote the sector more generally. Final_1: September 10, 2008 24 SMEs and Export Performance in South East Europe Appendix 1: Investigation of the potential endogeneity of the technology variable of interest Within BEEPS only one variable satisfied the testable requirements of a valid instrument for the potentially endogenous technology variable; namely, reinvested profit (reinvest_prof). The major problem with testing for endogeneity is the choice of instrument. First, this variable does not prove significant at any conventionally acceptable level of significance when added to the models either for the 2002 data (t-statistic=0.92) or for the 2005 data (tstatistic=0.64). Second, Staiger and Stock (1997) demonstrated that instrumental variables estimation and inference may be invalid if conducted with ‘weak instruments’. Accordingly, for both the 2002 and the 2005 specifications we regress the potentially endogenous variable (i.e. prli_tech) on the exogenous variables plus the instrument. In this regression, the tstatistic (robust to heteroskedasticity) measures the level of significance of the partial correlation of the instrument with the potentially endogenous variable (Wooldridge, 2002, pp. 83-86, 90-92 and 104-105). Buono and Coviello (2004, p.4) suggest that to reject the null of a weak instrument this t-statistic ‘as a rule of thumb must be greater than 3.5’. In the 2002 data, the robust t-statistic on the coefficient for the instrument is 3.94 and in the 2005 data it is 3.25. Although the t-statistic in the letter case does not exceed the benchmark one can safely argue that the null of a weak instrument can be rejected even for 2005 as it approaches very closely to 3.5. Although discussion of weak instruments is still developing and, to our knowledge, the focus of attention has been on linear models rather than on the models employed here, there is no reason to think that our investigation of the potential endogeneity of the prli_tech has been unduly impaired by weakness of the instrument. In addition, we gain confidence in our instrument not only from applying current statistical practice but also from economic interpretation. We find strong correlation between reinvested profit and the variable that identifies companies which have established new, upgraded a product line or introduced a new technology. In the presence of a perfect capital market, one would not expect such a strong correlation, because companies would be able to borrow (or issue equity) to finance their investment needs. Yet, in the presence of imperfect capital markets – indeed, of highly imperfect capital markets as expected in SEE – plagued by incomplete and asymmetric information (which lead to problems of adverse selection and moral hazard), one would expect investment to be constrained by ‘own finance’ (including, in particular, retained profit/earnings). This means that firms’ investment decisions in SEE depend heavily on their own financial sources. Theoretical underpinnings of this discussion Final_1: September 10, 2008 25 SMEs and Export Performance in South East Europe (i.e. profitability as the determinant of investment) rely on the premise that investment decisions of firms constrained with respect to external finance are dependent on their generated profit. Pugh (1998, p.98) explains that ‘to the extent that finance is a binding constraint relaxed by increased profits, we should expect a strong positive relationship between changes in profitability and changes in investment’. A relevant finding for our discussion can be found in Budina et al. (2000) who find that the investments of relatively large firms was not liquidity constrained, but that the opposite holds for smaller firms. We used our instrument in two related ways: to implement a standard test for the null of no endogeneity; and to implement instrumental variables estimation of our model. First, the testing procedure could not confirm that an endogenous relationship exists within either the 2002 or the 2005 models; both the Smith-Blundell test of exogeneity and the Wald test of exogeneity show that the models are appropriately specified with the investment variable as exogenous. Second, instrumental variables estimation confirmed this result and was thus able to add no further information.14 Of course, this investigation of potential endogenity does not establish in general that technical progress is exogeneous with respect to export performance. However, these findings are consistent with the contention that the particular definition of our technology variable is such as to minimise the impact of potential endogeneity. The results - obtained from Stata’s ivtobit estimator - are not valid because, in the absence of demonstrable endogeneity, instrumental variables estimation is not appropriate. With this caveat, the coefficients from both the 2002 and the 2005 samples were positive, much larger and, as expected, less precisely estimated than those reported in Table 3 above. 14 Final_1: September 10, 2008 26 SMEs and Export Performance in South East Europe References Acs, Z. J., Morck, R., Shaver, J. M. and Yeung, B. (1997), The Internationalisation of Small and Medium – Sized Enterprises: A Policy Perspective, Small Business Economics 9: 7 – 20. Allison, P. (2000), Multiple Imputation for Missing Data: A Cautionary Tale, Sociological Methods and Research 28: 301-309. Bartlett, W. and Prasinkar, J. (1995). Small Firms and Economic Transformation in Slovenia, Communist Economies and Economic Transformation 7 (1): 81-101. Baum, C. (2008). Stata tip 63: Modelling proportions, The Stata Journal, 8 (2): 299–303. Becchetti, L. and Rossi, P. S. R (2000), The Positive Effects of Industrial District on the Export Performance of Italian Firms, Review of Industrial Organisation 16: 53 – 68. Beck, T. (2001) Financial Development and International Trade: Is There a Link? World Bank Policy Research Working Paper No. 2608. Beck, T. (2003), Financial Dependence and International Trade, Review of International Economics 11 (2): 296-316. Beck, T., Demirgüç-Kunt, A. and Maksimovic, V. (2004), Financing Patterns around the World: Are Small Firms Different?, World Bank Working Paper. Available at: http://www.worldbank.org/research/projects/sme/ Financing%20Patterns091604.pdf Beck, T., Demirgüç-Kunt, A., Laeven, L. and Maksimovic, V. (2006), The Determinants of Financing Obstacles, Journal of International Money & Finance 25 (6): 932-952. Becker, B. and Greenberg, D. (2005), Financial Development and International Trade, University of Illinois at Urbana-Champaign, mimeo. Bellak, C. (2004), How Domestic and Foreign Firms Differ and why does it Matter?, Journal of Economic Surveys 18 (4): 483-514. Bennett, R. J. (1998), Business Associations and their Potential Contribution to the Competitiveness of SMEs, Entrepreneurship & Regional Development 10: 243-260. Bernard, B. A. and Wagner, J. (1998), Export Entry and Exit by German Firms, NBER Working Paper 6538 Bernard, B. A. and Jensen, J. B. (2004), Why Some Firms Export, The Review of Economics and Statistics 86 (2): 561-569 Bevan, A.A., Estrin, S. and Schaffer, M.E. (1999), Determinants of Enterprise Performance during Transition, Centre for Economic Reform and Transformation (CERT) Discussion Paper No. 99/03 (January). Bleaney, M. and Wakelin, K. (2002), Efficiency, Innovation, and Exports, Oxford Bulletin of Economics and Statistics 64 (3): 3-15. Bonaccorsi, A. (1992), On the relationship between firm size and export intensity, Journal of International Business Studies, 23 (4): 605-635. Brock, W. A. and Evans, D. E. (1989), Small Business Economics, Small Business Economics 1: 7 – 20. Bryan, J. (2006), Training and Performance in Small Firms, International Small Business Journal 24(6): 635660. Budina, N., Garretsen, H, and de Jong, E. (2000), Liquidity Constraints and Investment in Transition Economies: The Case of Bulgaria, Economics of Transition 8(2): 453-475 Buono I. and Coviello D. (2004) Instrumental Variables Regressions in Estimating the Effects of Institutions on Economic Development, mimeo. Available on request from ines.buono@upf.edu; or decio.coviello@upf.edu. Calof, J. L (1993), The Impact of Size on Internationalization, Journal of Small Business Management, 31 (4): 6069. Final_1: September 10, 2008 27 SMEs and Export Performance in South East Europe Carlin, J. B., Galati, J. C. and Royston, P. (2008), A new framework for managing and analyzing multiply imputed data in Stata, The Stata Journal, 8 (1): 49–67. Carlin, W., Glyn, A. and Van Reenen, J. (2001a), Export Market Performance of OECD Countries: An Empirical Examination of the Role of Cost Competitiveness, The Economic Journal 111: 128-162. Carlin, W., Fries, S., Schaffer, M. And Seabright, P. (2001b), Competition and enterprise performance in transition economies: evidence from a cross-country surey, EBRD Working Paper No. 63, June. Chevalier, A., Harmon, C., Walker, I. and Zhu, Y. (2004), Does Education Raise Productivity, or Just Reflect it?, The Economic Journal 114: 499-517. Clercq, D. D., Sapienza, H. J. and Crijns, H. (2005), The Internationalisation of Small and Medium-Sized Firms, Small Business Economics 24: 409-419. Demsetz, H. (1997), The Firm in Economic Theory: A Quiet Revolution, American Economic Review 87 (2): 426 – 429. Dex, S. and Scheibl, F. (2001), Flexible and Family-Friendly Working Arrangements in UK-Based SMEs: Business Cases, British Journal of Industrial Relations 39 (3): 411 – 431. Dex, S. and Scheibl, F. (2002), SME and Flexible Working Arrangements, Bristol: The Policy Press. EBRD (2005), Transition Report: Business in Transition, European Bank for Reconstruction and Development: London. Estrin, S., Meyer, K, E., and Bytchkova, M. (2006), Entrepreneurship in Transition Economies, in: Mark Casson et al., eds.: Oxford Handbook of Entrepreneurship, Oxford: Oxford University Press. Everett, J. and Watson, J. (1998) Small Business Failure and External Risk Factors, Small Business Economics, 11 (4): 371-390. Fujita, M., Krugman, P. R. and Venables, A. J. (1999), The Spatial Economy: Cities, Regions, and International Trade, Cambridge: The MIT Press. Futo, P., Hoggett, P., and Kallay, L. (1997), “Small Firms and Economic Transformation in Hungary”, The European Journal of Social Sciences, Vol. 10, No. 2, pp. 171-184. Gorodnichenko, Y., Svejnar, J. and Terrell, K. (2008), Globalization and Innovation in Emerging Markets, IZA Discussion Paper No. 3299 (January) Greene, W. (2003), Econometric Analysis. 5th edition. New Jersey: Prentice Hall. Grossman, G. M. and Helpman, E. (1994) Endogenous Innovation in the Theory of Growth, Journal of Economic Perspectives 8 (1): 23 – 44. Harel, O. and Zhou, X. H. (2005), Multiple Imputation – Review of Theory, Implementation and Software, Statistics in Medicine 00: 1-22. Hoshi, I., Balcerowicz, E. and Balcerowicz, L. (2003), Barriers to Entry in Early Transition, New York: Kluwer Publishers. Hutchinson, J. and Xavier, A. (2006), Comparing the Impact of Credit Constraints on the Growth of SMEs in a Transition Country with an Established Economy, Small Business Economics 27: 169 – 179. Iakovleva, T. (2005), “Entrepreneurial Orientation of Russian SMEs”, in Vinig, T. and Van Der Voort, R. (Ed.) The Emergency of Entrepreneurial Economics, Amsterdam: Elsevier. Keeble, D., Bryson, J. and Wood, P.(1991), Entrepreneurship and Flexibility in Business Services: The Rise of Small Management Consultancy and Market Research Firms in the UK, Small Business Research Centre, University of Cambridge Working Paper No. 13. Kenward, M. G. and Carpenter, J. (2007), Multiple Imputation: Current Perspectives, Statistical Methods in Medical Research 16: 199 - 218 Kim, L., Nugent, J. B. and Yhee, S. (1997), Transaction Costs and Export Channels of Small and Medium – Sized Enterprises: The Case of Korea, Contemporary Economic Policy 15: 104 – 120. Final_1: September 10, 2008 28 SMEs and Export Performance in South East Europe Kumar, N. and Siddharthan, N. S. (1994), Technology, Firm Size and Export Behaviour in Developing Countries: The Case of Indian Enterprises, The Journal of Development Studies 31 (2): 289-309. Leonidou, L. C. and Adams-Florou, A. S. (1999), Types and Sources of Export Information: Insights from Small Businesses, International Small Business Journal 17(3): 30-48. Maddala, G. (1977). Econometrics. London: McGraw-Hill. Manova, K. B. (2006), Credit Constraints in Trade: Financial Development and Export Composition. Available at SSRN: http://ssrn.com/abstract=805027. McMillan, J. and Woodruff, C. (2002), “The Central Role of Entrepreneurs in Transition Economies”, Journal of Economic Perspectives, Vol. 16, No. 3, pp. 153-170. Meijaard, J., Brand, M. J. and Mosselman, M. (2005), Organisational Structure and Performance in Dutch Small Firms, Small Business Economics 25: 83-96. Mills D. E. (1984), Demand Fluctuations and Endogenous Firm Flexibility, The Journal of Industrial Economics XXXIII: 55 – 71. Mills, D. E. and Schumann, L. (1985), Industry Structure with Fluctuating Demand, The American Economic Review 75 (4): 758 – 767. Moosa, I. (2002), Foreign Direct Investment: Theory, Evidence and Practice, New York: Palgrave Power, B. and Reid, G. C. (2005), Flexibility, Firm-Specific Turbulence and the Performance of the Long-lived Small Firm, Review of Industrial Organisation 26(4): 415-443. Pugh, G. (1998), The Profit Elasticity of Investment in West Germany and the Investment Diversion Effects of Unification, Applied Economics Letters 5: 97-99. Roberts, M. J. and Tybout, J. R. (1997), The Decision to Export to Columbia: An Empirical Model of Entry with Sunk Costs, American Economic Review 87 (4): 545-564. Royston, P. (2005a), Multiple Imputation of Missing Values: Update, The Stata Journal 5(2): 1-14. Royston, P. (2005b), Multiple Imputation of Missing Values: Update of ice, The Stata Journal 5(4): 527-536. Royston, P. (2007), Multiple Imputation of Missing Values: Further Update of ice, with an Emphasis on Interval Censoring, The Stata Journal, 7 (4): 445–464 Rubin, D. B. (1978), Multiple Imputations in Sample Surveys – A Phenomenological Bayesian Approach to Nonresponse. Proceedings of the Survey Research Methods Section, American Statistical Association 20-28. Available at: http://www.amstat.org/Sections/Srms/Proceedings. Rubin, D. B. (1987), Multiple Imputation for Nonresponse in Surveys, New York: Willey. Scase, R. (1998), The Role of Small Businesses in the Economic Transformation of Eastern Europe: Real But Relatively Unimportant?, International Small Business Journal, Vol. 16, No 1, pp 13-21. Schafer J. L. (1999), Multiple Imputation: A Primer, Statistical Methods in Medical Research 8: 3-15. Schafer J. L. and Graham J. W. (2002), Missing data: Our View of the State of the Art, Psychological Methods 7(2): 147-177. Schafer J. L. and Olsen, M. K. (1998), Multiple Imputation for Multivariate Missing-Data Problems: A Data Analyst's Perspective, Multivariate Behavioral Research 33: 545-571. Staiger, D. and Stock, J. H. (1997), Instrumental Variables Regression with Weak Instruments, Econometrica, 65: 557-586. Sterlacchini, A. (1999), Do Innovative Activities Matter to Small Firms in Non-R&D-Intensive Industries? An Application to Export Performance?, Research Policy 28: 819-832. Svejnar, J. and Commander, S. J. (2007), Do Institutions, Ownership and Competition Explain Firm Performance? Evidence from 26 Transition Countries, IZA Discussion Paper No. 2637 (February) Final_1: September 10, 2008 29 SMEs and Export Performance in South East Europe Tobin, J (1956). Estimation of Relationships for Limited Dependent Variables. Cowles Foundation Discussion Paper 3R (July). Vagliasindi, M. (2001), Competition across transition economies: an enteprise-level analysis of the main policy and structural determinants, EBRD Working Paper No. 68, December. Vagliasindi, M. (2006), Does Competition Policy Implementation Affect the Intensity of Competition?, September; http://www.ebrd.com/country/sector/econo/jrp27.pdf (accessed on 16 August 2009). van Buuren, S., Boshuizen, H. C. and Knook, D. L. (1999), Multiple Imputation of Missing Blood Pressure Covariates in Survival Analysis, Statistics in Medicine 18: 681-694. Wagner, J. (2001), A Note on the Firm Size – Export Relationship, Small Business Economics 17: 229-237. Wakelin, K. (1998), Innovation and Export Behaviour at the Firm Level. Research Policy 26: 829-841. Wood, P. (1991), Small Firms, Business Services and Flexibility, Small Business Research Centre, University of Cambridge Working Paper No. 3. Wooldridge, J. M. (2002), Econometric Analysis of Cross Section and Panel Data, Cambridge: MIT Press. Wooldridge, J. M. (2003), Introductory Econometrics: A Modern Approach, 2nd edition, Mason: South-Western Final_1: September 10, 2008 30