Assessing the microeconomic effects of public subsidies on the performance of firms in the czech food processing industry: A counterfactual impact evaluation

1 INTRODUCTION

The European Union (EU) allocates substantial amounts of financial resources to the European agribusiness sector every year, especially through the Common Agricultural Policy (CAP), but also through the European Regional Development Fund. The ongoing debate among the scientific community and policymakers focuses on the effects of these support programmes and deals with the question to what extent public support actually stimulates the development and competitiveness of the agribusiness sector. Therefore, there has been increased interest in conducting impact evaluations of support programmes in the last decade (e.g., Acs, Åstebro, Audretsch, & Robinson, 2016; Gagliardi & Percoco, 2017; Giua, 2017; Medeiros, 2014; Storey, 2017) and policy evaluation has become a major issue for many international organisations including the European Commission. Despite the vast amount of literature evaluating the effects of the CAP on farmers and the agricultural sector (e.g., Hansen & Herrmann, 2012; Kirchweger, Kantelhardt, & Leisch, 2015; Lankoski & Ollikainen, 2011), the effects of public support on the food processing sector in the EU have received less attention (e.g., Garcia-Alvarez-Coque, Mas-Verdu, & Sanchez García, 2015; Náglová, 2018; Schmitt, Lofredi, Berriet-Solliec, & Lepicier, 2004; Špička, 2018; Špička, Náglova, & Gurtler, 2017) and signify the challenge to agricultural economics. In this context, our study applies a counterfactual analysis approach in evaluating the EU support policy, as exemplified by the Operational Programme Enterprise and Innovation (OPEI) for the Czech Republic. The objective of OPEI was to increase the competitiveness of the Czech enterprises through public programmes facilitating financial instruments and subsidies to new and established companies. This study aims to identify adequate effects of the support programme on the participating firms within the Czech food processing industry. The counterfactual impact evaluation is defined by the European Commission (2017a) as “a method of comparison which involves comparing the outcomes of interest of those having benefitted from a policy or programme (the ‘treated group’) with those of a group similar in all respects to the treatment group (the ‘comparison/control group’), the only difference being that the comparison/control group has not been exposed to the policy or programme.”

Regardless of increasing emphasis on the counterfactual impact evaluation of public policies (e.g., Dvouletý, 2017; Dvouletý & Lukeš, 2016; Huergo & Moreno, 2017; Pellegrini & Muccigrosso, 2017; Porro & Salis, 2017), there are not many publications in outstanding agri-food economics journals on this topic and, moreover, they are, for the most part, geared toward agriculture rather than toward the food industry (e.g., Anderson, Jensen, Nelgen, & Strutt, 2016; Esposti & Sotte, 2013; Läpple & Hennessy, 2015; Mendola, 2007; Petrick & Zier, 2011; Wang, 2016). In fact, many studies conducting impact evaluations have been published in conference proceedings or as working/seminar papers (e.g., Bartova & Hurnakova, 2016; Fan, 2016; Kirchweger & Kantelhardt, 2012; Ratinger, Medonos, & Hruška, 2014).

Estimating the effects of the EU support policies subsidizing food firms with the aim to enhance their competitiveness is important for more reasons. First, the food processing industry plays a significant role in the EU economy, accounting in 2016 for 4.25 million jobs, turnover of €1,089 billion and 1.8% of EU gross value added (FoodDrink Europe, 2016). The economic and social importance of the food industry is evident also in the Czech Republic—the share of the food processing industry on the value added of the whole manufacturing industry was 7.5%, and on employment, it was 9.2% in 2016 (Czech Statistical Office, 2017). As stated by Zouaghi, Sánchez-García, and Hirsch (2017), the competitiveness of the agribusiness sector is one of the important factors of continuous economic growth, and moreover, Tong, Yu, Jensen, De La Torre Ugarte, and Cho (2016) emphasise the connection between food processing, national economy, and diet as the affordable supply of food products is crucial to the nation. Second, there is increased pressure on the application of impact evaluations of the EU support policies recently (Crowley & McCann, 2017; Dvouletý & Lukeš, 2016; or Slavík, Potluka, & Rybová, 2017), which follows also from the EU Regulation No. 1303/2013 (European Parliament & the Council of the EU, 2013) requiring the evaluations of EU support programmes “to improve the quality of the design of each programme, and verify whether its objectives and targets can be reached,” as mentioned in Article 53. Finally, the actual application of quantitative counterfactual impact evaluation methods on the evaluation of EU interventions in agribusiness is sparse, which opens the opportunity to fill this gap and also encourage other researchers to conduct empirical impact evaluations of public policies. The empirical findings have crucial implications for policymakers since they can contribute to designing of the interventions to improve their quality, and assist in the efficient allocation of resources based on scientific evidence.

Therefore, our study aims to contribute to the issue related to the effects of EU public policies in the Czech agribusiness sector. We assess the impact of the OPEI for the period 2007–2013, which targeted on the increase in the competitiveness and innovation performance of the Czech industry (European Commission, 2017b). The current study is conducted on the firm-level, and it is based on a quantitative research approach. We follow our previous research (Dvouletý & Blažková, 2017), which indicated that participation of the Czech food processing companies in the OPEI did not lead to the better financial performance of the supported enterprises. However, this study was mainly descriptive (e.g., based on t tests) and hence based on a less rigorous empirical approach. Therefore, in this study, we implement the counterfactual impact analysis (propensity score matching in combination with a difference in differences (DID) approach) to investigate the impacts of the support on the financial performance of the beneficiaries in the Czech food processing industry 2 years after the end of the intervention.

The structure of our study is as follows. First, previously published studies concerned with evaluations of the EU support programmes in the agribusiness sector are presented. Second, we described the EU public support programme under investigation, that is, the OPEI. Third, the data and methodology are introduced. This is followed by the section dealing with the results and discussion. Finally, the conclusions are drawn, with a link to several policy recommendations and suggestions for future research.

2 EMPIRICAL EVIDENCE OF THE EFFECTS OF EU AGRICULTURAL SUPPORT PROGRAMMES

In this section, we give an overview of previously published studies evaluating the effects of EU support programmes within the agribusiness sectors and summarize the existing empirical findings (Table 1). Since recent empirical research underlines the use of rigorous impact evaluation methods (e.g., Crowley & McCann, 2017; Huergo & Moreno, 2017), we focus only on studies based on the implementation of the counterfactual approach. Table 1 presents the important characteristics of each study, that is, country of the analysis, support programme, period of study, aided sectors, sample, outcome variables, particular empirical approach, and findings.

Regarding the focus of the studies, most of them evaluate the impact of the support on the economic development of the supported actors (measured by financial performance and/or employment), such as Schmitt et al. (2004), Michalek (2009), Pufahl and Weiss (2009), Salvioni and Sciulli (2011), Petrick and Zier (2011), Medonos, Ratinger, Hruška, and Špička (2012), Kirchweger and Kantelhardt (2012), Ratinger, Medonos, and Hruška (2013), and Mezera, Vilhelm, and Špička (2013). However, some of them are concerned with the impact on innovation activity, such as Czarnitzki and Lopes Bento (2010) and Nilsson (2017). Although it is generally assumed that public support should promote the performance and competitiveness of supported subjects, empirical research does not always provide consistent findings on this issue, as seen in Table 1. When looking at the productivity indicators (often used to assess performance and growth), negative impact of EU support is reported by Michalek (2009) and Pufahl and Weiss (2009), while Schmidt et al. (2004), Medonos et al. (2012), Ratinger et al. (2013), Mezera et al. (2013), and Nilsson (2017) observed positive effects. Also, the impact on employment is ambiguous—Schmidt et al. (2004) found a reduction in the unemployment rate and the number of long-term unemployed in the supported rural area as a consequence of the support. However, Salvioni and Sciulli (2011) and Petrick and Zier (2011) observed no significant impact on rural employment. As for innovation activity, Czarnitzki and Lopes Bento (2010) concentrated on innovative firms in the manufacturing industry across several EU countries and showed that supported firms would have invested significantly less in R&D and innovation if they had not received public support. Likewise, Ortner (2012), whose results are based only on a cost-benefit analysis, estimated that investment support renders private investments profitable.

With respect to previous studies in the Czech agribusiness environment, the effects of investment support have been addressed based on counterfactual analysis by Medonos et al. (2012), Ratinger et al. (2013, 2014) and Mezera et al. (2013). According to the authors, investment support brings benefits in the case of business development and improvement in productivity. However, Ratinger et al. (2013) have reported heterogeneity in impacts on the particular subsamples of businesses (divided according to the production conditions and firms size). Their study has shown that benefits are the highest on farms in less favoured areas and on medium-sized farms. On the contrary, there is according to their study a high deadweight of the investment support programme in the case of large farms. They have also highlighted the need to follow more dimensions of performance. Nevertheless, apart from Mezera et al. (2013), all Czech published studies dealt with agriculture, and in addition to that, no analysis of the effects of the OPEI in Czech agribusiness has yet been carried out.

As evident from previous studies, EU public support in the agribusiness sector has mostly positive effects on firms’ performance and investment activities. Nevertheless, the empirical results are not always unequivocal and, moreover, there is an insufficient quantity of relevant analyses. Therefore, this study contributes to this issue by employing counterfactual analysis as a rigorous approach (see, e.g., Michalek, 2009; Ratinger et al., 2013) toward the evaluation of EU public support.

3 THE SUPPORT OF THE CZECH FOOD PROCESSING INDUSTRY WITHIN THE OPEI

On the basis of the regulations adopted by the European Parliament in July 2006 (European Commission, 2006a, 2006b, 2006c, 2006d, 2006e), EU member states could draw financial support from the Structural Funds and the Cohesion Funds in the period 2007–2013. The Czech NUTS III regions, except the Capital of Prague, were included under the objective “Convergence” for the programming period of EU Structural Funds 2007–2013. Regions were allowed to use support from the EU Structural Funds (i.e., from the European Regional Development Fund and European Social Fund) and from the EU Cohesion Fund. The OPEI was created to provide exclusive support for small- and medium-sized companies (SMEs1) with the aim to increase by the end of the programming period the competitiveness of the Czech economy and to bring the innovation performance of the industry and services sectors closer to the level of leading industrial EU member states (European Commission, 2017b).

The OPEI ran during the period 2007–2013. The programme was administered by the Ministry of Industry and Trade of the Czech Republic and it was approved by the Resolution of the Government of the Czech Republic No. 392 of the year 2006 as the main programming document for the realisation of the policy for economic and social cohesion in the industry sector (Ministry of Industry & Trade of the Czech Republic, 2013). The OPEI, which follows the Operational Programme Industry and Enterprise (OPIE) implemented during years 2004–2006, ran between the years 2007–2013. Today, in the EU programming period 2014–2020, the OPEI continues as the OPEIs for Competitiveness (OPEIC). As stated in the document of the Ministry of Industry and Trade of the Czech Republic (2013), the main objective of the OPEI was “the improvement of the business environment in the Czech Republic as one of the key elements of the future successful development of the Czech economy” (Ministry of Industry & Trade of the Czech Republic, 2013, p. 6).

Since the programming period of the OPEI is already finished, the important issue is to evaluate whether the support contributed to better economic performance and efficiency of Czech enterprises and whether it increased the competitiveness of the sector and thus fulfilled its purpose.

The OPEI consisted of seven priority axes, and each of the axes was further divided into several programmes, depending on the area of support (European Commission, 2017b).

Generally, the proposals had to meet several criteria to be approved. Applicants could not have any debts toward public authorities, and they should not struggle from any financial problems. Proposals meeting the formal requirements were then assessed by the evaluation panel supervised by the Ministry of Industry and Trade. Unfortunately, there is no information available on the exact procedure and criteria for the assessment. However, we ascertained that the most important aspects of the assessment procedure included the quality of the project, its relevance to the particular priority axis, its feasibility and its cost structure.

It is worth mentioning that companies could apply for more than one project and therefore the supported companies could have been supported more than once, as can be seen in Table 2, which shows the number of supported companies and projects in particular sectors of the Czech food processing industry structured by the NACE classification. On the basis of the database of the governmental agency CzechInvest (2017) 203 food processing firms were identified participating from OPEI support with a total number of 337 supported projects, as can be seen in Table 2.

The aid programmes drawn by Czech food processing firms and the structure of supported projects are shown in Table 3. The total amount of funds from the OPEI allocated to the projects in the Czech food industry was CZK 2,379 mil,2 which was divided between a total number of 337 supported projects. Most applications were within the Development Support Programme, that is, 27% (91 projects), where the most frequent projects supported in the Czech food industry were those focused on the modernisation of production lines and equipment. As stated by Dvouletý and Blažková (2017), this shows the effort to increase the competitiveness of food processing firms in the Czech Republic through new technical facilities. From the viewpoint of the total amount of funds paid from the OPEI, the most utilized support programme was the programme Real Estate (CZK 622,271,000, i.e., 26.15% of the total funds) and the programme Innovation (CZK 602,001,000, i.e., 25.30% of the total funds), which is caused by the high financial demands in the case of innovation activities and maintenance and building restoration.

4 DATA AND EMPIRICAL APPROACH

For the programme evaluation, we implement counterfactual impact analysis, which is based on microdata obtained from the databases Albertina CZ Gold Edition (Bisnode, 2017a) and MagnusWeb (Bisnode, 2017b). The initial data set included financial data of all enterprises operating in the Czech food processing industry whose financial statements were available in the abovementioned databases. Particular food processing sectors are defined based on the three-digit level of the CZ-NACE classification (Table 2). In the case of missing data, the balance sheets and profit and loss statements of the particular companies were obtained from the websites of the Ministry of Justice of the Czech Republic (Ministry of Justice of the Czech Republic, 2017) to minimise the missing values. The obtained data set of 2,120 food processing firms was screened and firms with incomplete financial data (in the analysed periods before the programme (2005–2007) and after the programme (2014–2015)) or extreme values of financial results (such as negative values of assets and/or negative values of equity) were excluded.3 These steps resulted in the final dataset consisting of 747 firms in total, of which 143 firms were supported by the public programme and are then considered as Treated. The remaining firms are considered as the Control group. The collected financial data for the 143 supported companies account for 70% of the programme participants within the sector (based on the database of the CzechInvest, 2017, there were 203 firms within the Czech food processing industry participating from OPEI support). The distribution of the data sample is demonstrated in Table 4 showing the data set structure concerning the sector. We further report in Annex A1 and A2 the sample structure concerning region and size of the firm. As follows from Table 4, sectors with a higher number of firms participating in the OPEI (i.e., production of bakery products—CZ-NACE 107, production of other food products—CZ-NACE 108 and production of beverages CZ-NACE 110) have a higher frequency of firms in the control group (regardless of the size of the particular sector) to allow for proper matching. The largest representation of firms supported by the OPEI in these sectors may be caused by the size of these sectors (in terms of the number of firms) on the one hand, and on the other, by the fact that these sectors with a high degree of processed production belong to more profitable (as documented by Blažková & Dvouletý, 2018a), growing and up-and-coming Czech food sectors, which implies firms to be more active in taking market opportunities toward their economic growth. On the contrary, the sector CZ-NACE 101 (i.e., meat processing) with a large number of very small processors, a broad range of production and laboriousness is one of the least profitable Czech food sectors in the long run. The sector's situation is furthermore worsened due to increasing imports—while in 2003 import penetration4 ranged around 10% of domestic demand in this sector, since 2012, it has been over 40% (as reported by Blažková & Chmelíková, 2015).

For both treated and control groups, we compare the periods before the treated firms received the subsidy (2005–2007) and after the end of the programme (2014–2015). Taking into account the 2-year period after support, which is understood as a threshold by other empirical researchers (e.g., Antonioli, Marzucchi, & Montresor, 2014; Autio & Rannikko, 2016), our analysis identifies somewhat short-term effects.

Before implementing the evaluation methods, we define variables approximating the firm characteristics that the programme is likely to affect. To measure the impact of the support programme, we used variables indicating the effects on competitiveness and growth, which are stated as the main declared objectives of the OPEI programme (Ministry of Industry & Trade of the Czech Republic, 2013). Although some authors are engaged in the empirical assessment of competitiveness, there is no mutual agreement on the definition or the exact method of measuring competitiveness (Latruffe, 2010; Lerner, 2002). The assessment of competitiveness from a microeconomic point of view is usually based on traditional performance indicators (in the agri-food sector, e.g., Bezat-Jarzębowska & Rembisz, 2013; Madau, 2015). Firm-specific factors are also considered as the key sources of firm performance in agribusiness sectors (e.g., Blažková & Dvouletý, 2018b; Chaddad & Mondelli, 2013; Hirsch & Schiefer, 2016; Hirsch, Schiefer, Gschwandtner, & Hartmann, 2014). Therefore, we selected the indicators of performance based on the profit, value-added and costs that can be considered as applicable outcome variables to assess the competitiveness of firms, that is, PCM, ROA, Value Added per Labour cost and Growth of Labour Costs. As stated by Jambor and Babu (2016), financial indicators of performance provide a comprehensive picture of the competitive position of companies because companies with positive profits are able to create barriers to entry for other companies into the industry (theoretically, under the conditions of perfect competition, free entry would lead to a drop in profits to zero for all companies in the industry) and to maintain their market shares and a certain competitive advantage. As the expected outcome of the OPEI is also growth, we also used outcome variables regarded as indicators of the firms’ growth. In view of the fact that firms’ growth in the long term is not possible without innovations and investments (Geroski, 2005), as representative outcome variables we selected indicators based on the changes of assets and sales, that is, Assets Turnover, Growth of Tangible Assets, and Growth of Sales. The relationship between innovation and company growth measured by the rate of growth in sales was revealed, for example, by P. Geroski and Machin (1992), and between innovation and growth in assets by Freel (2000). In addition, the indicator Long-Run Risk was included as another outcome variable related to the increase in investments, since it reflects an increased need for funds, often covered by loans. Moreover, indebtedness is one of the key drivers of firm profitability in agribusiness sectors (i.e., it is related to the competitiveness and growth of businesses indirectly), as confirmed by recent empirical research in these sectors conducted by Gschwandtner and Hirsch (2018), Hirsch and Hartmann (2014), Hirsch et al. (2014), Blažková and Dvouletý (2018a) and Goddard, Tavakoli, and Wilson (2005).

All variables used for the analysis are defined in Table 5 and their summary statistics for the periods before and after intervention are reported in Tables 6 and 7.

5 ESTIMATION OF THE PROPENSITY SCORE

Counterfactual impact evaluation was introduced to social/economic research decades ago (Rosenbaum & Rubin, 1983). As the first step of our analysis, we need to estimate the propensity score, which is used to link the control sample of firms with the supported (Treated) sample, based on the most similar observable characteristics influencing the participation in the programme and the outcomes (Khandker, Koolwal, & Samad, 2010). Once we match both groups with different techniques (kernel matching, propensity score matching (PSM) and nearest neighbour matching), we are able to calculate the average treatment effect on the treated (ATET), informing us of the outcomes of the programme on the supported companies. For the calculation of the propensity score, we use logistic regression. The dependent variable in the regression is the dummy variable indicating whether the particular firm was supported by the scheme or not (Treated = 1). ATET is then calculated as an average difference in performance of the treated firms between the periods after and before the implementation of the programme, and at the same time, also as a difference between the supported and nonsupported groups (e.g., Abadie, Drukker, Herr, & Imbens, 2004; Angrist & Pischke, 2008; Becker & Ichino, 2002; Caliendo & Kopeinig, 2008; and Heckman, Ichimura, & Todd, 1997). Therefore, we apply propensity score matching in combination with a difference in differences (DID) approach. We utilize knowledge from previously published evaluation studies (e.g., Antonioli et al., 2014; Huergo & Moreno, 2017; and Pergelova & Angulo-Ruiz, 2014) to choose the right covariates/independent/matching variables. All calculations were made in STATA 14 software. The estimated matching logistic regression is depicted in Table 8. Selected covariates include variables that are regarded as factors determining firm profitability. Therefore, these variables enable matching firms with the same conditions of business success that basically influence the participation in the programme. On the basis of the literature focused on profitability drivers in food sectors (Chaddad & Mondelli, 2013; Goddard et al., 2005; Gschwandtner & Hirsch, 2018; Hirsch & Gschwandtner, 2013), the most important drivers of profitability can be considered firm size (in our analysis represented by the control variables Company Size and Assets) and capital structure (represented by Debt Ratio). Since there is evidence of the importance of the age of the firm as an important determinant of profitability (Demeter & Szász, 2016; Hirsch et al., 2014; Sørensen & Stuart, 2000), the covariate Year of Registration representing the age of the firm was used in the matching process. The legal form (represented by control variable denoted as Legal Form in our analysis), which influences performance through legislation (i.e. taxation and the availability of subsidies) is also considered an influential determinant of the profitability and growth of SMEs, as well as competition, as stated by Foreman-Peck, Makepeace, and Morgan (2006). The inclusion of the covariates Region and Sector is motivated by the latest studies focused on the profitability variance decomposition in the Czech and EU food processing industry conducted by Blažková and Dvouletý (2018b) and Hirsch and Schiefer (2016). In addition to these profitability determinants that have been reviewed in the food industry recently, we also include control variables representing revenues and costs (Profit/Loss, Trade Margin, and Personnel Costs), which have to be monitored when investigating profitability (Brealey, Myers, & Allen, 2017). The list of selected control variables (i.e., Year of Registration, Region, Legal Form, Company Size, Sector, and average preintervention (2005–2007) financial measures Profit/Loss, Assets, Trade Margin, Personnel Costs, and Debt Ratio) including the method of calculation is given in Table 5.

The estimated model (with the dependent variable indicating if the firm was supported or not) was found to be statistically significant (Models’ Wald χ² test p < 0.001) and the pseudo R-squared informs us that the model was able to explain 21% of the variability of the dependent variable. Our model also has some statistically insignificant variables, nevertheless, based on the previous empirical experience of other scholars (e.g., Angrist & Pischke, 2008; Antonioli et al., 2014; and Pergelova & Angulo-Ruiz, 2014), we also retain insignificant covariates in the model, because the variables help us to calculate the most precise propensity score. First of all, the year of registration (age), region and legal form of firms do not emerge as significant determinants of their participation in the programme. According to the firm size, we observe the highest likelihood of participation for medium-sized firms. With regard to the industry, lower odds of participation were observed for firms in production, processing and preservation of meat, vegetables and dairy products. The highest likelihood of participation was observed for firms focused on the manufacture of beverages (a reference category in the logistic regression). From the perspective of financial variables, it was more likely that companies with higher profits and personnel costs and with lower assets and debt ratio (see also previously reported Tables 6 and 7) were supported. This observation may indicate the strategy of picking-up (Cerulli, 2010) or retaining winners (Autio & Rannikko, 2016). Nevertheless, it may just be caused by the more heterogeneous characteristics of the control group (Hawawini, Subramanian, & Verdin, 2003).

To reduce the heterogeneity bias, we implement various matching techniques (kernel matching, propensity score matching and nearest neighbour matching). Implementation of three different matching techniques helps us to increase the robustness of our analysis (e. g. Čadil, Mirošník, & Rehák, 2017; Dvouletý, Longo, Blažková, Lukeš, & Andera, 2018; Khandker et al., 2010). It is important to mention that specifically nearest neighbour matching technique (that matches firms that are closest to each other regarding the propensity score) is different from the two previous techniques (that are based on weighted averages of almost all available firms), for details and formulas see e. g. Caliendo and Kopeinig (2008), Abadie et al. (2004) and Dehejia and Wahba (2002).

After the implementation of propensity score matching, the mean bias5 was reduced from 16.4 to 6.1, and the median bias decreased from 12.2 to 4.3. The standardized percentage bias across covariates is reported in Annex A3. In addition, the likelihood ratio test was unable to reject the null hypothesis assuming no differences between both groups based on the observable characteristics (LR χ² test p = 1.00). As a robustness check, we decided to test the differences in all eight financial outcome variables for both groups (Treated and Control), for the pre-intervention period (2005–2007), after the implementation of the matching procedures. The results are reported in Annex A4. They show that there are no statistically significant differences in terms of all eight outcome variables (please note that for the sake of simplicity, the variables Tangible Assets, Labour Costs, and Sales were tested during this period in nongrowth form), and thus both matched groups/samples (Treated and Control) in terms of these outcome measures were not statistically different. On the basis of these findings, we conclude that we were able to substantially reduce the bias between both groups; implying that both samples are now comparable and therefore we are allowed to estimate ATETs and to interpret the obtained estimates.

6 RESULTS AND DISCUSSION

Of the eight analysed indicators, we found six statistically significant average treatment effects on the treated (ATETs). ATET is quantified as a difference in differences (see, e.g., Khandker et al., 2010), that is, as a difference between both matched groups (Treated and Control) and as a difference between periods after the end of the programme (2014–2015) and before the programme had started (2005–2007)). The results are presented in Table 9. The variables representing price-cost margin (PCM), growth of tangible assets, growth of sales, value added per labour cost, and long-run risk, confirmed that, compared to the control group, firms that participated in the programme reported on an average higher values of the abovementioned indicators during the period of 2 years after participation in the programme. On the contrary, the variable representing assets turnover confirmed that, compared to the control group, firms that participated in the programme reported lower assets turnover during the period of 2 years after the end of the programme. The remaining two indicators (ROA and Growth of Labour Costs) could not prove any statistically significant impact of the programme. However, both variables indicated the positive insignificant impact of the programme.

Although the ROA indicator is considered to be the best measure of performance in many studies, the PCM indicator seems to be a more suitable measure when considering short-term changes in performance for more potential reasons. First, the main difference between these two performance measures is the fact that the PCM reflects better the changes in the operational activity of a firm than the ROA, since it is based on value added, whereas the ROA is dependent on profit. In the PCM costs directly related to production are included, that is, the PCM assesses the difference between the price of a particular production and the cost of this production, no matter at what time the production was made. The PCM can be understood as an indicator able to point out the trend of a particular production to profitability/loss, and therefore we regard it as a “tachometer of a firm's profit.” Second, supported firms invested more in tangible assets than firms in the control group (as seen from Table 9) that have led to the cost increase (i.e., depreciation and financial costs) not reflected in the PCM. And finally, investments lead to an increase in assets, which reduced the values of the ROA indicator in the case of supported firms.

Our results suggest the positive effect of the OPEI programme on the profitability of Czech food processing firms—the price-cost margin (PCM) improved significantly after the intervention and the results also indicate a positive impact on the return on assets (ROA) indicator, however, it is statistically insignificant (Table 9). Since, the PCM indicator is based on the sales and variable costs of the firm, it assesses the changes in the operational efficiency due to, for example, new technologies or product innovations (Megginson, Smart, & Lucey, 2008). Therefore, it can better reflect the effects of the evaluated support programme (support for development and innovation accounted for about 50% of supported activities, see Table 3) than the ROA indicator based on profit. Our findings are in line with the results found by Špička et al. (2017), who evaluated the effects of investment subsidies in the Czech meat processing industry through fixed-effect models.

The positive effects of the OPEI programme on the competitiveness of the firms are also supported by the use of the indicator of value added. Value added is often used as a measure of competitiveness due to the fact that the improvements in value-added indicate more effective internal processes, business organisation, productivity growth and stronger market position, as mentioned by Jones and Tilley (2003) and Capitanio, Coppola, and Pascucci (2010). Therefore, it is to be expected that investment subsidisation is positively correlated with the growth of value added, as shown, for example, in the study conducted by Bergström (2000) on the firm-level data of manufacturing firms in Sweden. For our analysis, we used the relative indicator, that is, Value Added per Labour Cost, reflecting the firms’ efficiency regarding labour, as used, for example, by Čadil et al. (2017), Špička et al. (2017), Nilsson (2017), and Michalek (2012). In our study, the supported companies manifested a statistically significant increase in value added per labour cost. Generally, the results in this context are inconsistent—while Špička et al. (2017), Nilsson (2017), and Michalek (2012) also found a positive impact of investment support on labour productivity, this has not been proven by Čadil et al. (2017).

Growth in the productivity of the supported firms in the Czech food processing industry is also evident from the results of the Growth of Labour Cost indicator, where the results suggest a positive impact of support, that is, the growth of labour cost in the group of supported enterprises. However, the effect is statistically insignificant. Although the positive sign of the coefficients indicates an increase in labour cost (as suggested by Čadil et al., 2017), it may be due to the creation of new employment opportunities), the value added per labour cost has also increased. This result can be assessed positively—despite the growth in labour costs, a statistically significant increase was proved in the value added per labour cost reflecting the productivity of labour costs.

Taking the Assets Turnover indicator into consideration, the estimates show the negative and statistically significant effects of support on the assets turnover of the companies in the treated group. It can be assumed that the significant increase in investment due to support slowed the turnover of total assets mainly due to a short period of analysis. In general, investment is considered as a long-term business with long-term returns, that is, in the first years after the investment usually not enough revenues are achieved (Brealey et al., 2017). Given that the short-term effects of the programme have been assessed, this result is logical, and it would be appropriate to monitor the impact on this indicator over a longer period of time. We may suppose that, although the decrease in the assets turnover can generally be regarded as a negative development, in the case of the short-term effects of the support this result can be considered as adequate and expected since, due to the OPEI programme, the rate of growth of tangible assets (Growth of Tangible Assets) in the supported companies has increased (Table 9). These results confirm that the programme fulfils its goal of increasing competitiveness through modernisation and investment.

On the basis of the result of the Growth of Sales indicator, we can suppose that the OPEI contributed to the firm growth of the beneficiaries, since the growth in sales reflects the short- and long-term changes in the firm and is the most common indicator to measure growth by managers and entrepreneurs (Coad & Hölzl, 2010). The positive and statistically significant effect of the support on this indicator was proven by all three matching techniques.

As mentioned earlier, the increase in investment (which is assumed to be based on the presented results) is usually accompanied by an increase in the need for funds (Brealey et al., 2017). When looking at the ATET for the Long-Run Risk indicator, we may support this claim. In the group of treated firms, the impact of the OPEI support on this outcome variable was positive and statistically significant. This indicates the increase in liabilities in the capital structure of the supported firms.

7 CONCLUSIONS

The key relevance of the European regional policy is to support economic and social cohesion by mitigating regional disparities (European Commission, 2017c). The particular strategy is to fund specific projects to contribute to the growth and competitiveness of the beneficiaries. During the current programming period of 2014–2020, the EU Cohesion Policy has received the third largest budget, that is, EUR 351.8 billion (European Commission, 2017c). However, the real empirical effects are very uneven throughout both regions and industries. Moreover, there are often only descriptive impact assessments of individual programmes available. We believe that all funded programmes and projects should be properly evaluated with the use of appropriate scientific methods. Approaches using both quantitative and qualitative research methodology should be used to assess programmes. Counterfactual impact evaluation may answer questions related to the real outcomes of the programme, demonstrating whether impact of the intervention on the selected key performance indicators (KPIs) was zero, positive, or even negative (Crowley & McCann, 2017; Dvouletý & Lukeš, 2016; Lerner, 2002; Potluka, Brůha, Špaček, & Vrbová, 2016).

In our study, we respond to this issue through the empirical analysis of the effects of EU public policy on the example of the Czech agribusiness sector. We have focused on the assessment of the impact of the particular programme of the EU Cohesion Policy—OPEI, which ran during the period of 2007–2013. We applied counterfactual impact analysis (propensity score matching in combination with a difference in differences (DID) approach) based on the data of firms with the aim of investigating the effects of this support on the financial performance of enterprises operating in the Czech food processing industry. As the main objective of the OPEI was to promote competitiveness and growth (European Commission, 2017b), we used eight indicators related to these attributes to estimate the effectiveness of the support. It should be mentioned that the high number of variables used for the evaluation of the support is, to the best of our knowledge, not usual in previously published studies on this topic, implying that this study provides evidence going beyond what has been reported by the previous work.

When looking at the effect of the OPEI support on competitiveness, that is, outcome variables based on profit, value added and labour costs, we found a positive effect on the performance of supported firms, which is, in the case of the price-cost margin and value added per labour cost, statistically significant, even with the use of various matching techniques. Our findings also suggest a positive and statistically significant impact of the OPEI on the growth of the supported firms regarding higher growth in tangible assets and higher growth in sales in comparison with the control group.

On the basis of our analysis, we can summarize that the OPEI support had a significant positive effect on the competitiveness and growth of food processing firms in the Czech Republic. The validity of the obtained results in the sector under investigation is supported by a high proportion of supported firms in the treated group (i.e., 70% of beneficiaries) and by the satisfactory number of outcome variables used for the analysis. However, the limitation lies in the short-term evaluation, which should be revisited after a longer period of time, since the effects of the investment support may emerge in the long term (e.g., Bondonio & Greenbaum, 2014; Bondonio, Biagi, & Ãk, 2016).

Although we were able to identify positive outcomes of the programme, we are unable to judge the economic effectivity of the public programme. Cost-benefit analysis should definitely become a part of the programme evaluations in the future. An interesting observation was raised by Wokoun, Kolařík, and Kolaříková (2016), who found that most of the supported entrepreneurs would realise the project/investment even without receiving a subsidy. Answering the question of “what if” is always something we cannot do retrospectively. Nevertheless, perhaps public stakeholders should follow the advice of Terjesen, Bosma, and Stam (2016) and Audretsch and Link (2017), who suggest allocating resources to elite programmes with high-growth potential only or investing to boost entrepreneurial ecosystems, from which all entrepreneurs might benefit. Our final comment is related to the need to impose a reporting duty on the supported firms. All supported entrepreneurs should be forced to regularly report their financial records together with information on the number of employees. Without this crucial step, wider implementation of the counterfactual analysis is not possible. For future empirical evaluations, public authorities should also make available a list of rejected applicants for the programme, which could represent an additional control group.