The Effect of Business Support on Employment in Manufacturing: Evidence from the European Union Structural Funds in Germany, Italy and Spain

Abstract

This study investigates employment growth in the business activities supported by the European Cohesion Policy. We examine cross-industry, cross-regional variations in a sample of fourteen manufacturing industries and seventy European regions (in Germany, Italy and Spain) and take advantage of detailed European Union Structural Funds data at the regional level for the period 2000–2006. We show that business support is positively associated with higher employment growth in industries that are initially smaller and in those with higher growth opportunities. The results suggest that direct support to businesses by the European Cohesion Policy contributes to the growth process of employment in different industries. Because previous estimated effects at the aggregate level may in fact conceal large differences across industries, we conclude that our empirical analysis contributes to the understanding of how Structural Funds can affect industrial and regional development as well as adjustment paths.

Notes

1. See Molle (2007) for a detailed discussion of the objectives of the EU Cohesion Policy.

2. See for details Commission Regulation (EC) No 1685/2000 of 28 July 2000 laying down detailed rules for the implementation of Council Regulation (EC) No 1260/1999 as regards eligibility of expenditure of operations co-financed by the Structural Funds (Official Journal L 193 of 27.07.2000).

3. See, for example, Funk and Pizzati (2003) for a collection of contributions regarding SF effectiveness, and Magrini (2004) for a survey focusing on regional convergence studies.

4. About Italian subsidies granted through the regional policies, see also Cerqua and Pellegrini (2011) and Bronzini and De Blasio (2006). For empirical works in other contexts, see among others, Harris and Trainor (2005) for the effects of capital subsidies in Northern Ireland, and Daly et al. (1993) on Canada.

5. See, for example, Beutel (1995) for a description and application of an input-output model that evaluates the effects of the SF. In addition, see Treyz and Treyz (2003) for an application of the REMI model and a review of macroeconometric models that are used to evaluate the effects of the SF. For an example of analysis of the effects of SF expenditures at the industrial level without using macroeconometric models, see the work of Pereira and Andraz (2005), who conduct a VAR analysis to assess the industry effects of ERDF expenditures in Portugal. See also Lima and Cardenete (2008), who use a social accounting matrix to simulate the effects of SF on Andalusia and find overall positive effects but some degree of differences across sectors.

6. Levine (2005) reviews the literature on the real effects of financial market development and characteristics and mentions several industry-level studies.

7. Twenty-three out of the 70 regions of Germany, Italy, and Spain are classified as Objective 1 regions.

8. Data are unavailable for the Spanish regions of Ceuta and Melilla and for the Italian region of Trentino-Alto Adige. Thus, these regions have not been included in the analysis.

9. The fourteen manufacturing industries are the following: DA (manufacture of food products, beverages and tobacco); DB (manufacture of textiles and textile products); DC (manufacture of leather and leather products); DD (manufacture of wood and wood products); DE (manufacture of pulp, paper and paper products; publishing and printing); DF (manufacture of coke, refined petroleum products and nuclear fuel); DG (manufacture of chemicals, chemical products and man-made fibers); DH (manufacture of rubber and plastic products); DI (manufacture of other non-metallic mineral products); DJ (manufacture of basic metals and fabricated metal products); DK (manufacture of machinery and equipment n.e.c.); DL (manufacture of electrical and optical equipment); DM (manufacture of transport equipment) and DN (manufacturing n.e.c.).

10. Given the size of the unit of analysis (industry–region), this variable might present some outliers (Table 2), and our estimation results might be affected by extreme values of both the dependent and independent variables. We address this problem as part of our robustness checks.

11. For the background study on the construction of the SF dataset, see, http://ec.europa.eu/regional_policy/sources/docgener/evaluation/pdf/expost2006/expenditure_final.pdf.

12. As an additional test, we also examine the relationship between the total amount of SF and manufacturing employment growth (see the section about the robustness checks).

13. Firm-level data for sales are obtained from a restricted version of the Amadeus Bureau Van Dijk database (2009) for the 2000–2006 period for some of the EU-15 countries. The UK is excluded because of the lack of data availability, whereas the three countries under analysis (Germany, Italy and Spain) and other European Regional Policy recipient countries (e.g. Greece, Ireland and Portugal) are excluded to ensure the exogeneity of the computed indicator. We apply a number of filters before computing the indicator: (a) to avoid problems of double counting, we exclude from the original database the consolidated financial statements of companies that have reported both consolidated and unconsolidated financial statements; (b) we use only active firms to avoid problems that may involve mergers, acquisitions and dismissals; and (c) we eliminate observations that had an annual real growth rate of sales higher than 100% and lower than −50% (these values roughly correspond to the 98th and 2nd percentiles of the firm-year distribution), as these observations might be sources of data unreliability and/or problems similar to those that are listed in the previous point and (d) we require firms to have non-missing data for sales during the period of analysis to ensure that the indicator is not driven by missing data for particular years. We find that 37,698 firm-year observations remain (i.e. data for 6283 firms operating in the manufacturing industries of Belgium, Finland, France, Luxembourg, Netherlands and Sweden).

14. The lack of annual information on SF business support limits our estimation to cross-section analysis. However, in the section about the additional evidence on the role of growth opportunities, after some assumptions to compute a measure of annual SF business support spending, we provide some evidence with panel data techniques.

15. For a detailed discussion, see, for instance, Barro and Sala-i-Martin (2004).

16. We do not include the single term of the industry-specific indicator GO as it is perfectly collinear to industry dummies.

17. Independent variables have been re-scaled around their mean values to facilitate their interpretation and to avoid potential collinearity problems in the interaction terms. A variance inflation factor analysis has been conducted.

18. The former effect is present in most of the works in the finance and growth literature analyzing industry-level data with a specification à la Rajan and Zingales (1998). The latter result provides further evidence of β-convergence process across EU-regions; see, for example, Barro and Sala-i-Martin (2004) and Magrini (2004).

19. We control for the potential endogeneity of SF support to business activities. In fact, the SF are not randomly allocated to regions; rather, because of allocation rules, they are concentrated in those regions with lower levels of GDP per capita and higher levels of unemployment. If industry employment growth influences both the changes in regional GDP per capita and unemployment, which affect the allocation of Structural Funds, then endogeneity problems may exist. However, the estimates of differential effects across industries reduce this problem, especially for the interaction term between business support and the exogenous growth opportunities indicator. Furthermore, we use the 1999 values of region's GDP (i.e. the year before the beginning of the period of analysis) to compute our control variables (i.e. annual average SF business support to GDP, GDP per capita, and GDP). Nevertheless, we attempt to use the total amount of SF to GVA received by the region in the previous programming period (1994–1999) as an instrumental variable for the business support to GDP ratio using data from the work of Fiaschi et al. (2009). The assumption is that the lagged values of SF are correlated with the SF business support (this correlation is confirmed by the results in the first stage) and that these values are not correlated with the error term. In the last row of Table 4, we report the endogeneity test (i.e. a robust version of the Durbin–Wu–Hausman test) for each model specification that tests the null hypothesis that the suspected endogenous regressors can actually be treated as exogenous. When the null hypothesis is not rejected (in our case in all model specifications), we report the OLS estimation results. The test is conducted for 67 regions, as we do not have instrumental data for the 3 German NUTS I regions that are included in the sample.

20. Considering that we presented the EU support to business as an alternative source of financing to private credit, this result is similar to the one obtained by Fisman and Love (2007), who find that a larger amount of private credit relative to the economy's GDP has positive growth effects on industries with higher growth opportunities.

21. The inclusion of both regional and industry fixed effects in our model specification perfectly fits with the model specifications proposed by Rajan and Zingales (1998) and Fisman and Love (2007).

22. Cook's D estimate is a common procedure to take into account the influence of outlying observations when using a least squares method. We choose the value of 4/(n – k − 1), where n is the number of observations and k is the number of regressors, as the cut-off point. We do not report the estimation results of any model specification after having detected the outliers using Cook's D measure. However, the estimation results are similar and confirm the main results.

23. As a further robustness check, in column 4 of Table 5, we show the estimation results using the ratio of the total amount of SF to GDP for the 2000–2006 period. The estimation results are similar in terms of significance and direction to those results that were obtained using the expenditures in business support.

24. We thank Angel Catalina (EU Commission DG-Regio) for having provided us with the data and the technical support. This panel dataset contains annual information only on the total amount of SF for each EU region. To build an annual measure of business support we reallocate the SF business support received by each region during the period 2000–2006 as a function of the annual allocation of the total SF.

25. The standard errors are finite-sample adjusted following Windmeijer (2005).

26. Hansen's test of overidentification restrictions for the instruments' validity is reported in the lower part of Table 6. The joint null hypothesis is that the instruments are valid, i.e. uncorrelated with the error term, and that the excluded instruments are correctly excluded from the estimated equation. P-values indicate that we do not reject the null hypothesis. Autocorrelation tests show no second-order autocorrelation.