Impact of Objective 1 Funds on Regional Growth Convergence in the European Union: A Panel-data Approach

Abstract

Esposti R. and Bussoletti S. Impact of Objective 1 funds on regional growth convergence in the European Union: a panel-data approach, Regional Studies 41, 1–15. This paper investigates the impact of Objective 1 structural funds expenditure on European Union regions by estimating an augmented conditional convergence econometric model. According to this model, growth convergence is influenced by the policy treatment, which affects the regional initial investment rate by interacting with other regional structural variables and eventually influencing its steady-state level. The convergence model is specified in a dynamic panel-data form and estimated using a database of 206 EU-15 regions observed from 1989 to 2000. A generalized method of moments (GMM) estimation is applied to obtain consistent estimates of both parameter β and impact of the Objective 1 policies.

Esposti R. et Bussoletti S. L'impact des fonds Objectif 1 sur la convergence des taux de croissance dans l'Ue: une façon par enquête permanente, Regional Studies 41, 1–15. Par moyen de l'estimation d'un modèle économétrique de convergence à conditions augmentées, cet article cherche à examiner l'impact des fonds structurels Objectif 1 engagés sur les régions de l'Ue. Selon ce modèle, la convergence des taux de croissance est influée par la politique, ce qui a un effet sur le taux d'investissement initial régional en ayant une action réciproque avec d'autres variables structurelles régionales et, à terme, en influant sur son niveau continu. Le modèle de convergence se voit préciser sous forme d'une enquête permanente dynamique et estimer à partir d'une base de données auprès de 206 régions de l'Ue des 15 entre 1989 et l'an 2000. Afin d'obtenir des estimations systématiques, une estimation GMM se voit appliquer à la fois au paramètre ß et à l'impact des politiques Objectif 1.

Convergence régionale Fonds Objectif 1 Enquête permanente Estimation GMM

Esposti R. und Bussoletti S. Auswirkungen der Ziel-1-Fonds auf die regionale Wachstumskonvergenz in der EU, Regional Studies 41, 1–15. In diesem Artikel werden die Auswirkungen der Zuwendungen aus den Ziel-1-Strukturfonds für die EU-Regionen anhand eines erweiterten konditionalen Konvergenz-Ökonometriemodells untersucht. Nach diesem Modell wird die Wachstumskonvergenz durch die politische Behandlung beeinflusst, die sich auf die regionale initiale Investitionsrate auswirkt, indem sie mit anderen regionalen Strukturvariablen in Wechselwirkung tritt und schließlich das stationäre Niveau der Investitionsrate beeinflusst. Das Konvergenzmodell wird in Form dynamischer Paneldaten spezifiziert und mit Hilfe einer Datenbank von 206 EU-15-Regionen geschätzt, die zwischen 1989 und 2000 beobachtet wurden. Um sowohl vom Parameter β als auch von den Auswirkungen der Ziel-1-Politiken einheitliche Schätzungen zu erhalten, wird eine GMM-Schätzung vorgenommen.

Regionale Konvergenz Ziel-1-Fonds Paneldaten GMM-Schätzung

Esposti R. y Bussoletti S. Impacto de los fondos del Objetivo 1 en la convergencia de crecimiento regional en la UE, Regional Studies 41, 1–15. En este artículo investigamos cuál es el impacto de los gastos de fondos estructurales del Objetivo 1 en las regiones de la UE mediante un cálculo según un modelo econométrico aumentado de convergencia condicional. Según este modelo, la convergencia de crecimiento está influenciada por el tratamiento político que afecta a la tasa de inversión inicial en las regiones al interactuar con otras variables estructurales regionales y que posteriormente afecta a su nivel estable. El modelo de convergencia está especificado en tipo de datos dinámicos de panel y se ha calculado usando una base de datos de 206 regiones EU-15 que se observaron de 1989 a 2000. Hacemos un cálculo mediante el método GMM para obtener estimaciones coherentes del parámetro β y del impacto de las políticas del Objetivo 1.

Convergencia regional Fondos del Objetivo 1 Datos del panel Estimación GMM

Keywords:

• Regional convergence
• Objective 1 funds
• Panel data
• Generalized method of moments (GMM) estimation

Notes

1. Distinguishing between conditional and club convergence may be indeed difficult (Islam, 2003). Club convergence is, in fact, a particular case of conditional convergence. The latter implies a region-specific steady-state level, whereas club convergence implies multiple steady-state equilibria, one for any group (club) of regions.

2. Besides β-convergence, the alternative concept of σ-convergence has been proposed (Barro and Sala -i-Martin, 1992; Quah, 1996). Whereas the former deals with the expected value of income growth conditional on its initial value, the latter concerns its statistical distribution across regions, over time or both. β-convergence is a necessary but not sufficient condition to have σ-convergence.

3. According to the European Commission (2001, p. 131):

• transfers from the structural funds added directly to demand and economic activity, but more importantly, since they were concentrated on investment […], they were aimed at increasing growth potential in the medium and long-term. […] The estimates of the ‘supply-side’ effects on growth […] become predominant in the long-term. […] Although structural policies are ultimately judged in terms of their effect in narrowing regional disparities in GDP per head of employment, it is their impact on the underlying factors which determine economic development which is prime consideration.

4. For a more complete picture on the whole set of issues, as well as approaches, about the EU structural policy evaluation, see Bachtler and Wren (2006).

5. In particular, Rodriguez-Pose and Fratesi (2004) provide further interesting details in this respect.

6. The authors acknowledge the interesting suggestions and critical remarks made by an anonymous referee on the detailed representation of Objective 1 SF in the adopted model. The comments have been particularly helpful in improving a previous version of the model.

7. Alternatively, as clarified in the fourth section, the region-specific term may be random.

8. Since T is always positive and largely > 1, for simplicity it is assumed that when T = 0, then lnT = 0.

9. The assumption is made that s _i remains constant, if not shocked by M, and its observed initial level corresponds to the steady-state level.

10. This virtual currency converts all national currencies into the common European currency (ecu-euro) and then adjusts for the different purchasing power within the countries.

11. The current NUTS classification actually comprises 211 NUTS II regions, but, owing to data availability, Ireland and NUTS I German regions Sachsen and Sachsen-Anhalt are included as single NUTS II regions.

12. Badinger et al. (2002) assume a much higher value (0.25). In the present case, however, different values of this term do not relevantly affect the estimation results.

13. For extensive information on data sources and treatment, including the regional attribution of multi-regional funds, see Bussoletti (2004).

14. This variable can be computed only for 1999; thus, the 1999 value has been maintained over the whole period 1989–2000. This assumption seems acceptable taking into account that infrastructural indicators usually show very little variations over time.

15. As suggested by an anonymous referee, different groupings of Objective 1 regions could be proposed to interpret the mentioned differences. Nonetheless, clustering Objective 1 regions by country remains the most meaningful solution, as the country-effect still strongly conditions the regional performance. Results computed and reported by regional clusters, however, do not imply club convergence. In the conditional convergence approach here adopted, any region has its own steady-state, though groups of regions may show similarities in how conditioning variables affect it.

16. Although it is out of the 1994–99 programming period, the year 2000 is included in the sample period in order to take account of the growth effect of the last year of payments (1999).

17. Nevertheless, several empirical studies suggest that, in finite samples, the two-step estimator may actually generate little, if any, efficiency improvement while the one-step estimator may also outperform the two-step counterpart in terms of robustness (Blundell and Bond, 1998; Carmeci and Mauro, 2003; Gaduh, 2002; Judson and Owen, 1999).

18. Bond et al. (2001) obtained a 2.4% convergence rate with a GMM-SYS estimator using the same data set and model specifications as adopted by Caselli et al. (1996), who reported a 13% GMM-DIFF estimate.

19. This statistic is distributed as χ² under the null hypothesis of all instruments orthogonal to the respective error terms and with degrees of freedom given by the difference between the number of moment conditions and of unknown parameters.

20. Both statistics are distributed as standard normal under the null hypothesis of no serial correlation.