What drives total factor productivity growth? An examination of spillover effects

ABSTRACT

This study analyses the role of regional characteristics, particularly regional knowledge spillovers, as drivers of total factor productivity (TFP) growth in 190 European regions. A new modelling approach for knowledge spillovers that is different from the commonly used neighbourhood matrices is proposed. Therefore, by estimating the impact of drivers on TFP growth, findings reveal a strong direct effect of TFP spillovers alongside factors such as research and development (R&D) spending, cultural values or perceived quality of governance. However, the results also show that the openness of regional society to progress is a prerequisite for the effective use of existing knowledge.

KEYWORDS:

• regional economic development
• total factor productivity growth
• driving forces
• European regions
• knowledge spillovers

JEL:

• R1
• R11

DISCLOSURE STATEMENT

No potential conflict of interest was reported by the authors.

Notes

1. The physical capital stock is calculated by applying the perpetual inventory method. The capital stock of a specific year results from the sum of the capital stock in the previous period and the flow of gross investments. The capital stock for the initial year is calculated following Ezcurra et al. (2009) and Vogel (2015). Furthermore, these studies assume an annual depreciation rate of 6%, which is also applied herein.

2. Naturally, the impact of the driving forces on performance, understood as the growth of TFP, is not only direct but also to some extent indirect and transmitted through a variety of channels. Innovation has clearly emerged as a key element for regional economic development among these indirect channels (Akçomak & ter Weel, 2009; OECD, 2010; Pike et al., 2017). The generation of knowledge and innovations ultimately leads to efficiency gains from improved production processes and better management practices, which have a positive impact on TFP (Adler et al., 2017). Although acknowledging this, the present approach focuses only on the direct effects from the driving forces on TFP growth. However, indirect effects are implicitly captured by the applied framework.

3. The presence of spatial effects is tested with a Moran’s I-test applying a weighting matrix constructed using the inverse of the squared distance (km) between the pairs of regions, truncated at 300 km. Furthermore, the matrix is normalized using its largest eigenvalue.

4. As above, the applied spatial weighting matrix is constructed using the inverse of the squared distance (km) between the pairs of regions, truncated at 300 km and normalized using its largest eigenvalue.

5. With NUTS (Nomenclature of Territorial Units for Statistics), Eurostat classifies European regions for statistical purposes into three levels: the national level NUTS-0 and the three subnational levels NUTS-1–3. The levels of analyses of the 190 regions used herein are a combination of NUTS-1 regions: Germany (16), Estonia (1), Greece (4), France (8), Latvia (1), Lithuania (1), Luxembourg (1), Malta (1) and the UK (12) and NUTS-2 regions: Austria (9), Belgium (11), Bulgaria (6), the Czech Republic (8), Denmark (5), Spain (16), Finland (5), Hungary (7), Ireland (2), Italy (21), the Netherlands (12), Poland (16), Portugal (5), Romania (8), Slovenia (2), Slovakia (4) and Sweden (8).

6. The findings remained the same whether we took EQI scores for 2013 or a linear combination of the regional EQI scores for both dates and the national EQI scores.

7. Comparable studies reveal R² values of 0.55 in Marrocu et al. (2013) or about 0.40 in Brixy (2014).

8. Neither the model with a radius of 200 km nor the model with a radius of 400 km had a better fit. Therefore, a radius of 300 km seems to be a useful range for the detection of spatial spillovers.

9. All results are available from the authors upon request. We are grateful to the two anonymous reviewers for demanding these robustness checks.

10. All results are available from the authors upon request.