Absorptive Capacity and FDI Spillovers: Evidence from Quantile Regressions

ABSTRACT

Using Turkish firm-level data for the period 2003–2010, we look at the relationship between foreign direct investment, absorptive capacity, and spillovers at a disaggregated level, and analyze whether firms operating in different quantiles of the productivity distribution respond differently to foreign presence and changes in absorptive capacity. The results reveal that, for a given level of foreign presence, it pays to increase the absorptive capacity of firms operating in lower quantiles. When it comes to inter-industry spillovers, it is counterproductive to increase the absorptive capacity of firms already operating in higher quantiles, as this diverts resources from other productive activities.

KEYWORDS:

• Absorptive capacity
• foreign direct investment
• productivity spillovers
• quantile regression
• Turkey

Acknowledgments

I would like to acknowledge Prof. Dr. Dieter Bender and Prof. Dr. Mathias Busse for their valuable comments and suggestions. I would also like to acknowledge the Turkish Statistical office (TurkStat) for granting me access to firm-level data.

Funding

This project was supported by the Ruhr University Research School PLUS, funded by Germany’s Excellence Initiative [DFG GSC 98/3].

Notes

¹ In our analysis, absorptive capacity is taken as an inverse of technology gap; i.e., a small technology gap translates into greater absorptive capacity, whereas a large technology gap implies lower absorptive capacity.

² The study shows a differential effect of absorptive capacity on different types of FDI spillovers, with laggard firms (i.e., firms further from the technology frontier) benefiting more from forward spillovers, while firms closer to the technology frontier tend to reap more benefits from horizontal spillovers. The former trend is consistent with a “catching-up” hypothesis, while the latter is aligned with a “technology-accumulation” hypothesis.

³ Here, TFP is computed as a Solow residual; i.e., it is that portion of value-added (sales value minus material and energy expenditure) which is unexplained by the conventional factor inputs, namely labor and capital (Solow 1956, Solow 1957). This approach makes the estimation of production function a pre-requisite for the computation of firm-level. Given the well-known simultaneity problem in production function estimation, a semi-parametric approach developed by Levinsohn and Petrin(2003) is employed, which separately estimates production function for each of the two-digit industrial classifications.

⁴ Equation (1) is estimated using a fixed-effect estimation technique, which allows us to incorporate not only the firm- but also industry-, region-, and time-fixed effects. However, when employing quantile regression for panel data, Equation (1) cannot be directly estimated. Therefore, we use the approach developed by Canay (2011), which estimates Equation (1) in a two-step procedure. In the first step, firm-fixed effect—i.e., —is recovered, and in the second step a new dependent variable—i.e. —is constructed to run a pooled quantile regression. The latter estimation procedure is in accordance with the methodology developed by Koenker and Bassett (1978). Refer to page 10 for an elaboration on Canay’s approach.

⁵ Previous studies conducted by Görg and Girma (2005) and Kim (2015) use pooled quantile regression technique to derive results, thereby ignoring the information contained in the panel component of the data. These studies omit the first step, as outlined by Canay’s (2011) approach. Our article improves on the methodology adopted previously by intertwining quantile regression with panel data.

⁶ Refer to Canay (2011) for a more detailed discussion of the methodology.

⁷ Based on our own calculations.

⁸ The sector-wise distribution of FDI inflow reveals that manufacturing and service sectors received 18% and 80% of the total FDI inflows between 2003 and 2010, respectively (our own compilation from the Annual FDI reports (2006–2011) published by Undersecretariat of Treasury (2006, 2007, 2008, 2009, 2010, 2011) and OECD (2014) FDI database). This can be compared to FDI inflows of 52% and 48% during the late 1990s and early 2000s in the manufacturing and service sectors, respectively (Sayek 2007). However, even though the share of FDI inflows in the manufacturing sector has decreased, this trend does not undermine the fact that the overall magnitude of FDI inflows in this sector has increased over the years.

⁹ These results are consistent with an earlier study conducted using firm-level data for Turkey from 2003 to 2010 by Fatima (2016).

¹⁰ Refer to Levinsohn and Petrin (2003) and Petrin et al. (2004) for a detailed explanation.

¹¹ Results are available upon request.

Additional information

Funding

This project was supported by the Ruhr University Research School PLUS, funded by Germany’s Excellence Initiative [DFG GSC 98/3].