Foreign direct investment and its impact on real wages: evidence from Turkish micro-level data

Abstract

This paper analyzes the role of foreign direct investment (FDI) on wages, using Turkish firm-level data from 2003 to 2010, a period which coincides with significant FDI inflows both in manufacturing and service sector firms in the region. We explore the possibility of increased foreign presence translating into shifts in either labor demand or supply curves thereby resulting in changing the total wage bill or wage per worker in the host country. To empirically test this relationship we employ a dynamic specification of the wage equation. After addressing endogeneity concerns, the results reveal that foreign presence measured in terms of intra- and inter-sectoral linkages is related to higher wage bills in the host economy, hence strengthening the argument for attracting greater foreign investment to enhance labor welfare.

Keywords:

• Foreign direct investment
• real wages
• firm-level panel data
• Turkey

JEL codes:

• F21
• F23
• F63
• O33
• J23
• J30

Acknowledgements

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 Turkish Statistical office (TurkStat) for granting access to firm level data.

Notes

This paper is part of my PhD thesis titled, ‘Productivity and Wage Spillovers from Foreign Direct Investment: Evidence from Turkish Micro-level Data’ submitted to the Institute of Development Research and Development Policy, Ruhr University Bochum, Germany.

1. Aitken, Harrison, and Lipsey (1996) and Lipsey and Sjöholm (2004) find evidence in favor of the fact that foreign-owned firms pay more than their domestic counterparts even after controlling for industry and worker characteristics. Peluffo (2015) using firm level data from Uruguay also find a wage gap between local and foreign firms. The latter study however does not venture into the domain of wage spillovers initiated by increased foreign presence in the region.

2. Wage share of skilled workers is defined as the total annual wage of skilled workers relative to the wages earned by both skilled and unskilled workers.

3. The aggregate change in the wage share of skilled workers may be attributed to (a) expansion and reallocation of production toward skill-intensive sectors due to trade shifts, change in tastes or changes in economic policies; and (b) skill upgrading within industries due to technological advancements (Meschi, Taymaz, and Vivarelli 2008). The former is referred to as the between-industry while the latter is referred to as the within-industry change in demand for skilled workers.

4. The capital stock series is not available in the data-set and is constructed using the perpetual inventory method. To calculate the capital stock series the required information includes the (1) investment series, (2) initial stock, and (3) depreciation rate. The firm-level database only contains information about yearly investment. The various investment categories are aggregated to obtain the total tangible and intangible investment in a given year. The initial capital stock is calculated using the steady state approach developed by Harberger (1978) and the depreciation rate is alternatively assumed to be 10, 15, 20, and 30%.

5. To address endogeneity concerns regarding TFP, a more natural approach would have been to use system GMM while treating both and TFP as endogenous variables. However since TFP is determined by multiple factors such as FDI, trade and in-house technology development (refer to Equation (5)) this is tantamount to taking all these factors to be endogenous in our final regression, Equation (6). Doing so resulted in rapid proliferation of instruments thereby making ‘asymptotic results about the estimators and related specification tests misleading’ (Roodman 2009, 5). To contain instrument proliferation, instruments were limited to certain lags. This led to loss of efficiency in our results as standard errors increased considerably with cuts in the number of instruments. To get around this problem of instrument proliferation, we instead substituted a lagged measure of TFP in Equation (6).

6. The capital construction assumes investment comprises tangible and intangible assets, with R&D in the latter. Ideally R&D expenditure should have been included as a separate variable which influences technology and hence production levels. However given that few firms reported R&D, its inclusion as a separate variable seemed redundant.

7. Refer to Appendix 1 for definition and construction of key variables.

8. BL_man (FL_man) measures foreign presence in downstream (upstream) manufacturing sectors while BL_ser (FL_ser) measure foreign presence in downstream (upstream) service sectors.

9. Results are available on request.

10. The total wage bill includes the gross payments to personal in terms of wages, salaries, allowances, overtime payments, social contributions, etc.

11. The labor cost is a more comprehensive measure as it includes wages plus employees’ contribution and premiums.

12. The system-GMM estimation requires that the errors in the model are serially uncorrelated. However, if this assumption fails then one possibility is to add more lags of the dependent variable to eliminate any serial correlation (Cameron and Trivedi 2009, 297). In our final model we add the second lagged variable of the dependent variable for this very purpose.

13. The aforementioned research reveals that a 1% increase in backward and forward linkages translate into an average increase in total factor productivity (TFP) of 1.8 and 2.1%, respectively.

14. Fatima (2016), using Turkish firm-level data reveals that a 1% increase in foreign presence in a sector translates into an average decrease in TFP of 0.5% for firms operating in that sector.

15. This issue was addressed by Meschi, Taymaz, and Vivarelli (2008) and Srour, Taymaz, and Vivarelli (2013) by re-estimating their models and performing Sargan/Hansen test on random subsamples comprising of about one-third of the firms in TurkStat database. The results revealed the Hansen test did not reject the null, hence reassuring the validity of the chosen instruments.