Internal Capital Markets and R&D Investment: Evidence from Korean Chaebols

ABSTRACT

This study examines how internal capital markets (ICMs) within business groups affect the financing of research and development (R&D) investments, focusing on Korean chaebols. We find that, on average, chaebol-affiliated firms exhibit lower R&D investment – cash flow sensitivity than standalone firms do. We also find that an affiliated firm’s R&D expenditure is significantly positively associated with the net amount of equity capital received from other affiliates but not significantly related to its own cash flow. These findings indicate that ICMs effectively mitigate affiliated firms’ R&D financing constraints. We further find that this financing constraint mitigation is more pronounced during financial crises and for high-tech firms. Finally, we find that, for chaebols, greater ICM effectiveness leads to better innovative performance. Overall, our study highlights the importance of ICMs for promoting corporate innovation via the reduction of R&D financing constraints.

KEYWORDS:

• R&D investment
• corporate innovation
• internal capital market
• business group
• chaebol

JEL:

• G31
• G32
• O32
• O33

Acknowledgments

We wish to thank Jiyoon Lee, Paresh Kumar Narayan (the editor), and the anonymous subject editor and reviewer for their guidance and constructive comments. Any remaining errors are ours.

Disclosure Statement

No potential conflict of interest was reported by the author(s).

Notes

1. We note that the detailed hypothesis development and related literature review are in Supplementary Document.

2. Indeed, R&D investment – cash flow sensitivity can also be interpreted as the investment opportunities embedded in the R&D projects. Previous studies have used Tobin’s Q as a measure to control for such investment opportunities. However, Tobin’s Q, as a proxy for unobservable investment opportunities, likely has a measurement error problem due to the gap between true investment opportunities and observable measures of Tobin’s Q⁠ (Erickson and Whited 2000). Therefore, following Mulier, Schoors, and Merlevede (2016), we use employment growth as an alternative to control for investment opportunities.

3. It should be noted that the Asian countries suffering the most during the 2007–2008 global financial crisis are most open to trade: Hong Kong, Singapore, Korea, and Taiwan. During this period, international investors’ appetite for risk evaporated, and the flow of capital shifted away from countries that had historically been viewed as more vulnerable (Bernanke 2009). This reversal in capital flows led to credit tightening in some of these Asian countries.

4. Our data begin in 2003 due to limitations on data regarding internal capital flows.

5. About 40% of the chaebol groups lost their status as chaebols in the aftermath of the 2008 financial crisis. However, the majority of these excluded chaebol groups retrieved their status as chaebols within a few years.

6. It should be noted that, on average, there are 5.5 listed chaebol affiliates per chaebol. However, some chaebols (SK, LG, and Samsung) indeed hold a relatively large number of affiliates. The turnover rates of the affiliates in the chaebol groups are low because chaebol affiliates rarely get delisted from the market.

7. It should be noted that we only count the number of patents that are filed and eventually granted later.

8. We control for these variables to examine how R&D investment responds to the change in internal cash flows. Indeed, controversial arguments abound in the previous research on the relationship between firm size and R&D investment. We still control for the size to purge the spurious correlation caused by the size effect. We also control for investment opportunities and external sources of finance (debt and equity).

9. For example, a possible concern is that a firm operates in an industry that starts growing rapidly. As a result, the firm receives net capital inflows from other firms in the same chaebol and increases its investment, including in R&D. R&D and ICM will be positively related not because of efficient ICMs, but simply because of the growth in that industry. Such firm- or industry-level R&D investment opportunities may determine both the amount of R&D investment and whether a firm is affiliated or standalone. Moreover, business groups benefit from political support more than standalone firms (Joe and Oh 2018; Ungson, Steers, and Park 1997), which can be the driver of the differential R&D performance between them. Accordingly, the R&D investment opportunity and political support are possible confounders and should be controlled for. However, because it is hardly observable, we control for fixed effects as their alternative proxies.

10. The sampled firms for Equation (1)(1) $R\text{amp}{D}_{it}={\beta }_{1}C{F}_{it}+{\beta }_{2}C{F}_{it}\times Chaebo{l}_i+\mathrm{\Theta }{ }^{\mathrm{\prime }}{X}_{it}+{\alpha }_i+{\gamma }_t+{\epsilon }_{it},$(1) consist of all affiliated and standalone firms, whereas those for Equations (2)(2) $R\text{amp}{D}_{it}={\beta }_{1}C{F}_{it}+{\beta }_{2}IC{M}_{it}+\mathrm{\Theta }{ }^{\mathrm{\prime }}{X}_{it}+{\alpha }_i+{\gamma }_t+{\epsilon }_{it},$(2) –(5(5) $\begin{array}{rl}& R\text{amp}{D}_{it}={\beta }_{1c_{1}k}C{F}_{it}+{\beta }_{2c_{1}k}IC{M}_{it}+{\beta }_{3c_{1}k}{I}_{c_1}\times IC{M}_{it}+\mathrm{\Theta }{ }^{\mathrm{\prime }}{X}_{it}+{\alpha }_i+{\gamma }_t+{\epsilon }_{it};\\ & R\text{amp}{D}_{it}={\beta }_{1c_{2}k}C{F}_{it}+{\beta }_{2c_{2}k}IC{M}_{it}+{\beta }_{3c_{2}k}{I}_{c_2}\times IC{M}_{it}+\mathrm{\Theta }{ }^{\mathrm{\prime }}{X}_{it}+{\alpha }_i+{\gamma }_t+{\epsilon }_{it};\\ & ⋮\\ & R\text{amp}{D}_{it}={\beta }_{1c_{N}k}C{F}_{it}+{\beta }_{2c_{N}k}IC{M}_{it}+{\beta }_{3c_{N}k}{I}_{c_N}\times IC{M}_{it}+\mathrm{\Theta }{ }^{\mathrm{\prime }}{X}_{it}+{\alpha }_i+{\gamma }_t+{\epsilon }_{it},\end{array}$(5) ) include only affiliated firms because there are no ICMs for standalone firms.

11. For comparison, we also examine the change in R&D investment – cash flow sensitivity for firms under financial crisis and for firms in high-tech industry. By doing so, we check whether our measure of financing constraint actually rise under financially constrained circumstances.

12. The industry classifications are based on two- or three-digit Korean Standard Industrial Classification (KSIC) codes. We follow the criteria for high-tech industries provided by Statistics Korea: Manufacture of pharmaceuticals, medicinal chemical and botanical products (code 20, 21); Manufacture of electronic components, computer, visual; sounding and communication equipment (code 26, 28, 29); Manufacture of medical, precision and optical instruments, watches and clocks (code 27); and Manufacture of aircraft, spacecraft and its parts (code 30, 31).

13. We use moving window statistics approach to obtain chaebol – year-level ICM effectiveness. Subsamples of year k-2 to k + 2 are used to get the ICM effectiveness for year k.

14. We do not use the value of ICM_Eff_ck if chaebol c consists of less than three sampled affiliates in year $k$, because we would be unable to obtain a reliable value of ICM effectiveness for such chaebols given the limited information available on their ICM activities.

15. The specific lag structure of instruments depends on the results for autocorrelation in residuals. If there is a serial correlation in residuals up to the k^th order, the lagged differences dated t-k and the lagged levels dated earlier than t-k are used as instruments for the difference and level equations, respectively.

16. For robustness checks, we have controlled for industry and year fixed effects instead of firm fixed effects, and indeed, the results are qualitatively the same.

17. The fact that cash flows are not significantly related to R&D expenditure after adding internal capital flows partly mitigates the reverse causality concern. If, alternatively, high R&D firms attract more equity investment by other affiliates, the correlation between their cash flow and R&D should be insignificant even without having internal capital flow as a control.

18. We have conducted a similar analysis using capital expenditure as a substitute for R&D expenditure. Because capital expenditure suffers less from the information asymmetry and uncertainty problems, the coefficient of ICM is now statistically insignificant, indicating that internal capital flow is less relevant in financing capital expenditure.

19. To further check whether it is ICMs that cause the reduction of financial constraint, we conduct difference-in-difference tests. The results are in line with Hypothesis 2, reported in Supplementary Document.

20. Some critiques say that global crisis periods should be considered in a long term to capture the lingering effects. Our result remains robust even after changing the crisis years from 2007–2008 to 2007–2010.

21. Of course, an alternative hypothesis is that ICM can systematically be correlated with other firm-specific characteristics such as size or profitability. However, when we subdivide our chaebol affiliate sample based on market capitalization and Tobin’s Q, we do not find a significant difference in ICM.