The transfer of risk taking along the supply chain

While our study and this literature focus on the influence of major customers’ financial conditions and policies on those of their suppliers, a related stream of research examines the effects of customer concentration on suppliers’ corporate practices (Kale and Shahrur 2007; Wang 2012; Itzkowitz 2013; Campello and Gao 2017; Chen et al. 2022; Wang et al. 2022). In a robustness check, our main inference continues to hold after controlling for customer concentration.

Another related line of research investigates the role of corporate governance within the supply chain. For example, Jandik and Salikhova (2023) show that suppliers’ social connections with customers affect their capital structure.

As another case in point, Samsung and LG, two biggest panel makers, have competed fiercely with each other in the global LCD industry through technological investments. Their market share, a common measure of a firm’s bargaining power, tends to be highly correlated with their technological advances (Lee and Kim 2013).

In the Online Appendix (Table OA-1), we provide additional evidence on the significant and positive relationship between risk taking and corporate performance.

There is evidence that suppliers with concentrated customer bases reduce their risk by holding less financial leverage (Kale and Shahrur 2007; Banerjee et al. 2008) and higher cash balances (Itzkowitz 2013). Apart from maintaining conservative financial policies, suppliers may also reduce their risk-taking behaviors. Recent research on the transfer of distress risk from major customers to their suppliers seems to support the precautionary view. Huang and Ren (2017) find that major customers’ bankruptcy risk has a positive impact on that of their suppliers, resulting in a positive link between the two parties’ credit ratings. Lian (2017) shows that suppliers’ probability of financial distress is positively influenced by their major customers’ distress risk.

To examine the persistence of the finding, we re-estimate Model (1) using suppliers’ risk taking from year t + 1 to year t + 5 while ensuring that the two parties maintain their customer-supplier relationships within this time horizon. The results (untabulated) show that the transfer of risk taking along the supply chain remains significantly positive until year t + 5, although both the magnitude and statistical significance of the coefficients on C_Risk decline over time. These results imply that, as long as the two sides maintain their customer-supplier relationships, the positive link between their risk-taking behaviors will persist.

As explained in Appendix 2, our matching procedure results in 30,707 major customer–supplier firm-year observations and 22,873 largest customer–supplier firm-year observations. Our baseline regressions use the latter sample with 10,875 firm-year observations that have sufficient data required for empirical analysis. In the current robustness test, we use the former sample (with 30,707 firm-year observations), which includes not only suppliers’ largest customers but also their major customers.

A potential limitation of direct measures of risk-taking activities is that they may not reflect all possible risk-taking decisions, especially those that are unobserved or those for which data are unavailable. Firms’ risk taking may take the forms of not only investments in capital expenditures, R&D, and intangible capital, but also investments in working capital and labor (human capital) (e.g., Khedmati et al. 2019). Likewise, firms that engage in risk taking may choose certain financial policies, through capital structure as well as operating leverage, financial leases, the choice of public vs. private debt and short vs. long-term debt, among others.

An important feature of our sample is that major customers, on average, are larger than their suppliers (see Table 2), suggesting that the former may have more resources and greater risk-taking capacities, such that the balance of power is skewed toward them.

For brevity, Panel A of Table 9 reports the findings when the dependent variable is σROA. The results when the dependent variable is measured as σRET or σCF are qualitatively similar; see the Online Appendix (Table OA-2) for more details.

In another (untabulated) analysis, we include all the above additional controls in our baseline model. The results are qualitatively unchanged.

The results reported in our Online Appendix (Table OA-3) show that the coefficients on the explanatory variables in the probit model are not significant after the matching. Also, the firm characteristics of the treated and control firms are generally not statistically different from one another.

In our Online Appendix (Table OA-3), we also find that the risk-taking link is statistically significant at the 1% level across all the three measures of risk taking for the propensity-score-matched sample.

In the Online Appendix (Tables OA-3 and OA-4), we further show that our main finding holds in system generalized method of moments (SYSGMM) regressions and falsification tests (pseudo analysis), which further helps alleviate endogeneity concerns.

A caveat of this analysis is that it does not examine all the possible contracting and risk management strategies firms may use. We only focus on the use of trade receivables as the data on other trade credit terms are not available in Compustat. On firm risk management, we examine diversification strategies rather than financial and contractual measures as the former, due to their longer horizon, are more comparable to risk-taking activities, such as Capex and R&D investments.