Product Recall Contagion in the Supply Chain

Abstract

Following a manufacturer's large product recall, its supplier's shareholders may perceive uncertain future demand for the supplier's products and react punitively, causing a drop in the supplier's stock return—that is, a contagion (or negative spillover). Moreover, shareholders’ information asymmetry may cause them to “screen” the supplier's information cues to determine the supplier's extent of demand uncertainty. The ideal screen is the supplier's proportion of sales revenue from the recalling manufacturer. However, not all suppliers disclose this information. Therefore, we propose that shareholders use a two-stage screening. The first screen is whether the supplier demonstrates transparency by voluntarily disclosing information about its customer portfolio. The second screen—available only to the subset of suppliers that disclose customer information—is the supplier's sales revenue from the recalling manufacturer. We used a sample of 896 U.S. public manufacturer–supplier dyads impacted by 27 large manufacturer recalls. An event study followed by cross-sectional regressions provides evidence of contagion. In addition, it reveals that the supplier's voluntary disclosure of customer information mitigates contagion, whereas revenue dependence aggravates it. Contextual (i.e., recall) variables also impact contagion. Our research study contributes to the supply-chain contagion literature, screening theory, and customer information disclosure literature. The findings inform supplier firm managers that their prior customer-related disclosures and the contextual variables can moderate contagion.

Keywords

Contagion Product Recall Supply Chain Disclosure of Customer Information

1 Introduction

Manufacturer–supplier relations are intertwined (e.g., Astvansh and Jindal, 2022; Hertzel et al., 2008). A supplier may become more prosperous because of a manufacturer's success (e.g., Li and Simcoe, 2021; Van Everdingen et al., 2009) but also suffer steep losses resulting from the manufacturer's failure (e.g., Hertzel et al., 2008; Kolay et al., 2016). For example, the manufacturer's product recall—notably a large one—can spur a sharp, near-term drop in the demand for the manufacturer's products (Borah and Tellis, 2016; Giannetti and Srinivasan, 2021; Liu and Shankar, 2015) and, by extension, forecast uncertain demand for the supplier's products. This demand uncertainty can translate into a drop in the supplier's imminent cash flow. Anticipating this uncertainty, the supplier's shareholders would likely drive down its stock price. Thus, a manufacturer's recall likely causes a contagion (or negative spillover) on the supplier's shareholder value (Fang et al., 2025). Unsurprisingly, the supplier's managers would want to know which prior customer-related disclosures may influence shareholders’ perceptions of demand uncertainty, thus mitigating or aggravating their punitive reactions. Our research answers this question.

We reason that the supplier's shareholders attempt to resolve their uncertainty about the demand for the supplier's products by “screening” for supplier-provided “cues” about its customers (Connelly et al., 2021). Thus, we rely on screening theory¹ (Stiglitz, 1975; Zhang et al., 2023) to explore a supplier's prior (i.e., in the pre-recall period) information cues that may lower its shareholders’ perceived demand uncertainty.

The ideal screen provides information about the supplier's dependence on the recalling manufacturer-customer for sales revenue. However, this information is often unavailable. Specifically, the U.S. federal law requires a U.S. public firm to disclose revenue from and the name of a customer that contributed at least 10% of the supplier's annual sales revenue (i.e., the customer is “major”). Further, the law states that the firm's disclosure of a “minor” customer's information is voluntary. Interestingly, the Financial Accounting Standards Board (FASB) states that the supplier's disclosure of major customers’ information is voluntary (Web Appendix A quotes the law and the FASB). Prior research has shown that many firms do not disclose customer information, and the U.S. Securities and Exchange Commission (SEC) has never taken disciplinary action in response to such law violations (e.g., Ellis et al., 2012). Therefore, the supplier's shareholders often cannot find information about the supplier's revenue dependence on the recalling manufacturer-customer.

We reason that the supplier's shareholders overcome the unavailability of their preferred screen by undertaking a two-stage screening. First, they check whether the supplier voluntarily disclosed customer information by going beyond legal requirements and accounting standards. The intuition is that the supplier's voluntary disclosure of customer information assuages shareholders’ perceived demand uncertainty after the manufacturer's recall, thereby mitigating their punitive reactions. Second, in cases where the supplier discloses customer information, shareholders screen the supplier's revenue dependence on the recalling manufacturer. The intuition is that the higher the supplier's dependence, the greater the shareholders’ perceived demand uncertainty and, thus, the more punitive their reactions.

Furthermore, the contextual recall variables may serve as a shareholder screen (Connelly et al., 2021; Qian et al., 2021). Prior research on shareholder reactions to automotive recalls (e.g., Eilert et al., 2017; Mukherjee et al., 2022a) has identified recall severity as the most relevant shareholder screen because it directly ties to customer demand for the recalled product and its components. Therefore, we follow this research to consider an exhaustive set of five proxies of recall severity: recall size, recall news volume, recall news sentiment, customer harm, and software (vs. non-software) recall (Astvansh and Eshghi, 2023; Astvansh et al., 2024c).

We test our conjectures in the context of 896 U.S. public manufacturer–supplier dyads affected by 28 large recalls announced by 11 manufacturers, which contracted with 46 suppliers. Following recall research in operations management (e.g., Mukherjee et al., 2022a; Thirumalai and Sinha, 2011) and marketing (e.g., Chen et al., 2009; Liu et al., 2017), we measure a supplier's shareholders reactions by the supplier's cumulative abnormal stock return (CAR) surrounding on the date of the manufacturer's recall. Next, we estimate two cross-sectional regressions. The first one regresses the supplier's CAR on whether the supplier voluntarily disclosed customer information in the year immediately preceding the recall year and on recall severity proxies. The second regression—estimated on the subsample of suppliers who disclosed their sales revenue from the recalling manufacturer—regresses the CAR on the supplier's revenue dependence (Jacobs and Singhal, 2020; Jacobs et al., 2022; Qiu et al., 2024).

The event study reports that, on average, a manufacturer-customer's large recall causes its supplier's stock returns to drop by 0.40%. Thus, evidence supports supply-chain contagion from recalls. Next, the first cross-sectional regression reports that shareholders are less punitive toward suppliers that disclosed (vs. those that did not disclose) customer information in the year before the recall year. Contextual variables—specifically, recall size, recall news volume, and recall news sentiment—are also associated with shareholder reactions. The second cross-sectional regression reports that the higher the supplier's revenue dependence on the recalling customer, the more punitive the shareholders’ reaction. Thus, the second screen also reduces shareholders’ uncertainty, albeit in an unfavorable direction for the supplier, because it increases the likelihood of a greater loss in demand.

As we elaborate in the discussion section, our findings contribute to (1) supply-chain contagion literature, (2) screening theory, and (3) customer information disclosure literature. First, we extend the supply-chain contagion literature (see Table B1 in Web Appendix B) by documenting that a supplier's prior customer-related disclosures and recall-related contextual variables can influence shareholders’ uncertainty and, by extension, their punitive reactions. These findings also inform the recall literature, which shows that a recall's consequences can propagate through the supply chain (Astvansh et al., 2024a, 2024b; Cleeren et al., 2017). Second, we contribute to screening theory by proposing a two-stage screening procedure when shareholders’ ideal screen is unavailable and demonstrating that the screens in the two stages impact their reactions asymmetrically. Third, we expand the sparse literature on how customer information disclosure affects a firm's shareholder value (Ellis et al., 2012; He et al., 2020). We show the double-edged nature of this disclosure in the context of a recall's supply-chain contagion.

Our findings advise supplier firms’ managers on managing shared supply-chain contagion. On the one hand, anticipating potential supply-chain contagion, the supplier's managers may highlight their firm's customer information disclosure to assuage shareholders’ perceptions that the manufacturer's recall may hurt the demand for the supplier's products. On the other hand, our findings alert managers to the disclosure's potential downside because it reveals revenue dependence on a recalling manufacturer whose product demand is likely to drop significantly. Further, our findings show that suppliers that decide not to disclose customer information can rely on firm- and recall-specific characteristics to mitigate supply-chain contagion.

2 Conceptual Framework

2.1 Shareholders’ Two-Stage Screening to Mitigate Demand Uncertainty

Information asymmetry “arises between those who hold information and those who could make better decisions if they had it” (Qian et al., 2021: 529). In our context, a supplier to a recalling manufacturer holds private information about its customer portfolio. This information is unavailable to the supplier's shareholders. The shareholders’ need for this information becomes salient when a manufacturer-customer issues a large recall. Prior research has shown that a large recall is likely to cause a sharp and immediate decline in the demand for (1) the recalled product (Cleeren et al., 2013; Liu and Shankar, 2015), (2) the manufacturer's non-recalled products (Giannetti and Srinivasan, 2021; Liu and Shankar, 2015), and (3) other products in the focal product category (Borah and Tellis, 2016). This evidence provides the supplier's shareholders with a reason to interpret/assume that the manufacturer's recall increases the uncertainty surrounding demand for the supplier's products. The increase in shareholders’ perceived demand uncertainty prompts them to infer that the supplier's future cash flows are at risk. The inference drives down the supplier's stock price (Chen et al., 2009; Liu et al., 2017). The customer recall's negative effect on the supplier's shareholder value is called contagion (or negative spillover). Table B1 in Web Appendix B summarizes the supply-chain contagion literature.

The supplier's shareholders may reduce their perceived demand uncertainty by screening the supplier-provided information cues about its customer portfolio, and by extension, the demand for its products (Connelly et al., 2021; Spence, 1974; Stiglitz, 1975). In our substantive context, screening refers to the supplier's shareholders using the supplier's observable cues to determine the level of demand uncertainty the supplier may face due to its manufacturer-customer's recall (Panagopoulos et al., 2018).

The ideal screen is the supplier's dependence on the recalling manufacturer-customer for sales revenue (Jacobs et al., 2022; Jacobs and Singhal, 2020; Qiu et al., 2024). The higher the supplier's revenue dependence on the recalling customer, the greater the proportion of a supplier's cash flows exposed to risk. Therefore, dependence should aggravate shareholders’ punitive reactions. However, publicly listed firms in the United States do not necessarily report their sales revenue from customers (see Web Appendix A). Therefore, the supplier's shareholders cannot rely solely on the revenue-dependence screen.

We propose that shareholders adopt a two-stage screening process to circumvent the unavailability of the revenue dependence cue. The first stage involves determining whether the focal supplier disclosed customer information in the year before the recall year. If the answer is affirmative, shareholders proceed to the second stage, using the supplier's revenue screen to assess the recalling customer's dependence on the supplier. More concretely, shareholders consider the proportion of annual sales revenue the supplier received from the recalling manufacturer-customer.

2.1.1 Stage 1: Supplier's Voluntary Disclosure of Customer Information

A supplier may voluntarily disclose proprietary information about its customer portfolio if the anticipated benefits of disclosure outweigh its anticipated costs (Ellis et al., 2012; He et al., 2020). Disclosure benefits involve reducing shareholders’ information asymmetry (Bayer et al., 2017) and, in turn, fostering favorable perceptions (Li, 2010; Ling et al., 2020). Shareholders of a disclosing supplier may screen for its voluntary disclosure of customer information and thus become aware of its alternatives to the recalling manufacturer-customer. Consequently, they can more precisely estimate the supplier's demand uncertainty triggered by the manufacturer's recall, leading them to react less punitively. However, the disclosed information is also available to the firm's rivals, who may learn about the supplier's customers and poach them, resulting in a decline in the supplier's sales revenue (Bayer et al., 2017). This plausibility leads the firm to anticipate the disclosure's proprietary costs.

We acknowledge the possibility that shareholders may interpret the disclosure as managerial overconfidence and fear that it may cause rivals to poach customers, thereby inducing greater (rather than lesser) demand uncertainty. Should this interpretation prevail, shareholders may react more punitively when the supplier voluntarily discloses customer portfolio information.

2.1.2 Stage 2: Supplier's Revenue Dependence on the Recalling Manufacturer-Customer

The contagion literature suggests that the extent to which a recall's costs propagate to a supplier is a function of the supplier's level of dependence on the recalling manufacturer-customer (Pfeffer and Salancik, 1978). The three factors of (1) importance, (2) discretion, and (3) number of alternatives, which compose resource dependence (Pfeffer and Salancik, 1978), suggest that revenue dependence is a diagnostic screen for the supplier's shareholders when a manufacturer-customer issues a large recall. First, the supplier relies on its manufacturer-customers for revenue, a critical resource that defines the supplier's ability to survive and grow (Heide and John, 1988). Second, customers have discretion over which suppliers they form relations with and whose contracts they terminate when faced with financial difficulties (Maitland et al., 1985; Wathne and Heide, 2000). Third, a customer that contributes a larger revenue share for the supplier is more difficult to replace than a smaller customer (Casciaro and Piskorski, 2005; Emerson, 1962). Thus, the greater the supplier's revenue dependence on the recalling customer, the more uncertain the demand for the supplier's products. Thus, revenue dependence may exacerbate shareholders’ punitive reactions.

2.2 Recall Context's Effect on Shareholder Reactions

Shareholders may also look up the “contextual variables” as a screen. Therefore, we consider recall severity—the most relevant recall variable—as a shareholder screen (Cleeren et al., 2017; Gao et al., 2015; Liu and Shankar, 2015).

Recall severity is multifaceted (Cleeren et al., 2017). Thus, we focus on five recall variables that shareholders may use to mitigate their perceived uncertainty. First, we examine recall size, defined as the number of affected products (Gao et al., 2015). A larger recall typically signals a broader customer reach, increasing the potential impact on brand reputation and product-market performance. Second, we examine the extent of the recall's news volume (Borah and Tellis, 2016; Liu and Shankar, 2015). News media exposure amplifies recall visibility, intensifying its deleterious effects on customers and shareholders. Third, we consider the news sentiment. Negative media portrayals amplify reputational damage, leading to more punitive shareholder reactions (Tetlock, 2007). Fourth, we also account for customer harm (Chakravarty et al., 2022), as recalls linked to bodily injury or death can significantly undermine customer trust and erode shareholder confidence. Fifth, we examine the effect of software-related defects to account for cases where the recall is relatively simple to address (e.g., upgrading) compared to nonsoftware-related defects.

3 Data and Method

3.1 Data

Measuring recalls’ contagion from a manufacturer-customer to a supplier requires an empirical setting in which manufacturers and suppliers are interdependent in the product market (Cho et al., 2021). The automotive industry meets this requirement because suppliers produce 70% of an automobile, on average (McGee, 2017), suggesting high interdependence.

An automotive supplier's shareholders may expect recalls to be frequent events (Astvansh et al., 2022a; Crouch et al., 2020; Stout, 2019). Consequently, a manufacturer's recall that affects a few automobiles will elicit little or no reaction from the supplier's shareholders (Jarrell and Peltzman, 1985). Indeed, many automobile recall studies sample “large” recalls (e.g., Gao et al., 2015; Giannetti and Srinivasan, 2021; Hoffer et al., 1988; Jarrell and Peltzman, 1985; Javadinia et al., 2023; Liu and Varki, 2021; Pupovac et al., 2022). Consistent with these precedents, we sample large recalls, defined as those affecting at least 1 million vehicles (Beattie et al., 2021). These recalls are large enough to attract the attention of the supplier's shareholders and frequent enough to create uncertainty about demand for the supplier's products.

We assembled our sample in three steps. (1) We identified large recalls announced by automobile manufacturers. (2) We found automotive suppliers listed on the major U.S. stock exchanges. (3) Not all automotive suppliers sell their parts/components to every manufacturer. Therefore, for each recall from Step #1, we matched suppliers from Step #2 to the recalling manufacturer. We describe each step next.

First, following prior event study research on automotive recalls (Astvansh and Eshghi, 2023; Liu et al., 2017; Liu and Varki, 2021), we searched Factiva, Google, The Wall Street Journal, and Automotive News, using keywords such as “product recall,” “automobiles recall,” “car recall,” “[name of 10 largest automobile manufacturers] recalled,” and “largest recalls in the automobile/car/automotive industry.” This search yielded 27 recalls affecting 1 million or more vehicles (see Table C1 in Web Appendix C) announced between 2010 and 2016.² Eleven manufacturers announced these 27 recalls. Some recalls were announced on the same day or within a few days. Following Warren and Sorescu's (2017) recommendation, we do not remove such recalls. Concretely, we treat all recalls announced within three days of one another and affecting the same supplier as a single event (as detailed subsequently, we use a three-day [−1, 1] event window), using the earliest recall as the focal event. As detailed subsequently, results are robust to excluding recall events falling within the same three-day window [−1, 1].

Second, we used several sources to identify the population of automotive suppliers. We start with the list of suppliers in the SIC code 3714 (“Motor Vehicle Parts and Accessories”) (Jacobs and Singhal, 2020). We reviewed automobile manufacturers’ websites and searched The Wall Street Journal, Financial Times, and Automotive News for news about manufacturer–supplier relations. Publicly available third-party sources, including the PricewaterhouseCoopers’ list, also indicate the largest suppliers in the automotive industry.³ We identified which suppliers are listed on the New York Stock Exchange or the Nasdaq Stock Market. This procedure led us to 46 automotive suppliers whose common stock is publicly traded in the United States.

Third, we matched a recalling manufacturer with its suppliers by reading their annual reports and websites, as well as external sources listed in Step #2. Next, we searched Factiva for the supplier's and the manufacturer's names to find mentions of their relations (e.g., “Magna Ford,” “Magna GM,” and “Magna Toyota”).⁴ After this matching step, we obtained 896 recall-specific distinct manufacturer–supplier dyads, which we use to measure a manufacturer's recall's impact on a supplier's shareholders’ reactions. The first of our two regressions uses these 896 observations to determine whether a supplier's voluntary disclosure of customer information affects shareholders’ reactions to a manufacturer-customer's recall. Not all suppliers disclose information about their sales revenue from manufacturer-customers (Jacobs and Singhal, 2020). Therefore, our second regression (after controlling for potential sample-selection bias) uses a subsample of 223 observations to test the effect of the supplier's revenue dependence on its shareholders’ reactions to the manufacturer-customer's recall. Each regression controls for firm- and recall-specific covariates.

3.2 Event Study

We use the event study method (Ba et al., 2013; Hendricks and Singhal, 2003) to measure a supplier's shareholders’ short-term reaction to a manufacturer-customer's large recall. Specifically, we calculate a supplier's abnormal stock return to a manufacturer's recall event. The abnormal return is the observed/actual return minus the expected return. We use the market model to calculate the expected return in four steps (Eshghi and Astvansh, 2024; Sorescu et al., 2017).

First, we regress the supplier s's pre-recall return on the market pre-recall return to obtain values for $\hat{α_{i}}$ and $\hat{β_{i}}$ :

\begin{aligned} R_{s, d} = α_{i} + β_{s} \times R_{m, d} + ε_{s, d}, \end{aligned}

(1)

R_s,d is the stock return on the supplier s's common stock on day d, and R_m,d is the return of a value-weighted market index m on day d. We use a period of 240 days (ranging from 250 days before the recall to 10 days before the recall; that is, [−250, −10]) to estimate the supplier's stock return (Koval et al., 2024). This window is long enough to estimate the parameters $\hat{α_{s}}$ and $\hat{β_{s}}$ . Further, it ends 10 days before the recall, allowing us to control for the event's influence on the parameter estimates.

Second, we use $\hat{α_{i}}$ and $\hat{β_{i}}$ to calculate supplier s's expected return E[R_s,d] on the day of the event:

\begin{aligned} E [R_{s, d}] = \hat{α_{s}} + \hat{β_{i}} R_{m, d} . \end{aligned}

(2)

Third, we calculate the abnormal return (AR) as the actual return on the day minus the expected return on the same day:

\begin{aligned} A R_{s, d} = R_{s, d} - E [R_{s, d}] . \end{aligned}

(3)

Fourth, the cumulative AR (CAR) in the event window [d₁, d₂] is

\begin{aligned} C A R_{s} [d_{1}, d_{2}] = \sum_{d_{1}}^{d_{2}} A R_{s, d} . \end{aligned}

(4)

3.3 Variables

We discuss below our study's outcome and explanatory variables. Table C2 in Web Appendix C lists all variables in our regression, their measures, and data sources.

Outcome variable : Supplier's shareholders’ reactions to a manufacturer's recall. Consider an automobile manufacturer m that announces a large recall r on date d in year y. Further, the supplier s supplies automotive key/components to m in year y. We are interested in measuring supplier s's shareholders’ reactions to m's recall r. As the preceding section details, we measure s's shareholder reactions by s's CAR to the recall event r. This return is computed in the event window [d₁, d₂] surrounding the recall announcement date d. Thus, estimating CAR requires researchers to select the number of days before the event (i.e., d₁) and the number of days after the event (i.e., d₂). Warren and Sorescu (2017) recommended selecting an event window that (1) minimizes confounding factors’ influence on the supplier's stock return, (2) minimizes the chances of capturing multiple recalls clustered in time, and (3) accounts for the possibility of information leakage before the event and delayed reaction to the event. Following this recommendation, we select a three-day event window [−1, 1] to measure a supplier's CAR. This window has been used in prior event studies on recalls (e.g., Davidson and Worrell, 1992; Gao et al., 2015) and OM event studies on supply chain (e.g., Jacobs and Singhal, 2020; Schmidt et al., 2020).

Explanatory variable #1 : Supplier's voluntary disclosure (of customer information). U.S. law requires a publicly traded supplier to disclose in its annual report (i.e., the Form 10-K the firm files with the SEC) the sales revenue it received from each “major” customer—that is, a customer from whom the supplier received at least 10% of its total sales revenue in the focal year. The law implies that a supplier has discretion in reporting its sales revenue from “minor” customers—customers from whom it received <10% of its total sales revenue in the focal year (Rawlins, 2008). Interestingly, the Financial Standards Accounting Board (FASB) states that the supplier “need not disclose the identity of a major customer or the amount of revenues that each segment reports from that customer” (https://fasb.org/page/ShowPdf?path = fas131.pdf&title = FAS%20131%20(AS%20ISSUED), p. 15). The inconsistency between the law and the FASB has perhaps prevented the SEC from enforcing the law (Ellis et al., 2012) (see Web Appendix A).

We leverage this voluntariness to reason that following a large recall by a focal supplier's manufacturer-customer, the supplier's shareholders examine whether the supplier voluntarily disclosed customer information, specifically, (1) major customers’ names and (2) sales revenue it received from nonmajor customers and the names of such customers. Assuming a manufacturer's recall in year y, we set the supplier's voluntary disclosure (of customer information) to 1 if it disclosed the customer information in year y − 1, and 0 otherwise.

Explanatory variable #2 : Supplier's revenue dependence is the ratio of the focal supplier s's sales revenue from the recalling manufacturer-customer in year y − 1 divided by the supplier's total sales revenue in y − 1 (Jacobs and Singhal, 2020; Jacobs et al., 2022).

Explanatory variable #3 : Recall size is the natural logarithm of the number of vehicles affected by the focal recall (e.g., Eilert et al., 2017; Gao et al., 2015).

Explanatory variable #4 : Recall news volume is the number of news articles specific to the focal recall and published in the following six news publishers: The Wall Street Journal, USA Today, The New York Times, Washington Post, New York Post, and Financial Times (Astvansh et al., 2022b; Gao et al., 2015).

Explanatory variable #5 : Recall news sentiment is the sum of the sentiment scores of recall-specific news articles published by the six news publishers listed above. We compute each news article's each word's sentiment using lexicons that rate the sentiment on −1 to +1 scale. Thus, higher values indicate more positive sentiment (Astvansh et al., 2022b). More concretely, following prior marketing (e.g., Mukherjee et al., 2022b) and OM (e.g., Gour et al., 2022) research, we use the R-package Syuzhet with the get_sentiment function, which relies on four lexicons that each capture different aspects of sentiment to ensure comprehensiveness: Syuzhet, AFINN, Bing, and NRC. Recall news sentiment is the sum of the sentiment scores of all words within the text of the articles announcing the recall. We ensure consistency across recalls by focusing on the three most important articles (ranked by Factiva).

Explanatory variable #6 : Customer harm equals 1 if the news articles report that the defect caused injuries and/or death to drivers, passengers, and the public (e.g., Eilert et al., 2017; Javadinia et al., 2023).

Explanatory variable #7 : Software recall equals 1 if the recall relates to a software issue, and 0 otherwise (Darby et al., 2023; Thirumalai and Sinha, 2011).

Covariates : Our regression controls for nine supplier-, five recall-, and two manufacturer-specific variables that may impact shareholders’ reactions to the supplier (see Table C2). Table C3 in Web Appendix C provides reasons for including these covariates.

3.3.1 Regression Specification

We estimate equations (5) and (6) below.

\begin{aligned} S u p p l i e r' s s h a r e h o l d e r s' r e a c t i o n_{s, r [- 1, 1]} \\ = α_{0} + α_{1} S u p p l i e r' s v o l u n t a r y d i s c l o s u r e_{s} \\ + α_{2} R e c a l l s i z e_{r} + α_{3} R e c a l l n e w s v o l u m e_{r} \\ + α_{4} R e c a l l n e w s s e n t i m e n t_{r} + α_{5} C u s t o m e r h a r m_{r} \\ + α_{6} S o f t w a r e r e c a l l_{r} \\ + \sum α_{7 - 15} S u p p l i e r - s p e c i f i c c o v a r i a t e s_{s} \\ + \sum α_{16 - 19} R e c a l l - s p e c i f i c c o v a r i a t e s_{r} \\ + \sum α_{20 - 22} M a n u f a c t u r e r - s p e c i f i c c o v a r i a t e s_{m} \\ + Θ_{t} + ε_{s, r}, \end{aligned}

(5)

$α_{0}$ is the intercept; $\sum α_{7 - 15}, \sum α_{16 - 19}, \sum α_{20 - 22}$ are coefficients for supplier s, recall r, and manufacturer m specific covariates, respectively; $Θ_{t}$ is a vector of dummy variables for year; and $ε_{r, s}$ is random error.

\begin{aligned} S u p p l i e r' s s h a r e h o l d e r s' r e a c t i o n_{s, r [- 1, 1]} \\ = δ_{0} + δ_{1} S u p p l i e r' s r e v e n u e d e p e n d e n c e_{s} \\ + δ_{2} R e c a l l s i z e_{r} + δ_{3} R e c a l l n e w s v o l u m e_{r} \\ + δ_{4} R e c a l l n e w s s e n t i m e n t_{r} + δ_{5} P h y s i c a l h a r m_{r} \\ + δ_{6} S o f t w a r e d e f e c t_{r} \\ + \sum δ_{7 - 15} S u p p l i e r - s p e c i f i c c o v a r i a t e s_{s} \\ + \sum δ_{16 - 19} R e c a l l - s p e c i f i c c o v a r i a t e s_{r} \\ + \sum δ_{20 - 22} M a n u f a c t u r e r - s p e c i f i c c o v a r i a t e s_{m} \\ + Θ_{t} + η_{s, r}, \end{aligned}

(6)

$δ_{0}$ is the intercept; $\sum δ_{7 - 15}, \sum δ_{16 - 19}, \sum δ_{20 - 22}$ are coefficients of supplier s, recall r, and manufacturer m level covariates, respectively; $Θ_{t}$ is a vector of dummy variables for year; and $η_{r, s}$ is the random error.

Multiple suppliers can supply to a manufacturer. Therefore, we estimate supplier-clustered standard errors. This clustering estimates coefficients that are robust to (1) within-supplier correlations (i.e., equivalent to random effects) and (2) heteroscedasticity (Eilert et al., 2017).

3.3.2 Supplier's Disclosure Decision's Potential Endogeneity

A supplier firm strategically decides whether to disclose customer information. Unobserved managerial characteristics (e.g., disclosure orientation) may be correlated with this decision and directly affect shareholders’ reactions. Omitting these characteristics makes the disclosure decision plausibly endogenous to our specification of shareholders’ reactions. We control for the disclosure decision's endogeneity using the control function (CF) method (Lu et al., 2018; Petrin and Train, 2010) because the “approach is better suited for addressing endogeneity for a non-continuous (independent) variable” (Papies et al., 2017: 589). We also present the estimates without endogeneity control.

The first stage of the CF method uses the logit model to regress the disclosure decision on covariates listed in Table C2. The first-stage logit regression also includes a variable, which is excluded from the second-stage regression (conceptually, an instrument): a binary variable that equals 1 if a prominent U.S. news publisher reported on the focal supplier's business in the year of the disclosure and 0 otherwise. We reason that our excluded variable is likely relevant—that is, it is associated with the potentially endogenous variable of disclosure decision. Shareholders lack information about a firm (a supplier, in our context). A prominent news publisher's mentions of the firm attenuate the shareholders’ information asymmetry (e.g., Liu et al., 2017; Noack et al., 2019). Therefore, all else equal, a supplier covered by a prominent news organization has less incentive to reduce its shareholders’ information asymmetry by disclosing customer information (Merton, 1987). Conversely, a supplier that lacks prominent news media coverage has more reason to disclose customer information to reduce shareholders’ information asymmetry (Tourani-Rad and Kirkby, 2005). Therefore, we expect prominent news coverage to be negatively associated with the supplier's decision to voluntarily disclose customer information. Further, we reason that our instrument meets the exclusion restriction—that is, it is uncorrelated with unobserved determinants of the shareholders’ reactions. We reason so because the supplier's prominent media coverage occurred before the recall. Thus, the efficient market hypothesis (Malkiel and Fama, 1970) suggests that the coverage is known to the supplier's shareholders and factored into its stock price before the recall announcement.

3.3.3 Potential Selection Bias in the Subsample of Suppliers That Disclosed Revenue Dependence on the Recalling Manufacturer

Only a subsample of suppliers disclosed revenue dependence on the recalling manufacturer. Therefore, the subsample may be selective, thus biasing the estimated coefficients (Wooldridge, 2010). We control for this potential bias by estimating Heckman's (1979) two-stage selection model. The first-stage model is a binary probit regression of whether the focal supplier s reported the sales revenue it received from the recalling manufacturer m in year y − 1. The second stage explains the supplier's shareholders’ reactions, including the inverse Mills ratio (λ), computed from the first-stage regression coefficients.

The first-stage regression requires an excluded variable that affects the supplier's decision variable but not the outcome variable, shareholders’ reactions. Our exclusion variable equals 1 if the supplier is headquartered in the United States or Canada, and 0 otherwise. Prior research has reasoned that the U.S. and Canadian stock markets are more mature than those of other countries. Therefore, shareholders scrutinize firms headquartered in these two countries less than they scrutinize firms headquartered elsewhere (Ling et al., 2021; Nahata et al., 2014). By extension, suppliers headquartered outside the United States and Canada have a greater incentive to disclose their customer revenue, limiting shareholders’ perceived information disadvantage and promoting market transparency (Cashman et al., 2019; Chakrabarti et al., 2009). Therefore, the supplier's headquarters location should be associated with its decision to disclose sales revenue received from a manufacturer-customer, thereby meeting the relevance criterion. Further, the headquarters’ location should not be associated with the error term of the stock returns model for two reasons. First, relocating headquarters from one country to another is a resource-demanding process. Thus, managers cannot easily make such a decision and implement it. Second, the headquarters location should not be of primary concern for shareholders during recalls. Thus, on average, a manufacturer's recall should not cause shareholders to assess the supplier's future performance based on the supplier's headquarters country. We present results with and without Heckman's correction.

4 Results

4.1 Model-Free Results

Table C4 (Web Appendix C) reports our variables’ mean and standard deviation (SD). It also reports Pearson correlation coefficients between key variables.

The average value of suppliers’ cumulative abnormal stock returns to a manufacturer-customer's recall is 0.40% on the day of the announcement (t = −3.181, p < .01). The insight is that the supplier's shareholders react punitively to a manufacturer-customer's recall. Shareholders’ reactions are negative in the following five event windows: ([0, 1], [−1, 0], [−1, 1], [−2, 2], and [−5, 5]). They are significantly different from zero for [0, 1] (−0.39, t = −2.219, p < .05) and [−1, 0] (−0.33, t = −1.858, p < .10), suggesting significant contagion (or negative spillover) from the manufacturer to its supplier. These results support the efficient market hypothesis as suppliers’ shareholders react instantaneously to new and diagnostic information. In addition, the results suggest two key findings: (1) information leakage occurs a day before the manufacturer-customer announces the recall, and (2) shareholders continue to react a day after the announcement.

4.2 Supplier's Voluntary Disclosure of Customer Information → Suppliers’ CAR to a Manufacturer's Recall

Table 1's Column I reports the estimates from the regression that assumes the supplier's customer information disclosure is exogenous. Columns II and III present estimates from the control function method, which controls for the disclosure's potential endogeneity. Column II shows that being covered (vs. not) by a prominent U.S. news publisher is negatively associated with the supplier's voluntary disclosure of customer information (Column II: b = −0.747, p < .05). This finding is consistent with intuition and logic that coverage by a prominent news publisher attenuates the supplier's incentive to attenuate shareholder's information disadvantage by disclosing customer information. Thus, our instrument likely meets the relevance criterion.

Table 1.
Supplier's voluntary disclosure of customer information → shareholders’ reactions to a manufacturer–customer's product recall.

DV: CAR_{[−1, 1]} DV: Supplier's voluntary disclosure: yes versus no DV: CAR_{[−1, 1]}

I II III

Two-stage control function

No endogeneity correction First-stage Second-stage

Supplier's voluntary disclosure: yes versus no 0.857 1.701*

(0.322) (0.531)

Recall size −0.800* 0.0225 −0.792*

(0.240) (0.271) (0.238)

Recall news volume −0.433* −0.0442 −0.432*

(0.0752) (0.108) (0.112)

Recall news sentiment −0.132* 0.0180 −0.133*

(0.0338) (0.0411) (0.0369)

Customer harm −0.662 0.133 −0.657

(0.509) (0.518) (0.465)

Software recall 2.280 −1.287 2.375*

(1.022) (1.558) (0.890)

Supplier's customer breadth −0.148 1.575* −0.219

(0.127) (0.291) (0.201)

Supplier's customer depth 0.171 −0.0491 0.181*

(0.0800) (0.0988) (0.0997)

Supplier's customer equity 3.109* 6.346* 2.766

(1.747) (2.104) (2.469)

Supplier's R&D 2.207 23.11* 0.809

(3.293) (4.285) (3.845)

Supplier's liquidity 0.409 −1.321* 0.499***

(0.163) (0.244) (0.182)

Supplier's age 0.0109* −0.104* 0.0160

(0.00594) (0.0129) (0.00662)

Supplier's profitability −0.397 12.39* −1.579

(1.321) (1.915) (1.978)

Supplier's size 0.315* −0.616* 0.357*

(0.110) (0.123) (0.117)

Supplier's gross margin −1.782 −1.950 −1.766

(1.126) (2.296) (1.316)

Recall scope: global versus US −0.796 0.0146 −0.800

(0.689) (0.609) (0.644)

Recall initiator: government versus manufacturer −0.575 −0.110 −0.573

(0.514) (0.389) (0.453)

Recall initiator: not available versus manufacturer −0.0674 0.0497 −0.0805

(0.606) (0.582) (0.610)

Clustered recall 1.104 −0.285 1.075

(0.754) (1.005) (0.838)

Additional recall 0.121 0.190 0.124

(0.492) (0.510) (0.450)

Manufacturer's gross margin −4.033 −2.280 −3.648

(2.977) (4.741) (4.237)

Region of manufacturer's headquarters: 1.594* 0.178 1.587*

Asia versus the United States (0.424) (0.508) (0.405)

Region of manufacturer's headquarters: 1.387 −0.354 1.386

Europe versus the United States (0.956) (1.068) (0.949)

Whether a prominent U.S. news publisher covered the supplier (excluded variable for Heckman's first-stage regression) −0.747

(0.302)

Control function −0.442

(0.184)

Constant −1.677* −3.206* −1.866***

(0.480) (0.671) (0.559)

Observations 896 896 896

Year FE Yes Yes Yes

Adjusted (pseudo) R² 0.071 0.484 0.073

	DV: CAR_{[−1, 1]}	DV: Supplier's voluntary disclosure: yes versus no	DV: CAR_{[−1, 1]}
Supplier's voluntary disclosure: yes versus no	0.857**		1.701***
	(0.322)		(0.531)
Recall size	−0.800***	0.0225	−0.792***
	(0.240)	(0.271)	(0.238)
Recall news volume	−0.433***	−0.0442	−0.432***
	(0.0752)	(0.108)	(0.112)
Recall news sentiment	−0.132***	0.0180	−0.133***
	(0.0338)	(0.0411)	(0.0369)
Customer harm	−0.662	0.133	−0.657
	(0.509)	(0.518)	(0.465)
Software recall	2.280**	−1.287	2.375***
	(1.022)	(1.558)	(0.890)
Supplier's customer breadth	−0.148	1.575***	−0.219
	(0.127)	(0.291)	(0.201)
Supplier's customer depth	0.171**	−0.0491	0.181*
	(0.0800)	(0.0988)	(0.0997)
Supplier's customer equity	3.109*	6.346***	2.766
	(1.747)	(2.104)	(2.469)
Supplier's R&D	2.207	23.11***	0.809
	(3.293)	(4.285)	(3.845)
Supplier's liquidity	0.409**	−1.321***	0.499***
	(0.163)	(0.244)	(0.182)
Supplier's age	0.0109*	−0.104***	0.0160**
	(0.00594)	(0.0129)	(0.00662)
Supplier's profitability	−0.397	12.39***	−1.579
	(1.321)	(1.915)	(1.978)
Supplier's size	0.315***	−0.616***	0.357***
	(0.110)	(0.123)	(0.117)
Supplier's gross margin	−1.782	−1.950	−1.766
	(1.126)	(2.296)	(1.316)
Recall scope: global versus US	−0.796	0.0146	−0.800
	(0.689)	(0.609)	(0.644)
Recall initiator: government versus manufacturer	−0.575	−0.110	−0.573
	(0.514)	(0.389)	(0.453)
Recall initiator: not available versus manufacturer	−0.0674	0.0497	−0.0805
	(0.606)	(0.582)	(0.610)
Clustered recall	1.104	−0.285	1.075
	(0.754)	(1.005)	(0.838)
Additional recall	0.121	0.190	0.124
	(0.492)	(0.510)	(0.450)
Manufacturer's gross margin	−4.033	−2.280	−3.648
	(2.977)	(4.741)	(4.237)
Region of manufacturer's headquarters:	1.594***	0.178	1.587***
Asia versus the United States	(0.424)	(0.508)	(0.405)
Region of manufacturer's headquarters:	1.387	−0.354	1.386
Europe versus the United States	(0.956)	(1.068)	(0.949)
Whether a prominent U.S. news publisher covered the supplier (excluded variable for Heckman's first-stage regression)		−0.747**
		(0.302)
Control function			−0.442**
			(0.184)
Constant	−1.677***	−3.206***	−1.866***
	(0.480)	(0.671)	(0.559)
Observations	896	896	896
Year FE	Yes	Yes	Yes
Adjusted (pseudo) R²	0.071	0.484	0.073

Notes: Standard errors (SEs) are reported in parentheses. The regression for Column I uses SEs clustered by suppliers. The regressions for Columns II and III (i.e., control function) use SEs bootstrapped 500 times. CAR = cumulative abnormal stock return; FE = fixed effects.

***p < .01, **p < .05, *p < .1.

Columns I and III show that the supplier's voluntary disclosure positive affects CAR_[−1,1] (Column I: $b_{S u p p l i e r' s v o l u n t a r y d i s c l o s u r e} = 0.857$ , p < .05; Column III: $b_{S u p p l i e r' s v o l u n t a r y d i s c l o s u r e} = 1.701$ , p < .01). In other words, disclosure attenuates shareholders’ punitive reactions (i.e., as we reported earlier, a manufacturer recall lowers its supplier's shareholder value by 0.40% on the day of the recall, on average). The finding is consistent with our logic that the supplier's voluntary disclosure of customer information serves as shareholders’ screen for the recall's implications for the supplier's short-term cash flows.

Next, we focus on five recall-specific regressors that proxy for recall severity and may thus serve as a shareholder screen. The coefficient estimates in Columns I and III carry the same sign and are similar in magnitude. Therefore, we report the estimates for Column III. Consistent with our intuition and prior research (e.g., Liu et al., 2017), recall size (b_Recall size = b = −.792, p < .01) and recall news volume (b_{Recall news volume} = −.432, p < .01) are negatively associated with shareholder reactions. However, contrary to our expectations and prior research (e.g., Noack et al., 2019), recall news sentiment (b_{Recalls news sentiment} = −.133, p < .01) is negatively associated with shareholder reactions. In other words, the more positive the news, the more negative the shareholder reaction. We turned to signaling theory and recall literature to make sense of the seemingly counterintuitive negative sign. We found two explanations. Signaling theory suggests that signal inconsistency—for example, media using positive language to convey negative news—can confuse shareholders, increasing uncertainty and punitive reactions (Connelly et al., 2021). Product recall literature has revealed that news publishers may use milder language when reporting large recalls, thus maintaining relations with their corporate advertisers (Beattie et al., 2021). Shareholders may perceive favorable coverage of a large recall as an attempt to obscure negative news, reinforcing signal inconsistency and raising negative reactions.

Additionally, while customer harm exhibits the expected negative coefficient, the association does not reach statistical significance (b_{Customer harm} = −.657, p > .1), a finding consistent with prior research (Eilert et al., 2017). Finally, shareholders react less negatively to a software (vs. nonsoftware) recall (Column I: b_{Software recall} = 2.375, p < .01), suggesting that shareholders are less punitive toward software defects, which can be addressed by less costly over-the-internet changes (Darby et al., 2023). Collectively, these results suggest that contextual variables (i.e., recall variables in our case) shape the supplier's shareholders’ perceived uncertainty, thus shaping their reactions. These characteristics are outside the supplier's managerial control. However, the evidence would help managers understand the contextual variables’ explanatory value in determining shareholder reactions.

4.3 Supplier's Revenue Dependence on the Recalling Manufacturer-Customer → Supplier's CAR to the Manufacturer's Recall

Table 2 reports how the supplier's revenue dependence on the recalling manufacturer-customer impacts its shareholders’ reactions to the manufacturer's recall. Column I displays the results without the Heckman correction term's inclusion. Results suggest that the higher the supplier's dependence on the recalling manufacturer, the more negative the supplier's shareholders’ reactions to the recall (b_{Revenue dependence} = −0.041, p < .05).

Table 2.
Supplier's revenue dependence on manufacturer → shareholders’ reactions to a manufacturer–customer's product recall.

DV: CAR_{[−1, 1]} DV: Revenue dependence observed DV: CAR_{[−1, 1]}

I II III

Two-stage Heckman model

Without correction for sample selection First stage of Heckman's model Second stage of Heckman's model

Supplier's revenue dependence −0.0412** −0.0414*

(0.0193) (0.0243)

Recall size −1.000 0.0286 −1.015

(0.463) (0.167) (0.681)

Recall news volume −0.752* 0.00558 −0.775***

(0.188) (0.0618) (0.237)

Recall news sentiment −0.168* −0.0121 −0.140

(0.0842) (0.0271) (0.112)

Customer harm −1.027 0.235 −1.337

(1.528) (0.306) (1.371)

Software recall 3.490* −0.213 3.822

(1.753) (1.694) (1.805)

Supplier's customer breadth −0.224 0.823* −1.202

(0.386) (0.144) (0.876)

Supplier's customer depth 0.0976 0.468*** −0.478

(0.109) (0.0530) (0.477)

Supplier's customer equity −9.847* 0.362 −10.86*

(5.471) (1.105) (5.941)

Supplier's R&D 3.663 2.909 −2.166

(7.524) (2.117) (11.89)

Supplier's liquidity 0.266 −0.272* 0.539

(0.222) (0.104) (0.399)

Supplier's age 0.000717 −0.0104 0.0225

(0.00919) (0.00449) (0.0238)

Supplier's profitability 8.899* 1.876 6.026

(2.922) (0.888) (6.334)

Supplier's size −0.0850 −0.167*** 0.0962

(0.165) (0.0579) (0.291)

Supplier's gross margin −9.002 −1.041 −7.067

(5.528) (1.017) (6.790)

Recall scope: global versus US −1.251 0.263 −1.456

(0.789) (0.344) (0.939)

Recall initiator: government versus manufacturer 0.450 −0.174 0.638

(0.908) (0.223) (0.921)

Recall initiator: not available versus manufacturer 0.265 −0.605* 0.940

(1.410) (0.362) (1.404)

Clustered recall 0.0824 0.594 −0.757

(1.755) (0.557) (2.440)

Additional recall 0.885 −0.218 1.394

(1.628) (0.330) (1.520)

Manufacturer's gross margin 19.91 −4.197 24.55

(15.86) (2.924) (18.80)

Region of manufacturer's headquarters: 3.152* −0.926* 4.391***

Asia versus the United States (0.920) (0.308) (1.391)

Region of manufacturer's headquarters: −0.998 0.0365 −1.116

Europe versus the United States (2.542) (0.630) (3.214)

Supplier's headquarters: North America (vs. Rest of the World) −0.387*

(0.199)

Inverse Mills ratio −2.414

(1.860)

Constant −2.063 −0.619 1.255

(1.314) (0.393) (2.967)

Year FE Yes Yes Yes

Observations 223 896 (223 selected) 223

Fit statistics R² 0.272 Pseudo-R² = 0.428 Wald chi² test = 64.08 (p < .01)

	DV: CAR_{[−1, 1]}	DV: Revenue dependence observed	DV: CAR_{[−1, 1]}
Supplier's revenue dependence	−0.0412**		−0.0414*
	(0.0193)		(0.0243)
Recall size	−1.000**	0.0286	−1.015
	(0.463)	(0.167)	(0.681)
Recall news volume	−0.752***	0.00558	−0.775***
	(0.188)	(0.0618)	(0.237)
Recall news sentiment	−0.168*	−0.0121	−0.140
	(0.0842)	(0.0271)	(0.112)
Customer harm	−1.027	0.235	−1.337
	(1.528)	(0.306)	(1.371)
Software recall	3.490*	−0.213	3.822**
	(1.753)	(1.694)	(1.805)
Supplier's customer breadth	−0.224	0.823***	−1.202
	(0.386)	(0.144)	(0.876)
Supplier's customer depth	0.0976	0.468***	−0.478
	(0.109)	(0.0530)	(0.477)
Supplier's customer equity	−9.847*	0.362	−10.86*
	(5.471)	(1.105)	(5.941)
Supplier's R&D	3.663	2.909	−2.166
	(7.524)	(2.117)	(11.89)
Supplier's liquidity	0.266	−0.272***	0.539
	(0.222)	(0.104)	(0.399)
Supplier's age	0.000717	−0.0104**	0.0225
	(0.00919)	(0.00449)	(0.0238)
Supplier's profitability	8.899***	1.876**	6.026
	(2.922)	(0.888)	(6.334)
Supplier's size	−0.0850	−0.167***	0.0962
	(0.165)	(0.0579)	(0.291)
Supplier's gross margin	−9.002	−1.041	−7.067
	(5.528)	(1.017)	(6.790)
Recall scope: global versus US	−1.251	0.263	−1.456
	(0.789)	(0.344)	(0.939)
Recall initiator: government versus manufacturer	0.450	−0.174	0.638
	(0.908)	(0.223)	(0.921)
Recall initiator: not available versus manufacturer	0.265	−0.605*	0.940
	(1.410)	(0.362)	(1.404)
Clustered recall	0.0824	0.594	−0.757
	(1.755)	(0.557)	(2.440)
Additional recall	0.885	−0.218	1.394
	(1.628)	(0.330)	(1.520)
Manufacturer's gross margin	19.91	−4.197	24.55
	(15.86)	(2.924)	(18.80)
Region of manufacturer's headquarters:	3.152***	−0.926***	4.391***
Asia versus the United States	(0.920)	(0.308)	(1.391)
Region of manufacturer's headquarters:	−0.998	0.0365	−1.116
Europe versus the United States	(2.542)	(0.630)	(3.214)
Supplier's headquarters: North America (vs. Rest of the World)		−0.387*
		(0.199)
Inverse Mills ratio			−2.414
			(1.860)
Constant	−2.063	−0.619	1.255
	(1.314)	(0.393)	(2.967)
Year FE	Yes	Yes	Yes
Observations	223	896 (223 selected)	223
Fit statistics	R² 0.272	Pseudo-R² = 0.428	Wald chi² test = 64.08 (p < .01)

Notes: Standard errors (SEs) are reported in parentheses. The regression for Column I uses SEs clustered by suppliers. The regressions for Columns II and III (i.e., Heckman's two-stage model) use SEs bootstrapped 500 times. CAR = cumulative abnormal stock return; FE = fixed effects.

***p < .01, **p < .05, *p < .1.

Column II presents the estimates from the first-stage regression of Heckman's model. Consistent with our expectation, a supplier headquartered in the United States or Canada is less likely to disclose its dependence on the recalling manufacturer than a counterpart headquartered in other countries (b_{HQ in US/Canada} = −0.387, p = .052). Thus, the excluded variable likely meets the relevance criterion. Statistical evidence suggests that the variable may also meet the exclusion restriction. Specifically, the pseudo-R² of the first-stage regression is 43%, well above the 24% benchmark obtained by Certo et al.'s (2016) simulations to define a strong exclusion restriction. Additionally, the correlation coefficient between revenue dependence and the inverse Mills ratio (λ) is .05, exceeding Certo et al.'s (2016) benchmark of .30. In other words, the power of the test of λ is greater as the correlation between λ and the potentially endogenous explanatory variable is smaller. Despite a smaller subsample, the recall variables’ coefficients are similar to those in Table 1.

4.4 Insights From Covariates’ Coefficients

The supplier's customer depth's coefficient is positive and significant in the full-sample regression (Table 1: $b_{S u p p l i e r' s c u s t o m e r d e p t h} = 0.181$ , p < .1) but insignificant in the subsample regression (Table 2, $b_{S u p p l i e r' s c u s t o m e r d e p t h} = 0.098$ , p > .1). These findings are consistent with the customer portfolio management literature, which stipulates that a deeper portfolio reduces the supplier's demand uncertainty, assuaging shareholders’ punitive reactions (Cohen and Li, 2020; Korcan and Patatoukas, 2016; Leung and Sun, 2021). Similarly, supplier liquidity's and size's coefficients are positive and significant in the full sample ( $b_{S u p p l i e r' s l i q u i d i t y} = 0.409$ , p < .05; $b_{S u p p l i e r' s s i z e} = 0.315$ , p < .01) but insignificant in the subsample ( $b_{S u p p l i e r' s l i q u i d i t y} = 0.266$ , p > .01 $b_{S u p p l i e r' s s i z e} = - 0.085$ , p > .1). A supplier's liquid assets could help limit the contagion. In contrast, larger suppliers are likely to have more resources to withstand the adverse effects of recalls and more opportunities to find alternative sources of revenue, attenuating shareholders’ negative reactions.

Customer depth's, liquidity's, and size's coefficients’ significance in the full sample and insignificance in the subsample are consistent with screening theory, which posits that screens vary according to their strength (Gulati and Higgins, 2003). In our context, the supplier's revenue dependence on the manufacturer-customer is a strong screen for the supplier's shareholders, overshadowing weaker ones. Thus, shareholders strongly consider alternative screens, such as customer depth, liquidity, and size, when the revenue dependence screen is unobservable.

4.5 Robustness Tests

We undertake six robustness tests on the full sample and the subsample. (1) We treat year as a continuous variable because variance inflation factors in the full model exceed 10, mostly due to the correlation between year-fixed effects and some regressors (in that model, the maximum VIF is 5.1). (2) We include the manufacturer's reputation and market share covariates and use year as a continuous variable because adding two new covariates with year-fixed effects amplifies multicollinearity. (3) We measure abnormal return using the Fama-French model (Fama and French, 1993) instead of the market model. (4) We Winsorize the outcome variable at the 5th percentile to limit the influence of extreme values. (5) We exclude clustered recall events within the three-day window [−1, 1]. (6) We sample suppliers headquartered in the United States and Canada. The sign, magnitude, and significance levels of the estimated coefficients of the alternative samples and/or regressions are consistent with those reported in Tables 1 and 2. Columns 1.1–6.2 in Table D1 in Web Appendix D present the estimates.

Because the endogeneity correction term for the supplier's voluntary disclosure variable was significant (Table 1, Column III), we repeated all robustness tests related to the supplier's voluntary disclosure variable with the endogeneity correction, yielding similar results (Table D2 in Web Appendix D).

5 Discussion

Manufacturer–supplier relations are intertwined, particularly in product quality (Barnett and King, 2008; Hertzel et al., 2008; Yu et al., 2008). Therefore, intuition suggests that a manufacturer-customer's product recall—particularly, a large one—can raise uncertainty about the demand for the manufacturer's products and, by extension, the supplier's products (Freedman et al., 2012; Liu and Varki, 2021; Mukherjee et al., 2022a, 2022b; Zavyalova et al., 2012). The supplier's shareholders experience this uncertainty and bid down the supplier's stock price. Our post-hoc analysis reveals that a manufacturer's large recall can erase 1.15% of a supplier's shareholder value,⁵ amounting to US$23 million for an average supplier in our sample. Further, an average manufacturer in our sample contracts with 33 suppliers, resulting in a substantial loss in suppliers’ cumulative shareholder value. The manufacturer-customer's recall's negative impact on a supplier's shareholder value (i.e., contagion or negative spillover) constitutes the starting point of our research. We invoke screening theory to propose the supplier's prior voluntary disclosure of customer information and recall contextual variables as screens that may moderate shareholders’ uncertainty and, by extension, their reactions.

5.1 Theoretical Contributions

Prior research on supply-chain (or more specifically, customer-to-supplier) contagion (see the revised E-Companion's new Table B1 in Web Appendix B) has shown that customer-related negative information can adversely impact a supplier's shareholder value. A manufacturer's decision to recall defective products is a notable omission in the literature. One can argue that the recall is “just another” negative information and thus theoretically identical to other customer-related negative information. However, the manufacturer's recall is an admission of low product quality, which is customer-related negative information directly related to the firm's supply-chain management. More importantly, recall information is substantively distinct from negative information about the manufacturer's accounting, finance, or management failure (e.g., low earnings, bankruptcy filing, and misconduct). As a result, one cannot confidently extrapolate prior research findings to contagion in the recall context. We show that a manufacturer's recall reduces its supplier's shareholder value. This finding contributes to the supply-chain contagion literature by replicating the contagion (or negative spillover) effect in the OM-relevant context of a customer's admission of a product-quality defect. Although managerially relevant, we do not frame this finding as a contribution because prior research has shown contagion, albeit in non-recall contexts.

We offer contributions to (1) the supply-chain contagion literature (see Table B1), (2) screening theory, and (3) the customer information disclosure literature.

First, the supply-chain contagion literature (see Table B1) has examined the presence of contagion and explained heterogeneity through managerially uncontrollable variables (e.g., headquarters’ country/region) (e.g., Hendricks et al., 2020; Qiu et al., 2024). We contribute to this literature by demonstrating that the supplier's prior customer-related disclosures and recall contextual variables can mitigate shareholders’ perceived uncertainty, influencing their punitive reaction to the manufacturer's recall (Eckert, 2020). Interestingly, the supplier's revenue dependence on the recalling customer, disclosed as part of the superset of customer information, aggravates the reaction. Thus, we extend the contagion literature by demonstrating that managerially controllable and contextual variables can influence contagion.

Second, prior research on screening theory (Bergh et al., 2020; Connelly et al., 2021; Sanders and Boivie, 2004) has implicitly assumed that a less-informed decision-maker (e.g., the supplier's shareholders in our context) undertakes a single-stage screening to alleviate their uncertainty and make their decision. We extend the screening theory by suggesting that shareholders’ preferred screen may be unavailable, leading them to adopt a two-stage screening approach. In our recall context, the ideal (and obvious) screen is the supplier's revenue dependence on the recalling manufacturer. However, the supplier may not have disclosed this information in its most recent annual report. This nondisclosure is legally compliant and arguably preferred because it prevents the supplier from disclosing proprietary information to rivals and thus avoids the costs of proprietary information (Ellis et al., 2012; He et al., 2020). Therefore, shareholders undertake a two-stage screening. First, they check whether the supplier voluntarily disclosed customer information in the most recent annual report. The supplier's transparency in revealing customer information may attenuate shareholders’ perceived uncertainty about the supplier's customer relations, mitigating their punitive reaction. Second, if the supplier disclosed its revenue dependence on the recalling manufacturer-customer, the dependence amplifies shareholders’ demand uncertainty, aggravating their punitive reaction. We reason that this two-stage procedure is a novel addition to screening theory.

Third, the literature on customer information disclosure presents two-sided arguments about whether shareholders value this disclosure (Bayer et al., 2017; Ellis et al., 2012; Xu et al., 2024). On the one hand, the disclosure signals the firm's transparency, boosting shareholder value. On the other hand, it can allow the firm's rivals to poach its customers, impeding shareholder value. We contribute to this literature by documenting disclosure's asymmetrical effects in the substantive context of customer recall, which impacts the supplier's shareholder value. Disclosing “general” customer information helps limit shareholder-value loss. In contrast, disclosing “specific” customer information (i.e., revenue dependence on the recalling manufacturer) amplifies shareholder-value loss. Thus, our findings generalize the two-sided arguments, expanding the disclosure's benefits and costs when demand uncertainty is high (Cohen and Li, 2020; Fang et al., 2011; Korcan and Patatoukas, 2016; Patatoukas, 2012).

5.2 Managerial Implications

Our findings alert supplier firm managers that a manufacturer-customer's recall can induce shareholders’ punitive reaction, reflected in a drop in the supplier's shareholder value. More importantly, we reveal several factors that can mitigate or aggravate this contagion.

We suggest managers consider the contagion effect of a customer's recall when deciding whether to disclose customer information. The disclosure can attenuate contagion by revealing the supplier's alternative sources of revenue. At the same time, disclosure may exacerbate contagion by revealing the extent of revenue in jeopardy. Notably, nondisclosure is not necessarily a panacea. For example, if the supplier's performance declines due to a customer's large recall, the lack of information about the supplier's revenue dependence on that customer may induce shareholder distrust. They may fear the worst-case cash flow drop and impose the highest penalty on the supplier. Disclosure's two-sided implications highlight the delicate balance managers must strike between transparency and anticipatory management of recall risks, especially in industries characterized by frequent recalls—for example, an automotive supplier can expect about three large-scale customer recalls each year. We performed a post hoc analysis to empirically determine when a voluntary disclosure strategy may be effective in preventing contagion. Results show that if a supplier's customers account for < 21% of its annual sales revenue, it should voluntarily disclose its customer relationship information. Conversely, if major customers account for 21% or more of the supplier's annual sales revenue, disclosure's aggravation outweighs mitigation, suggesting that the supplier is better off not disclosing the relationships.

Additionally, when contagion looms, the supplier may fare better by leveraging several firm- and recall-specific characteristics. Specifically, larger, older suppliers with greater liquidity and better customer relations are less susceptible to contagion. By contrast, larger recalls and those covered broadly and positively by news publishers exacerbate contagion. Recalls triggered by software versus nonsoftware defects yield milder contagion. Thus, managers can proactively mitigate contagion by improving financial flexibility and strengthening customer ties. If these characteristics provide sufficient mitigation, managers may evaluate whether the disclosure's costs (i.e., revealing dependence) dominate the reduced benefits (i.e., transparency about alternative revenue sources). Lastly, empirically, recall characteristics’ mitigation is more potent when the firm does not disclose customer information than when it does—recall characteristics’ effects are stronger in Model 1 than in Model 2—further highlighting the usefulness of knowing which characteristics blunt large recalls’ effects if firms decide not to be transparent about customer information.

5.3 Limitations and Future Research

Our research provides initial proof-of-concept evidence that suppliers can, to some extent, mitigate a customer recall's negative effect. We hope our findings serve as a basis to generate hypotheses that can be formally tested in the future. In addition to the need to more firmly establish the current findings, we propose that further research could examine three specific extensions. First, we provide evidence from the U.S. automotive industry. Future research should consider checking our findings’ generalizability to (1) other industries (e.g., such as food and medical devices), and (2) other countries where institutions and regulations differ from those in the United States (e.g., China). Second, we focus on the upstream (i.e., customer-to-supplier) contagion. Future research could test for downstream contagion. For example, when Takata Corporation announced its airbag recall, how did Takata's customers’ shareholders react? We posit that a mechanism different from our proposed mechanism (of demand uncertainty) might explain the answer. Third, future research could investigate whether contagion occurs and how it can be mitigated for other customer-related negative events, such as production halts and employee strikes (Astvansh et al., 2023; Astvansh and Simpson, 2025). In summary, our research represents a valuable first step at the intersection of manufacturer-specific operational failures and manufacturer–supplier relations, providing avenues for future research.

Supplemental Material

sj-docx-1-pao-10.1177_10591478251397690 - Supplemental material for Product Recall Contagion in the Supply Chain

Supplemental material, sj-docx-1-pao-10.1177_10591478251397690 for Product Recall Contagion in the Supply Chain by Ljubomir Pupovac, Vivek Astvansh, François Carrillat and Renaud Legoux in Production and Operations Management

Footnotes

Acknowledgements

This article is the first essay of Ljubo's doctoral thesis, which won the Pennsylvania State University Institute for the Study of Business Markets’ (ISBM's) 2019 Doctoral Support Competition. The authors thank Gary Lilien for his valuable work on earlier versions of this article and Thomas Schreiner for his assistance with the text analysis.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Vivek Astvansh

Supplemental Material

Supplemental material for this article is available online (doi: ).

Notes

How to cite this article

Pupovac L, Astvansh V, Carrillat F and Legoux R (2026) Product Recall Contagion in the Supply Chain. Production and Operations Management 35(5): 2003–2018.

References

Astvansh

Antia

Tellis

(2024b) Product recall: A synthesis of multidisciplinary findings, and research directions. Marketing Letters: 1–13.

Astvansh

Antia

Tellis

(2024a) What is (and isn’t) a product recall? Journal of Public Policy & Marketing 43(4): 295–317.

Astvansh

Ball

Josefy

(2022a) The recall decision exposed: Automobile recall timing and process data set. Manufacturing & Service Operations Management 24(3): 1457–1473.

Astvansh

Duffek

Eisingerich

(2023) How can companies recover from liability-invoking failures? Exploring the role of uncertainty avoidance in facilitating consumer compliance across national cultures. Journal of International Marketing 31(3): 1–18.

Astvansh

Eshghi

(2023) The effects of regulatory investigation, supplier defect, and product age on stock investors’ reaction to an automobile recall. Journal of Business Research 167: 114052.

Astvansh

Jindal

(2022) Differential effects of received trade credit and provided trade credit on firm value. Production and Operations Management 31(2): 781–798.

Astvansh

Simpson

(2025) A firm’s operational risk: Data set and empirical evidence. Manufacturing & Service Operations Management. Forthcoming.

Astvansh

Suri

Damavandi

(2024c) Brand warmth elicits feedback, not complaints. Journal of the Academy of Marketing Science 52(4): 1107–1129.

Astvansh

Wang

Shi

(2022b) The effects of the news media on a firm's voluntary product recalls. Production and Operations Management 31(11): 4223–4244.

10.

Lisic

Liu

, et al. (2013) Stock market reaction to green vehicle innovation. Production and Operations Management 22(4): 976–990.

11.

Barnett

King

(2008) Good fences make good neighbors: A longitudinal analysis of an industry self-regulatory institution. Academy of Management Journal 51(6): 1150–1170.

12.

Bayer

Tuli

Skiera

(2017) Do disclosures of customer metrics lower investors’ and analysts’ uncertainty but hurt firm performance? Journal of Marketing Research 54(2): 239–259.

13.

Beattie

Durante

Knight

, et al. (2021) Advertising spending and media bias: Evidence from news coverage of car safety recalls. Management Science 67(2): 698–719.

14.

Bergh

Peruffo

Chiu

, et al. (2020) Market response to divestiture announcements: A screening theory perspective. Strategic Organization 18(4): 547–572.

15.

Borah

Tellis

(2016) Halo (spillover) effects in social media: Do product recalls of one brand hurt or help rival brands? Journal of Marketing Research 53(2): 143–160.

16.

Casciaro

Piskorski

(2005) Power imbalance, mutual dependence, and constraint absorption: A closer look at resource dependence theory. Administrative Science Quarterly 50(2): 167–199.

17.

Cashman

Harrison

Seiler

, et al. (2019) The relation between intrafirm distances and information opacity: Evidence from stock market liquidity. Journal of Real Estate Research 41(4): 639–668.

18.

Certo

Busenbark

Woo

, et al. (2016) Sample selection bias and Heckman models in strategic management research. Strategic Management Journal 37(13): 2639–2657.

19.

Chakrabarti

Gupta-Mukherjee

Jayaraman

(2009) Mars–Venus marriages: Culture and cross-border M&A. Journal of International Business Studies 40: 216–236.

20.

Chakravarty

Saboo

Xiong

(2022) Marketing's and operations’ roles in product recall prevention: Antecedents and consequences. Production and Operations Management 31(3): 1174–1190.

21.

Chen

Ganesan

Liu

(2009) Does a firm's product-recall strategy affect its financial value? An examination of strategic alternatives during product-harm crises. Journal of Marketing 73(6): 214–226.

22.

Cho

DeMiguel

Hwang

(2021) Cover-up of vehicle defects: The role of regulator investigation announcements. Management Science 67(6): 3834–3852.

23.

Cleeren

Dekimpe

Van Heerde

(2017) Marketing research on product-harm crises: A review, managerial implications, and an agenda for future research. Journal of the Academy of Marketing Science 45(5): 593–615.

24.

Cleeren

Van Heerde

Dekimpe

(2013) Rising from the ashes: How brands and categories can overcome product-harm crises. Journal of Marketing 77(2): 58–77.

25.

Cohen

(2020) Customer-base concentration, investment, and profitability: The US government as a major customer. The Accounting Review 95(1): 101–131.

26.

Connelly

Shi

Cheng

, et al. (2021) Short sellers: A screening theory perspective on B2B relationships. Journal of Business Research 134: 393–404.

27.

Crouch

Pourazad

, et al. (2020) Investigating country image influences after a product-harm crisis. European Journal of Marketing 55(3): 894–924.

28.

Damavandi

Astvansh

(2025) Unveiling regulatory operations: A data set of the determinants, process, and outcomes of product defect investigations by the US automotive safety regulator. Manufacturing & Service Operations Management 27(1): 181–199.

29.

Darby

Ketchen

DJ Jr

Ball

, et al. (2023) CEO stock ownership, recall timing, and stock market penalties. Manufacturing & Service Operations Management 25(5): 1909–1930.

30.

Davidson

WN III

Worrell

(1992) Research notes and communications: The effect of product recall announcements on shareholder wealth. Strategic Management Journal 13(6): 467–473.

31.

Eckert

(2020) Risk and risk management of spillover effects: Evidence from the literature. Risk Management and Insurance Review 23(1): 75–104.

32.

Eilert

Jayachandran

Kalaignanam

, et al. (2017) Does it pay to recall your product early? An empirical investigation in the automobile industry. Journal of Marketing 81(3): 111–129.

33.

Ellis

Fee

Thomas

(2012) Proprietary costs and the disclosure of information about customers. Journal of Accounting Research 50(3): 685–727.

34.

Emerson

(1962) Power-dependence relations. American Sociological Review 27(1): 31–41.

35.

Eshghi

Astvansh

(2024) Stock investors’ reaction to layoff announcements: A meta-analysis. Human Resource Management Journal 34(3): 792–809.

36.

Fama

French

(1993) Common risk factors in the returns on stocks and bonds. Journal of Financial Economics 33(1): 3–56.

37.

Fang

Palmatier

Grewal

(2011) Effects of customer and innovation asset configuration strategies on firm performance. Journal of Marketing Research 48(3): 587–602.

38.

Fang

Astvansh

Tong

, et al. (2025) How do brands change their advertising spending in response to a rival's product recall? Production and Operations Management 34(6): 1387–1408.

39.

Freedman

Kearney

Lederman

(2012) Product recalls, imperfect information, and spillover effects: Lessons from the consumer response to the 2007 toy recalls. Review of Economics and Statistics 94(2): 499–516.

40.

Gao

Xie

Wang

, et al. (2015) Should ad spending increase or decrease before a recall announcement? The marketing–finance interface in product-harm crisis management. Journal of Marketing 79(5): 80–99.

41.

Giannetti

Srinivasan

(2021) The cloud and its silver lining: Negative and positive spillovers from automotive recalls. Marketing Letters 32(4): 397–409.

42.

Gour

Aggarwal

Kumar

(2022) Lending ears to unheard voices: An empirical analysis of user-generated content on social media. Production and Operations Management 31(6): 2457–2476.

43.

Gulati

Higgins

(2003) Which ties matter when? The contingent effects of interorganizational partnerships on IPO success. Strategic Management Journal 24(2): 127–144.

44.

Chen

Chan

(2020) Know your competitors: Customer identity disclosure by competitors and a firm's production efficiency. International Review of Finance 22(4): 777–792.

45.

Heckman

(1979) Sample selection bias as a specification error. Econometrica: Journal of the Econometric Society 47(1): 153–161.

46.

Heide

John

(1988) The role of dependence balancing in safeguarding transaction-specific assets in conventional channels. Journal of Marketing 52(1): 20–35.

47.

Hendricks

Jacobs

Singhal

(2020) Stock market reaction to supply chain disruptions from the 2011 Great East Japan Earthquake. Manufacturing & Service Operations Management 22(4): 683–699.

48.

Hendricks

Singhal

(2003) The effect of supply chain glitches on shareholder wealth. Journal of Operations Management 21(5): 501–522.

49.

Hertzel

Officer

, et al. (2008) Inter-firm linkages and the wealth effects of financial distress along the supply chain. Journal of Financial Economics 87(2): 374–387.

50.

Hoffer

Pruitt

Reilly

(1988) The impact of product recalls on the wealth of sellers: A reexamination. Journal of Political Economy 96(3): 663–670.

51.

Jacobs

Singhal

(2020) Shareholder value effects of the Volkswagen emissions scandal on the automotive ecosystem. Production and Operations Management 29(10): 2230–2251.

52.

Jacobs

Singhal

Zhan

(2022) Stock market reaction to global supply chain disruptions from the 2018 US government ban on ZTE. Journal of Operations Management 68(8): 903–927.

53.

Jarrell

Peltzman

(1985) The impact of product recalls on the wealth of sellers. Journal of Political Economy 93(3): 512–536.

54.

Javadinia

Gill

Jayachandran

(2023) Recall environment and post-recall stock market response. Journal of the Academy of Marketing Science 52(4): 1171–1194.

55.

Kolay

Lemmon

Tashjian

(2016) Spreading the misery? Sources of bankruptcy spillover in the supply chain. Journal of Financial and Quantitative Analysis 51(6): 1955–1990.

56.

Korcan

Patatoukas

(2016) Customer-base concentration and inventory efficiencies: Evidence from the manufacturing sector. Production and Operations Management 25(2): 258–272.

57.

Koval

Zaefarian

Iurkov

(2024) How do strategic alliance formations create shareholder value? An application of the event study methodology in the B2B context. Industrial Marketing Management 117: 79–91.

58.

Leung

Sun

(2021) Policy uncertainty and customer concentration. Production and Operations Management 30(5): 1517–1542.

59.

(2010) The information content of forward-looking statements in corporate filings—A naïve Bayesian machine learning approach. Journal of Accounting Research 48(5): 1049–1102.

60.

Simcoe

(2021) Competing or complementary labels? Estimating spillovers in Chinese green building certification. Strategic Management Journal 42(13): 2451–2476.

61.

Ling

Wang

Zhou

(2021) The geography of real property information and investment: Firm location, asset location and institutional ownership. Real Estate Economics 49(1): 287–331.

62.

Ling

Zhang

(2020) How do industry peers influence individual firms’ voluntary disclosure strategies? Review of Quantitative Finance and Accounting 54(3): 911–956.

63.

Liu

Varki

(2021) The spillover effect of product recalls on competitors’ market value: The role of corporate product reliability. Journal of Business Research 137: 452–463.

64.

Liu

Shankar

(2015) The dynamic impact of product-harm crises on brand preference and advertising effectiveness: An empirical analysis of the automobile industry. Management Science 61(10): 2514–2535.

65.

Liu

Shankar

Yun

(2017) Crisis management strategies and the long-term effects of product recalls on firm value. Journal of Marketing 81(5): 30–48.

66.

Ding

Peng

, et al. (2018) Addressing endogeneity in operations management research: Recent developments, common problems, and directions for future research. Journal of Operations Management 64: 53–64.

67.

Maitland

Bryson

Van de Ven

(1985) Sociologists, economists, and opportunism. Academy of Management Review 10(1): 59–65.

68.

Malkiel

Fama

(1970) Efficient capital markets: A review of theory and empirical work. The Journal of Finance 25(2): 383–417.

69.

McGee

(2017) Carmakers face threat from new drivers of profit. Financial Times, (August 08). Available at https://www.ft.com/content/40065b50-715e-11e7-93ff-99f383b09ff9 (accessed January 5, 2024)

70.

Merton

(1987) A simple model of capital market equilibrium with incomplete information. Journal of Finance 42(3): 483–510.

71.

Mukherjee

Dutta

De Bruyn

(2022b) Did clickbait crack the code on virality? Journal of the Academy of Marketing Science 50(3): 482–502.

72.

Mukherjee

Ball

Wowak

, et al. (2022a) Hiding in the herd: The product recall clustering phenomenon. Manufacturing & Service Operations Management 24(1): 392–410.

73.

Nahata

Hazarika

Tandon

(2014) Success in global venture capital investing: Do institutional and cultural differences matter? Journal of Financial and Quantitative Analysis 49(4): 1039–1070.

74.

Noack

Miller

Smith

(2019) Let me make it up to you: Understanding the mitigative ability of corporate social responsibility following product recalls. Journal of Business Ethics 157: 431–446.

75.

Panagopoulos

Mullins

Avramidis

(2018) Sales force downsizing and firm-idiosyncratic risk: The contingent role of investors’ screening and firm’s signaling processes. Journal of Marketing 82(6): 71–88.

76.

Papies

Ebbes

Van Heerde

(2017) Addressing endogeneity in marketing models. In: Leeflang

PSH

Wieringa

Bijmolt

THA

, et al. (eds) Advanced Methods for Modelling Marketing. Switzerland: Springer, pp.581–627.

77.

Patatoukas

(2012) Customer-base concentration: Implications for firm performance and capital markets. The Accounting Review 87(2): 363–392.

78.

Petrin

Train

(2010) A control function approach to endogeneity in consumer choice models. Journal of Marketing Research 47(1): 3–13.

79.

Pfeffer

Salancik

(1978) The External Control of Organizations: A Resource Dependence Perspective. New York: Harper and Row.

80.

Pupovac

Carrillat

Michayluk

(2022) Slicing vs chunking product-harm crisis: Antecedents and firm performance implications. European Journal of Marketing 56(7): 1856–1884.

81.

Qian

Crilly

Wang

, et al. (2021) Why do banks favor employee-friendly firms? A stakeholder-screening perspective. Organization Science 32(3): 605–624.

82.

Qiu

Yeung

, et al. (2024) Contagious stock price crashes along the supply chain. Production and Operations Management 33(8): 1679–1699.

83.

Rawlins

(2008) Give the emperor a mirror: Toward developing a stakeholder measurement of organizational transparency. Journal of Public Relations Research 21(1): 71–99.

84.

Sanders

Boivie

(2004) Sorting things out: Valuation of new firms in uncertain markets. Strategic Management Journal 25(2): 167–186.

85.

Schmidt

Wuttke

Ball

, et al. (2020) Does social media elevate supply chain importance? An empirical examination of supply chain glitches, Twitter reactions, and stock market returns. Journal of Operations Management 66(6): 646–669.

86.

Sorescu

Warren

Ertekin

(2017) Event study methodology in the marketing literature: An overview. Journal of the Academy of Marketing Science 45(2): 186–207.

87.

Spence

(1974) Competitive and optimal responses to signals: An analysis of efficiency and distribution. Journal of Economic Theory 7(3): 296–332.

88.

Stiglitz

(1975) The theory of “screening,” education, and the distribution of income. The American Economic Review 65(3): 283–300.

89.

Stout (2019) 2019 Automotive Defect & Recall Report. https://www.stout.com/en/insights/report/2019-automotive-defect-and-recall-report

90.

Tetlock

(2007) Giving content to investor sentiment: The role of media in the stock market. The Journal of Finance 62(3): 1139–1168.

91.

Thirumalai

Sinha

(2011) Product recalls in the medical device industry: An empirical exploration of the sources and financial consequences. Management Science 57(2): 376–392.

92.

Tourani-Rad

Kirkby

(2005) Investigation of investors’ overconfidence, familiarity and socialization. Accounting & Finance 45(2): 283–300.

93.

Van Everdingen

Fok

Stremersch

(2009) Modeling global spillover of new product takeoff. Journal of Marketing Research 46(5): 637–652.

94.

Warren

Sorescu

(2017) Interpreting the stock returns to new product announcements: How the past shapes investors’ expectations of the future. Journal of Marketing Research 54(5): 799–815.

95.

Wathne

Heide

(2000) Opportunism in interfirm relationships: Forms, outcomes, and solutions. Journal of Marketing 64(4): 36–51.

96.

Wooldridge

(2010) Econometric Analysis of Cross Section and Panel Data. 2nd ed. Boston: MIT Press.

97.

Cai

Zhang

(2024) To tell or not to tell? Examining voluntary disclosure of customers’ identities under capital market pressure. Journal of Accounting and Public Policy 45: 107211.

98.

Sengul

Lester

(2008) Misery loves company: The spread of negative impacts resulting from an organizational crisis. Academy of Management Review 33(2): 452–472.

99.

Zavyalova

Pfarrer

Reger

, et al. (2012) Managing the message: The effects of firm actions and industry spillovers on media coverage following wrongdoing. Academy of Management Journal 55(5): 1079–1101.

100.

Zhang

Shi

Connelly

(2023) Screening theory and its boundaries: Investigation of screen credibility, necessity, and salience in the context of corporate venture capital. Academy of Management Journal 67(5): 1359–1391.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.09 MB

0.00 MB

	DV: CAR_{[−1, 1]}	DV: Supplier's voluntary disclosure: yes versus no	DV: CAR_{[−1, 1]}
	I	II	III
		Two-stage control function
	No endogeneity correction	First-stage	Second-stage
Supplier's voluntary disclosure: yes versus no	0.857**		1.701***
	(0.322)		(0.531)
Recall size	−0.800***	0.0225	−0.792***
	(0.240)	(0.271)	(0.238)
Recall news volume	−0.433***	−0.0442	−0.432***
	(0.0752)	(0.108)	(0.112)
Recall news sentiment	−0.132***	0.0180	−0.133***
	(0.0338)	(0.0411)	(0.0369)
Customer harm	−0.662	0.133	−0.657
	(0.509)	(0.518)	(0.465)
Software recall	2.280**	−1.287	2.375***
	(1.022)	(1.558)	(0.890)
Supplier's customer breadth	−0.148	1.575***	−0.219
	(0.127)	(0.291)	(0.201)
Supplier's customer depth	0.171**	−0.0491	0.181*
	(0.0800)	(0.0988)	(0.0997)
Supplier's customer equity	3.109*	6.346***	2.766
	(1.747)	(2.104)	(2.469)
Supplier's R&D	2.207	23.11***	0.809
	(3.293)	(4.285)	(3.845)
Supplier's liquidity	0.409**	−1.321***	0.499***
	(0.163)	(0.244)	(0.182)
Supplier's age	0.0109*	−0.104***	0.0160**
	(0.00594)	(0.0129)	(0.00662)
Supplier's profitability	−0.397	12.39***	−1.579
	(1.321)	(1.915)	(1.978)
Supplier's size	0.315***	−0.616***	0.357***
	(0.110)	(0.123)	(0.117)
Supplier's gross margin	−1.782	−1.950	−1.766
	(1.126)	(2.296)	(1.316)
Recall scope: global versus US	−0.796	0.0146	−0.800
	(0.689)	(0.609)	(0.644)
Recall initiator: government versus manufacturer	−0.575	−0.110	−0.573
	(0.514)	(0.389)	(0.453)
Recall initiator: not available versus manufacturer	−0.0674	0.0497	−0.0805
	(0.606)	(0.582)	(0.610)
Clustered recall	1.104	−0.285	1.075
	(0.754)	(1.005)	(0.838)
Additional recall	0.121	0.190	0.124
	(0.492)	(0.510)	(0.450)
Manufacturer's gross margin	−4.033	−2.280	−3.648
	(2.977)	(4.741)	(4.237)
Region of manufacturer's headquarters:	1.594***	0.178	1.587***
Asia versus the United States	(0.424)	(0.508)	(0.405)
Region of manufacturer's headquarters:	1.387	−0.354	1.386
Europe versus the United States	(0.956)	(1.068)	(0.949)
Whether a prominent U.S. news publisher covered the supplier (excluded variable for Heckman's first-stage regression)		−0.747**
		(0.302)
Control function			−0.442**
			(0.184)
Constant	−1.677***	−3.206***	−1.866***
	(0.480)	(0.671)	(0.559)
Observations	896	896	896
Year FE	Yes	Yes	Yes
Adjusted (pseudo) R²	0.071	0.484	0.073

	DV: CAR_{[−1, 1]}	DV: Revenue dependence observed	DV: CAR_{[−1, 1]}
	I	II	III
		Two-stage Heckman model
	Without correction for sample selection	First stage of Heckman's model	Second stage of Heckman's model
Supplier's revenue dependence	−0.0412**		−0.0414*
	(0.0193)		(0.0243)
Recall size	−1.000**	0.0286	−1.015
	(0.463)	(0.167)	(0.681)
Recall news volume	−0.752***	0.00558	−0.775***
	(0.188)	(0.0618)	(0.237)
Recall news sentiment	−0.168*	−0.0121	−0.140
	(0.0842)	(0.0271)	(0.112)
Customer harm	−1.027	0.235	−1.337
	(1.528)	(0.306)	(1.371)
Software recall	3.490*	−0.213	3.822**
	(1.753)	(1.694)	(1.805)
Supplier's customer breadth	−0.224	0.823***	−1.202
	(0.386)	(0.144)	(0.876)
Supplier's customer depth	0.0976	0.468***	−0.478
	(0.109)	(0.0530)	(0.477)
Supplier's customer equity	−9.847*	0.362	−10.86*
	(5.471)	(1.105)	(5.941)
Supplier's R&D	3.663	2.909	−2.166
	(7.524)	(2.117)	(11.89)
Supplier's liquidity	0.266	−0.272***	0.539
	(0.222)	(0.104)	(0.399)
Supplier's age	0.000717	−0.0104**	0.0225
	(0.00919)	(0.00449)	(0.0238)
Supplier's profitability	8.899***	1.876**	6.026
	(2.922)	(0.888)	(6.334)
Supplier's size	−0.0850	−0.167***	0.0962
	(0.165)	(0.0579)	(0.291)
Supplier's gross margin	−9.002	−1.041	−7.067
	(5.528)	(1.017)	(6.790)
Recall scope: global versus US	−1.251	0.263	−1.456
	(0.789)	(0.344)	(0.939)
Recall initiator: government versus manufacturer	0.450	−0.174	0.638
	(0.908)	(0.223)	(0.921)
Recall initiator: not available versus manufacturer	0.265	−0.605*	0.940
	(1.410)	(0.362)	(1.404)
Clustered recall	0.0824	0.594	−0.757
	(1.755)	(0.557)	(2.440)
Additional recall	0.885	−0.218	1.394
	(1.628)	(0.330)	(1.520)
Manufacturer's gross margin	19.91	−4.197	24.55
	(15.86)	(2.924)	(18.80)
Region of manufacturer's headquarters:	3.152***	−0.926***	4.391***
Asia versus the United States	(0.920)	(0.308)	(1.391)
Region of manufacturer's headquarters:	−0.998	0.0365	−1.116
Europe versus the United States	(2.542)	(0.630)	(3.214)
Supplier's headquarters: North America (vs. Rest of the World)		−0.387*
		(0.199)
Inverse Mills ratio			−2.414
			(1.860)
Constant	−2.063	−0.619	1.255
	(1.314)	(0.393)	(2.967)
Year FE	Yes	Yes	Yes
Observations	223	896 (223 selected)	223
Fit statistics	R² 0.272	Pseudo-R² = 0.428	Wald chi² test = 64.08 (p < .01)