Labeling Regulation or Product Liability? Managing Product Risks When Consumer Label-Reading E orts Count

Transcription

1 Labeling Regulation or Product Liability? Managing Product Risks When Consumer Label-Reading E orts Count Maria Arbatskaya and Maria Vyshnya Aslam April 24, 2017 Abstract This paper compares regulatory and liability approaches to product safety. Labeling regulation requires a producer to disclose that their product contains harmful content when its level (unknown to consumers) exceeds a certain threshold. Requiring a higher level of transparency for warning labels encourages more consumer label-reading effort and provides further incentives for producer care through a vigilance e ect. By contrast, a stronger product liability encourages producer care but has a lulling effect on consumer care. The reason is that consumers always view producer care and consumer care levels as strategic substitutes, while the rm views them as strategic complements when product liability is weak. We argue that an intervention policy has to be chosen with caution and the endogeneity of consumer label-reading e ort is not to be overlooked. Keywords: product risk, warning labels, labeling regulation, mandatory disclosure, product liability. JEL numbers: K13; L15; L13; D83. Arbatskaya (Corresponding Author): Department of Economics, Emory University, Atlanta, GA Phone: (404) Fax: (404) marbatsemory.edu. Aslam: CDC, National Center for HIV, Viral Hepatitis, STD, and TB Prevention, O ce of Director, Atlanta, GA Maria.Aslamcdc.hhs.gov. We are thankful to the Editors, an anonymous referee, and participants at the International Industrial Organization Conference, the Southern Economic Association Meetings, and the 4th Workshop on Consumer Search for their comments. We are also grateful to Alexander Stein and Naina Hussain for their skillful research assistance.

2 1. Introduction Every day, consumers decide whether to buy products that may present health risks. They shop for yogurt and pastries that may contain life-threatening allergens, purchase laundry detergents containing benzene, and combat a disease with medications that may have adverse side e ects. Since safety information is costly to acquire, consumers are typically not fully informed of the risks they face. There is a general consensus that regulation or litigation e orts could correct the asymmetric awareness of buyers and sellers about a product s safety level. The main debate concerns the design of the best policy. We argue that an intervention policy must be chosen with caution. Although disclosure mandates can ensure that manufacturers provide information to consumers about the risk attributes of a product (for example, by using a warning label), the mandates do not guarantee that consumers read and understand the warning labels. It is important then to set transparency standards for the labels in such a way that consumers have incentives to read them. In a similar vein, a strong liability system that compensates consumers for damages does not provide buyers with incentives to take precautions. 1 The main goal of this paper is to analyze the e ects of labeling regulation and litigation on producer and consumer care and the expected harm associated with products that have risk attributes. Food allergen labeling regulation provides a tting example to motivate our study because to date the most successful method of managing allergies is to avoid the food containing the allergens. It is, therefore, important for consumers to be informed about allergy-related health risks, and food allergen labeling plays a key role in the long-term management of the disease. The federal mandate for disclosing allergen information in packaged foods requires the labeling of major allergens if a manufacturer adds them in any amount. It also sets uniform criteria for positioning the warning on food labels and speci es the standards for label transparency. However, the mandate requires manufacturers to indicate the presence of an allergen rather than its exact amount. 2 1 This is well-recognized in the literature on torts. See a comprehensive survey by Daughety and Reinganum (2013). 2 The national standards for labeling food allergens are set by the FDA s Food Allergen Labeling and Consumer Protection Act of 2004 (FALCPA). FALCPA mandates labeling the eight most common allergens responsible for 90 percent of all food allergies milk, eggs, sh, crustacean shell sh, tree nuts, peanuts, wheat, and soybeans. The mandate sets uniform standards for displaying the warning on food labels. First, FALCPA 1

3 A comprehensive review of the papers on the theory and practice of quality disclosure is provided in Dranove and Jin (2010). When disclosure is costly, a rm may choose to reveal its quality only if it is above a certain threshold level. The lack of full disclosure can then provide a rationale for the use of disclosure mandates. But even if rms reveal information about harmful content to consumers, it does not necessarily mean that consumers would read and understand the message. An increasing sophistication of the message or higher cost of processing information makes it harder for consumers to become fully informed (Anderson and Renault, 2006; Rasmusen, 2001). In our paper, a consumer s ability to nd a warning depends on the label s transparency and the amount of consumer reading e ort. We also assume that the rm is subject to a binary mandatory disclosure regulation and product liability. In addition to regulatory actions, litigation e orts may also be an alternative way to manage product risks. In general, stronger liability increases producer care because it increases the rm s liability costs and thus provides incentives for improving product quality (Polinsky and Rogerson, 1983; Viscusi and Moore, 1993; Daughety and Reinganum, 2013). In the case of food allergies, however, there is little litigation history in the U.S. despite the prevalence of food allergies. This suggests that product liability related to food allergies is weak in the sense that plainti s have di culty in proving their cases and, consequently, are not likely to recover damages. We maintain this as an assumption in this paper as we examine how labeling regulation and product liability a ect both the consumers incentives to nd and understand warning labels and the rm s incentives to invest in reducing product risks. Our paper analyzes bilateral accidents the situation that arises when both producer and consumer care decisions have an impact on the probability of harm. 3 Consumer care is in the form of label-reading e ort, while producer care is in the form of precautions taken bans the use of Latin terminology and requires listing the major allergens in plain and clear English. Second, it bans ne print. Finally, it outlines the standards for positioning the warning. The allergen may be mentioned in parentheses immediately after the ingredient: casein (milk). Alternatively, the statement contains (allergen source) may immediately follow or be adjacent to the list of ingredients (e.g., contains peanuts ). FALCPA regulation of the display of allergy information is aimed at making it easier for consumers to nd and understand warning labels. NIH, In contrast to bilateral accidents, unilateral accidents are situations where one party (e.g., the rm) is solely responsible for accidents. Shavell (2007) provides a comprehensive classi cation of accident liability cases. 2

4 by the producer to reduce the harmful content in a product. 4 The level of care chosen by the rm stochastically determines the amount of harmful content and the rm is mandated to post a warning if this amount exceeds a threshold level. Consumers do not know the level of producer care or the true risk associated with the product. They instead have to expend label-reading e ort in order to nd, read, and understand warning messages and then update their beliefs about the product s risk based on the outcome of their investigation. Quite intuitively, the consumer choice of label-reading e ort depends on the strength of product liability, the transparency of labels, consumer risk perception, susceptibility to harmful content, and the cost of reading e ort. Consumer care is also negatively a ected by producer care (i.e., consumers view care levels as strategic substitutes). Following Viscusi (1984), we call this a lulling e ect safety improvements may produce an unintended e ect of reducing consumer e ort aimed at avoiding an accident. In our model, the interactive e ect of consumer and producer care on harm reduction is negative because consumer label-reading e ort is less e ective at reducing the expected harm when the product is safer. By contrast, for the rm, consumer and producer care levels are strategic complements provided that liability is su ciently weak. The reason is that when liability is weak, the rm primarily cares about the e ect of the consumer label-reading e ort on the demand (not the expected harm), and the levels of care have a positive interaction in the demand. When consumers are vigilant, the rm chooses a higher level of care to avoid the need to post the warning that reduces its demand. We call this strategic e ect on producer care a vigilance e ect. We then study how alternative policy tools of a stronger liability rule, a higher transparency of warning labels, and educational campaigns informing consumers of potential risks a ect equilibrium levels of producer and consumer care, expected harm, consumer payo s, and pro ts. Our ndings have important implications for the choice of policies aimed at reducing the expected harm from a product. A higher transparency of the warning and higher consumer risk perception tend to increase the equilibrium consumer and producer 4 For example, there are many ways to manage risks associated with the presence of allergens in packaged foods. These include compiling a master list of all ingredients and raw materials used (including food additives, avors, and colorings), obtaining documentation from the suppliers of ingredients, managing production scheduling, shipping, handling, and storage, following cleaning procedures, testing for allergens, employee training, and program evaluation. 3

5 care and thus decrease the expected harm. By contrast, a stronger liability system shifts the burden of care from consumers to the rm, which tends to increase producer care and reduce consumer care, with an ambiguous e ect on the expected harm. We argue that it is important to treat consumer care as endogenous. With exogenous consumer label-reading e ort, both a higher warning label transparency and a stronger liability reduce expected harm, but the unintended consequences of policy changes (such as lulling e ect) are present when consumer care is actually endogenous. We summarize the direct e ects and strategic e ects of policy changes in Figure 1. [Figure 1 HERE] Finally, we examine whether there is too much or too little consumer and producer care in the equilibrium. We nd that in the equilibrium, producer care is underprovided compared to the socially optimal level, while consumer reading e ort is socially excessive when liability is low. The reason for this is that a higher label-reading e ort by consumers imposes a negative externality on the rm due to a reduction in demand: the rm prefer that consumers do not read warning labels when liability is low. On the other hand, consumers prefer that the rm provides a safer product. The rest of the paper is organized as follows. The basic setup of our model is presented in Section 2. The consumer problem is analyzed in Section 3, and the optimal rm strategy is derived in Section 4. In Section 5, we look at the equilibrium levels of care in a simultaneousmove game between a rm and consumers and discuss alternative policy changes such as a stronger liability rule, higher transparency of the warning label, and a higher perception of risk among consumers. Section 6 summarizes our ndings and concludes. All the proofs are delegated to Appendix A. 2. Model Setup A rm manufactures a product that may contain a certain amount of harmful content a, such as an allergen. We assume that the product is safe for consumers if the harmful content a does not exceed the consumer tolerance level a 0 > 0. On the other hand, if the 4

6 harmful content a exceeds a 0, then the product is (potentially) unsafe. In this case, under mandatory disclosure, the rm is required to include a warning message on the product s label. We assume that the rm complies with the regulation and provides a warning on the product s label if and only if a > a 0. 5 The rm decides on the level of care, y 0, that stochastically determines the harmful content a in its product. The marginal cost of production m (y) is twice continuously di erentiable and strictly increasing in producer care: m 0 (y) > 0. Let G(a; y) be the twice continuously di erentiable cumulative distribution function for the harmful content a in the product, given the level of producer care y; G(a; y) is de ned on [a; a] with a 0 2 (a; a) and G(a;y) a 0. A higher producer care y shifts the distribution of harmful content in the rstorder stochastic dominance sense: G(a;y) 0. The probability that the product is unsafe is then P (y) = 1 G(a 0 ; y) 2 (0; 1) and P (y) is decreasing in producer care y: P (y)= 0. 6 Each consumer wants to buy at most one unit of the product. Consumers have a value v for the product, but with probability they incur a harm (normalized to one) from consuming an unsafe product. We call consumer susceptibility. To focus on the strategic interaction between producer and consumer care, we assume that both the rm and consumers are risk neutral and treat price as exogenous, assuming that price p 2 (m(y); v). The rm has to compensate consumers for the harm associated with the consumption of the product. The compensation depends on the strength of liability l 2 [0; 1), which depends on the liability system and courts practices. High (low) levels of l correspond to strong (weak) liability, while l = 0 in the case with no liability. The uncompensated cost to the consumer su ering harm is then 1 l > 0. Consumers do not know the level of harmful content a in the product and producer care y. They know about mandatory disclosure, inspect product labels in search of a warning and then make a decision on whether to buy the product. Consumers may have misperceptions 5 For simplicity, we assume here that the disclosure standard is set to match exactly the level of allergen at which consumers start facing risk. Our analysis would be very similar if we assume that the mandated threshold level for disclosure is lower than the allergen tolerance level a 0, which could be the case because FALCPA requires rms to disclose any detectable levels of an allergen, regardless of the tolerance level a 0. 6 We place additional restrictions on the distribution function G(a 0 ; y) and marginal cost m (y) to guarantee that the producer maximization problem has a unique interior solution: and l G(a0;0) > m 0 (0). 2 G(a 0;y) 2 = G(a0;y) 2 < m 00 (y) m 0 (y) 5

7 about product safety. They believe that the probability that the product is unsafe is P e (y) = P (y) 2 (0; 1), where > 0 is the risk perception of consumers: = 1 corresponds to the correct prior assessment of risk; < 1 corresponds to the underestimation of risk; and > 1 corresponds to the overestimation of risk. After examining the product, consumers update their beliefs about product safety. It is costly to read labels. We assume that there is a constant marginal cost to a labelreading e ort, c > 0. In choosing the optimal e ort, consumers trade o the extra bene t from examining the label and the cost of doing so. The bene t comes from a decrease in the likelihood that an unsafe product is consumed. The probability that a consumer nds and correctly understands the warning depends on the label s transparency b 0 and consumer reading e ort x 0. The transparency of the warning b re ects its prominence, clarity, or readability (e.g., label font size). Higher levels of label transparency b indicate a warning label that is easier for consumers to nd and understand. We assume that, when present, the warning will be discovered and understood by the inspecting consumer with probability f(x; b) 2 [0; 1], which satis es the Inada conditions with respect to x. In particular, the discovery function is such that no e ort implies no discovery for any positive transparency level b > 0, a higher label-reading e ort is more likely to result in discovery of the warning, and there are diminishing returns to e ort. We also assume that for any reading e ort x > 0, the probability of discovery increases as transparency of the label improves and that there is complementarity between transparency and reading e ort. 7 [Figure 2 HERE] Figure 2 shows the timing of decisions by the rm and consumers. The rm chooses the amount of care y and consumers decide how much e ort, x, to allocate to reading the label. The product s harmful content a is realized and the rm posts the warning if it is required by the mandate (when a > a 0 ). Consumers do not know the producer care y or the level 7 Using notations, we assume the discovery function f(x; b), de ned for x 0 and b 0, is a twice continuously di erentiable function of x and b that satis es the following conditions: f(0; b) = 0; f x > 0 for any b > 0, 2 f f x < 0, and lim 2 x!1 x = 0, lim x!0 x For example, f(x) = bx with b > 0 and 2 (0; 1) satis es all these properties. f = 1; f(x; 0) = 0; f b > 0 for any x > 0, 2 f xb > 0; lim x = 0. b!0 f 6

8 of harmful content a, but their label-reading e orts may result in discovery of the warning. Based on the outcome of the inspection, consumers form an assessment of the probability that the product is unsafe and choose whether or not to buy it. In making their decisions, consumers know susceptibility, the strength of the liability l, mandated transparency of the warning label b, and the price of the product p. After we examine separately the consumer s and rm s problems in the next two sections, we will look for the Nash Equilibrium in a game where the rm and consumers simultaneously make their care decisions. 3. The Consumer s Problem To better assess the safety of the product, the consumer can examine the product s label for a warning. The rm is mandated to disclose that the product is unsafe (a > a 0 ) by including a warning on the product s label. In this case, the consumer discovers it with probability f(x; b), where x is the consumer s reading e ort and b is the label s transparency. In contrast, when the product is safe (a a 0 ), no warning is issued by the rm. After spending some label-reading e ort, the consumer forms a posterior probability that the product is unsafe. The detection of a warning is bad news for the consumer because the product turns out to be unsafe. In contrast, if the consumer does not detect a warning after spending some reading e ort, it is good news the product is safer than previously thought. Let P ND = P r(unsafejno Detection) denote the posterior belief of the consumer when a warning is not detected. In Lemma 1 below, we show that P ND (x; y) P e (y) < 1, with the rst inequality being strict whenever x > 0. The consumer payo in the case of detection, u D v p (1 l) 0, is lower than in the case of no detection, u ND v p P ND (1 l) 0, and one of three outcomes holds. First, it could be that the consumer buys the product regardless of the outcome of the inspection, because 0 u D < u ND and the consumer s expected utility is EU = u 0 cx, where u 0 v p P e (y)(1 l). This happens when consumer susceptibility is very low: D, where D v p 1 l. At the other extreme, for high levels of susceptibility, the consumer never buys the product because u D < u ND 0, and the consumer obtains cx. In both cases, we would obtain a corner solution for the consumer s e ort selection problem: 7

9 the consumer does not spend any label-reading e ort. Third, consider a more interesting case of warning labels that are informative in that they a ect the consumer s decision whether or not to buy the product. In this case, u D < 0 < u ND and such a consumer would buy the product if and only if the warning is not discovered. The (ex ante) expected utility of a consumer who spends e ort x in examining a product before knowing the outcome of the inspection and making the optimal purchase decision based on the inspection is: EU = cx + Pr (No Detection) max fu ND ; 0g + Pr (Detection) max fu D ; 0g, (1) which includes the cost of reading the label. Lemma 1 derives the expected consumer utility for the case when u D < u ND 0. Lemma 1. Suppose u D < 0 < u ND. Then, the expected consumer utility is: EU = u 0 cx + Bf(x; b), (2) where u 0 = v p P e (y)(1 l) is the expected consumer utility from buying the product without an inspection and B P e (y) ((1 l) (v p)) > 0 is the net bene t to the consumer from discovering the warning. A consumer maximizes the expected utility (2) by choosing a reading e ort x. An interior solution for the optimal label-reading e ort x > 0 is implicitly de ned by the rst-order condition: B f (x ; b) x = c: (3) In deciding on the optimal level of e ort, the consumer trades o the marginal bene t and marginal cost of reading the label. The marginal bene t is due to a higher probability of detecting a warning. The detection of the warning allows the consumer to reduce the likelihood of harm, but it also means that the consumer does not buy the product and therefore loses the net value v p. Let E be the susceptibility of a consumer who is indi erent between participating and not participating in the market, EU (x ( E ); E ) = 0. Note that E exists and is unique because deu=d = EU= < 0 at the optimal solution. Consumers with susceptibility 8

10 > E choose to avoid the product group altogether. They would choose not to participate in the market because it is costly to search for warnings and it is unlikely that the product will bring a positive net bene t given the risk they face. Proposition 1 describes the optimal consumer label-reading e ort for the consumer who participates in the market ( < E ) and nds warnings to be informative ( < D ). Proposition 1. Consumer E ort. For any susceptibility 2 ( D ; E ), there exists a unique interior equilibrium consumer label-reading e ort x > 0. It is positively related to the label transparency b, consumer susceptibility, and consumer risk perception, and it is negatively related to the net value of the product v p, cost of inspection c, and strength of liability l. The consumer reading e ort is negatively related to the producer care y. The consumer prefers a higher level of producer care y. The results are intuitive. The marginal bene t of reading the label in (3) is high and the consumer spends more label-reading e ort when the consumer susceptibility to harmful content is high, the consumer perceives the product to be risky ( is high), and the net value of the product v p is low. Quite intuitively, label transparency b also encourages consumers to spend more label-reading e ort. On the other hand, label-reading e ort is lower when the marginal cost of consumer care c and liability l are higher. The relationship between consumer and producer levels of care depends on how the labelreading bene t B is in uenced by the producer care. From the consumer perspective, care levels by the rm and consumers are strategic substitutes; that is, consumers reduce their care level when the rm produces a safer product. The lulling e ect is due to the fact that label-reading e ort is less bene cial to the consumer when the probability that the product is unsafe is lower. [Figure 3 HERE] Figure 3 shows how the equilibrium consumer e ort varies with consumer susceptibility. Consumers with D buy the product regardless of whether they nd the warning or not. Therefore, such consumers remain rationally ignorant. Consumers with 2 ( D ; E ) buy the 9

11 product only if they do not discover a warning. It is optimal for these consumers to spend some e ort reading labels and the optimal e ort increases with susceptibility. Finally, consumers with E never buy the product and therefore do not enter the market in the rst place. In what follows, we focus on the case of a homogeneous group of consumers with susceptibility 2 ( D ; E ) and normalize the number of consumers to one. Such a consumer chooses to participate in the market, potentially buys the product, and may bene t from reading labels. Suppose a consumer spends e ort x on reading the label. Then, the consumer demand for the product is: D(x; y) = 1 P (y)f(x; b); (4) because the consumer buys the product if and only if the warning is not discovered. Lemma 2 follows. Lemma 2. Properties of Demand. The demand for the product is negatively related to consumer care x and positively related to producer care y. For given levels of care, demand decreases in the label transparency b : D = D x; y; b + : (5) A consumer who buys the product after exerting e ort x faces the expected harm of H(x; y) = P (y) (1 f(x; b)) (6) because harm occurs when the product is risky, the warning label is not detected, and the exposed consumer is harmed. Lemma 3 follows. Lemma 3. Properties of the Expected Harm H. The expected harm is negatively related to consumer care x and producer care y. For given levels of care, the expected harm decreases with a higher level of label transparency b and it increases in consumer susceptibility : H = H x; y; b; + : (7) Lemma 3 shows that the expected harm is directly a ected by label transparency b and consumer susceptibility. It also depends indirectly on a number of parameters through the endogenous care decisions made by the rm and consumer. 10

12 4. The Producer s Problem A rm maximizes its pro ts by choosing the amount of care y to spend designing and manufacturing the product. The rm s costs consist of two components: production costs and liability costs. We assume that liability costs are proportional to the expected consumer harm and denote them by L = lh, where the strength of product liability l is weak. We start by looking at how a monopoly chooses the amount of care y, given consumer care, price p, and transparency of the label b. 8 The rm s pro ts are: = (p m(y)) D (x; y) lh(x; y): (8) Higher levels of producer care are associated with higher marginal costs of production and demand and lower liability costs. The demand depends on producer care because producer care a ects the distribution of harmful content in the product, making it safer, which reduces the probability of mandated disclosure. Since the rm knows the actual risk that its product poses, the likelihood of disclosure depends on the true distribution of harmful content. It follows, that under mandatory disclosure, the rm is not just minimizing the total costs but also has to account for the e ect of its care on the demand. This is in contrast with asymmetric information models of producer care, in which consumer demand typically does not depend on the care chosen by the rm because consumers make decisions based on their beliefs about the safety of the product and not on the actual safety (see an excellent discussion in Daughety and Reinganum, 2013). Let us examine how the rm s choice of care y a ects its pro ts: = D (x; y) m0 (y) D (x; y) + (p m(y)) l H(x; y) : (9) The rst term is the extra cost of care, which is proportional to the demand. The second term is the bene t due to the higher demand that is realized because of a lower likelihood of disclosure. The last term is the bene t from lower liability costs due to higher producer care. The pro t-maximizing level of care for the rm, denoted by y, is implicitly de ned by 8 In the analysis of the rm s decision-making, we choose to focus on the producer care decision and treat price p 2 (m(y); v) as exogenous. We also assume that the government sets a transparency standard b for warning labels and the rm complies with it. 11

13 the rst-order condition = 0. Note that when there is no liability (l = 0), the rm may still choose a positive care level y > 0 because of the disclosure mandate. Proposition 2 demonstrates how the parameters of interest a ect the rm s choice of care. Proposition 2. Suppose product liability is su ciently weak ( l is su ciently low). The pro t-maximizing level of producer care y is then a strategic complement to the consumer reading e ort x. Producer care y increases in liability l, label transparency b, price p, and consumer susceptibility to harmful content. The rm prefers that consumers spend less label-reading e ort x. Note that producer care does not directly depend on consumer risk perception, consumer valuation of the product v, and the marginal cost of reading labels c. These parameters directly in uence only consumer decisions. From the rm s perspective, whether the care levels of the rm and consumers are complementary depends on the sign of 2 x. In the proof of Proposition 2, we show that 2 x > 0 holds for su ciently low l: l < p m(y). For such low levels of liability l, the rm bene ts from selling an unsafe product, prefers that consumers do not read warning labels, and considers its care and consumer care to be strategic complements. We call the positive impact of consumer label-reading care on producer care a vigilance e ect. When consumers are vigilant, they search extensively for warnings and are likely to nd them, and in this case they do not buy the product. To prevent the reduction in the demand due to consumer label-reading e orts, the rm responds by choosing a higher level of care because this reduces the probability that the rm posts a warning under mandatory disclosure. When liability is weak, the rm primarily cares about this demand preservation, not the e ect of consumer care on the liability costs. 5. Equilibrium To nd the equilibrium in the model, where consumers and the rm simultaneously make their care decisions, we have to combine our ndings from the previous two sections. We will continue to assume that the liability rule is su ciently weak and that consumers are somewhat susceptible to harmful content, at least enough to care about warning labels but 12

14 still willing to enter the market. Speci cally, we assume l < p m(y) and 2 ( D ; E ). From the consumer perspective, the reading e ort and the rm s choice of care are strategic substitutes. For the rm, they are strategic complements. Therefore, the consumer s best response is downward-sloping while the rm s best response is upward-sloping. The intersection of the best response functions is the equilibrium pair of care decisions (x E ; y E ). From the proofs of Propositions 1 and 2 we know that: x = R x y; l ; b ; ; v; p; ; c (10) and y = R y x ; l ; b ; ; v; p; ; c : (11) We can then de ne the equilibrium expected harm H = H (x E (z) ; y E (z) ; z), expected utility EU = EU (x E (z) ; y E (z) ; z), and pro ts = (x E (z) ; y E (z) ; z), where z = (l; b; ; v; p; ; c) is a vector of all the parameters. We will next examine how alternative policy tools a ect the equilibrium levels of care, the expected harm, consumer payo s, and the rm s pro ts Changes in the Strength of Product Liability In this subsection, we discuss the e ects of changes in the strength of product liability l. The reason for advocating a stronger liability is to give rms larger incentives to produce a safer product. We say that an intended policy goal is reached when this is the case. From Proposition 1, we know that for any level of producer care, consumer label-reading e ort decreases in liability l. At the same time, by Proposition 2, producer care increases in l, keeping consumer reading e ort constant. Taking into account the strategic interaction between consumer and producer care established in Propositions 1 and 2, we nd that the overall e ect is a reduction in consumer care and an increase or decrease in producer care. Proposition 3. A marginal increase in the strength of product liability l results in lower equilibrium consumer care x E. Producer care y E may increase or decrease as liability rises. If producer care y E increases in liability, then consumers prefer higher levels of liability. If it does not increase, then the expected harm rises as the product liability strengthens. 13

15 Figure 4 shows what happens to the equilibrium as product liability becomes stronger, assuming the intended goal of the policy change is reached. The equilibrium moves from E 0 to E 1, with higher producer care and lower consumer reading e ort. [Figure 4 HERE] A stronger liability guarantees consumers a higher compensation and thus discourages them from carefully reading labels. This may reduce producer care through the vigilance e ect, but the rm also has an incentive to respond to stronger liability with a higher level of care. Overall, the equilibrium producer care increases in liability strength when the indirect vigilance e ect is not too strong compared to the direct e ect of stronger liability on the rm s incentives for care. Interestingly, stronger product liability can be counter-productive since it can also increase the expected harm. This happens when both consumers and the rm reduce their levels of care. A stronger product liability has a negative direct e ect on the rm s pro ts. However, there is a positive indirect e ect due to a reduction in consumer reading e ort. If the direct e ect dominates, then the rm favors lower levels of liability (in other words, the rm prefers a higher burden of proof in product liability cases). Changes in liability have ambiguous e ects on the equilibrium expected utility as well. When a stronger product liability reaches its intended e ect on producer care, consumers are better o. They enjoy higher producer care and obtain a higher compensation in case of an accident. If the consumer reading e ort were determined exogenously, then the rm prefers a lower liability standard, while consumers prefer a higher liability standard. The overall impact of a stronger liability on the expected harm would be negative. Thus, we show that the endogeneity of consumer e ort can change these predictions. A policy that only allows for supply-side responses may fail to achieve a desired outcome (such as a reduction in the expected harm) when demand-side responses are present. One has to think about the unintended consequences of changes in regulation and the liability system through their e ects on consumer care. 14

16 5.2. Changes in the Transparency of Warning Labels In this subsection, we discuss the e ects of changes in the transparency standard for warning messages. The intended goal of greater transparency of warning messages is to encourage consumers to spend more label-reading e ort. From the rm s perspective, higher transparency increases the probability that the consumer discovers an existing warning. For a given level of product safety, this implies a higher loss in demand. Therefore, when the government mandates higher label transparency, a rm responds by increasing its level of care. From the consumer perspective, the overall impact of higher transparency on consumer care is less clear. On the one hand, higher transparency directly induces more label-reading e ort. On the other hand, an increase in producer care that results from higher transparency indirectly discourages consumers from expending label-reading e ort. When the direct e ect on the reading e ort dominates the indirect e ect through producer care, the overall e ect of higher transparency on consumer care is positive and the intended goal of the policy is reached. In this case, the expected harm and pro ts decrease as labels become easier to read. In any case, consumers prefer more transparent labels. Proposition 4 summarizes the e ects of changes in label transparency. Proposition 4. A higher label transparency b increases the equilibrium producer care ye. Higher label transparency b may decrease or increase the equilibrium consumer care x E. If consumer care increases in label transparency, then both the expected harm and pro ts are lower when labels are more transparent. Consumers always favor more transparent labels. Figure 5 illustrates how the equilibrium changes as the transparency of labels b increases, assuming that the intended goal of the policy change is reached. [Figure 5 HERE] Proposition 4 implies that one of the ways to reduce the expected harm to consumers is to mandate a higher transparency of the health risk warnings. This can induce both the rm and consumers to choose higher levels of care. In this case, more transparent labels unambiguously decrease the expected harm. First, higher transparency makes it 15

17 more likely that consumers discover the warning, which directly reduces the expected harm. Additionally, it induces consumers to inspect the labels more attentively and stimulates the rm to choose a higher level of care, both of which further decrease the expected harm. In equilibrium, consumers favor more transparent labels because they are easier to nd and read and because higher transparency prompts the rm to o er a safer product. On the other hand, rm s pro ts tend to decrease with a higher transparency of warning labels as more visible warnings result in lost demand. However, higher transparency decreases liability costs. Under a su ciently weak liability standard, the rst e ect dominates and the rm prefers labels to be less transparent. What our results then suggest is that, under weak liability, rms may not have su cient incentives to make warning messages easy to nd and comprehend. When rms have an ability to choose the way the label is displayed and how transparent it is, they may not opt to make the warnings visible. This would explain why policymakers set the minimum transparency standard for product labels. For instance, FALCPA bans ne print and mandates labeling allergens in English rather than in Latin Changes in Consumer Perceptions Product risk perception parameter directly a ects only consumer decisions. If consumers underestimate the risk ( < 1), they spend less time and e ort on inspecting warning labels than if they had more accurate expectations. Risk perception does not directly in uence the rm s choice of care because its pro ts depend on the probability that the label is posted, which in turn depends on the actual risk level. Liability costs also depend on the actual rather than the perceived expected harm. Figure 6 shows the equilibrium e ects of a higher consumer risk perception. [Figure 6 HERE] Proposition 5. A higher risk perception by consumers results in a higher equilibrium care by the rm and consumers, x E and y E. The expected harm and pro ts are lower when consumers perceive the product to be more risky (higher ). 16

18 Intuitively, if consumers believe that the product is more risky, they will read labels more carefully. Under weak liability, this prompts the vigilance e ect the rm increases its level of care as well. As a result, higher consumer risk perception in equilibrium results in higher levels of care by both the rm and consumers. The expected harm is sure to decrease, as well as the rm s pro ts, but the (objective) expected consumer utility EU=1 (which is based on the true perception of risks, = 1) is sure to increase with because of higher producer care Comparing Policy Tools In Figure 1, we summarize the strategic e ects (the lulling e ect on consumers and the vigilance e ect on the rm) as well as the direct e ects of policy changes. The impacts of di erent policies on the equilibrium market outcomes are further compared in Table 1. The rst column lists a set of policy tools: the strength of liability l, transparency of the label b, and consumer risk perception. Columns (2) through (6) indicate how a certain policy (e.g., a more transparent label) a ects the equilibrium consumer and producer care, the rm s pro ts, consumers expected utility, and the expected harm. Negative or positive signs denote the corresponding negative or positive impacts of a policy on a target outcome and a question mark indicates an ambiguous overall e ect of a policy. [Table 1 HERE] Part A of Table 1 summarizes the impacts of various policy changes on the equilibrium market outcomes assuming that the label-reading e ort is xed. With exogenous consumer reading e ort, we nd that both a higher label transparency and a stronger liability always reduce the expected harm and have identical impacts on all market outcomes, including consumers expected utility and the rm s pro ts. Therefore, requiring more transparent labels seems to have the same e ect as strengthening product liability. Yet, as Part B of Table 1 shows, the endogeneity of consumer e ort is not to be overlooked. Our analysis of the equilibrium with endogenous consumer and producer care demonstrates that the impacts of alternative policies are more nuanced and that unintended consequences 17

19 of policy changes are present. Our framework also allows us to consider the impact of improving consumer risk perception, which may serve as a major tool in reducing the expected harm to buyers from consuming risky products. Our results demonstrate that policies aimed at boosting consumer care may have di erent implications than those aimed at promoting producer care. From the rm s perspective, producer care is a strategic complement to the consumer reading e ort. This means that a rm exerts more care if consumers spend more e ort on inspecting labels (vigilance e ect). At the same time, from the consumer perspective, consumer label-reading e ort is a strategic substitute to producer care, implying that an increase in producer care results in lower labelreading e ort. Thus, if policymakers target producer care, they need to consider the negative e ect on consumer care (the unintended lulling e ect ) and recognize that the overall impact of such a policy on the expected harm can be ambiguous. For example, a stronger product liability shifts the burden of care from consumers to the rm, increasing producer care and reducing consumer care, with an ambiguous e ect on the expected harm. If policymakers target consumer care instead, then there is an additional unintended bene t: a policy that stimulates consumers to spend more time reading warning messages also stimulates the rm to provide a higher level of care. For example, a higher label transparency reduces the expected harm directly and also indirectly by inducing higher levels of both producer and consumer care. An educational campaign that ensure more adequate risk perception of unknown risks not only stimulates consumer e ort but also gives producers incentives to invest in product safety. We also nd that alternative policies hardly appeal to everyone. Higher transparency of a warning message b boosts the expected consumer utility. At the same time, it tends to reduce the rm s pro ts. A stronger liability rule l tends to increase the expected consumer utility but has an ambiguous e ect on the rm s pro ts. This suggests that the optimal policy choice may depend on the weights assigned to rm and consumer gains. In terms of the expected harm, higher warning transparency and policies that ensure higher risk perception tend to reduce the expected harm from a product if product liability is weak. In contrast, higher liability shifts the burden of care from consumers to producers, with unclear e ects on the the expected harm. 18

20 We nish this section by examining from a social perspective whether equilibrium care levels are too low or too high. We de ne welfare as W = EU + and then analyze how the equilibrium consumer care x E and producer care y E of care x SO and y SO that maximize welfare W. compare to the socially optimal levels Proposition 6. In the equilibrium, producer care is underprovided compared to the socially optimal level. By contrast, consumer care is socially excessive if liability is low. Proposition 6 states that the rm underprovides care in the equilibrium. The reason is that the rm faces the social cost of increasing its care m 0 (y)d but captures only a share of the bene t. The bene t comes from a decrease in harm and an increase in the demand due to fewer warnings being posted. The rm captures only a share of each of these bene ts because of the non-appropriability of consumer surplus (v > p) and less than full liability (l < 1). In contrast, consumers label-reading e orts are excessive in the equilibrium when liability is weak. The consumer faces the social marginal cost of consumer care c. Another cost is due to demand reduction, which is higher for society as a whole since v m(y) > v p. The bene t comes from a decrease in harm, which is also higher for the society as a whole for l > 0. However, if liability is low, then the consumer bears most of the harm and consumer care is too high. We conclude that in the equilibrium, the producer care is too low compared to the socially optimal level and if liability is low, then the label-reading e ort is too high. x Another way to see that Proposition 6 holds is to realize that in the equilibrium W x = < 0 and W = EU > 0 since consumer label-reading e ort decreases pro ts (negative externality) while producer care increases the consumer expected utility (positive externality). These results have implications for the welfare-maximizing choices of liability and label transparency. Intuitively, stronger liability (assuming its intended goal is reached) reduces consumer label-reading e ort and increases producer care. Thus, both care levels move in the direction of improving welfare and thus a stronger liability is socially desirable. At the same time, we cannot determine if higher or lower transparency is desirable from the welfare point of view. Higher label transparency improves producer care but also leads to higher consumer label-reading e ort; it lowers expected harm and is bene cial to consumers but 19

21 hurts the rm s pro ts. 6. Conclusion This paper examines the joint e ects of safety regulation and product liability on consumer product safety. Since consumers may not often have adequate information about product risks, warnings posted on product labels help them make better choices. However, for labels to be e ective, they need to be read and understood by consumers. Food allergen labeling regulation is an interesting example of the mandated disclosure of product risks to consumers. The government requires that rms disclose the presence of major allergens in packaged goods and also regulates the display of the warning. In our model, a rm chooses a level of care that stochastically determines the safety level of the product. The rm has to place a warning on the product s label if the level of harmful content exceeds the consumer tolerance level. In this context, we examine how the rm s choice of care and consumers choice of label-reading e ort depend on the liability system and on the mandated transparency of warning messages. Importantly, both consumer and producer e orts are endogenous and interdependent. We nd that from the rm s perspective, producer and consumer care levels are strategic complements when liability is weak. Therefore, producers care more about product safety if consumers spend more e ort on inspecting the warnings. By contrast, from the consumer perspective, producer and consumer care levels are strategic substitutes. In other words, higher producer care has a lulling e ect on consumer care. These strategic considerations have implications for the choice of policy tools to manage product safety. A higher warning transparency (assuming its intended goal is reached) and higher consumer risk perception increase the equilibrium consumer and producer care and thus decrease the expected harm. By contrast, stronger product liability shifts the burden of care from consumers to the rm. This increases producer care and reduces consumer care, with an ambiguous e ect on the expected harm. Importantly, we argue that consumer safety-related e orts should be treated as being endogenous. Otherwise, a higher transparency of warning labels and stronger product liability would have similar impacts on all market outcomes, including consumer and producer care, expected harm, expected utility, and pro ts. 20

22 Ultimately, the answer to the question of whether a stronger liability and/or improved label transparency are desirable depends on the policy objective. If regulators are concerned with the expected harm and consumer payo s, then mandatory disclosure of harmful content with a requirement of high warning message transparency is likely to work. So will an information campaign that educates consumers about unknown potential risks in products. On the other hand, if the objective is the overall welfare, then stronger liability is the preferred policy tool. The rm also tends to favor stronger liability because transparency of labels stimulates consumer label-reading e ort and dampens the demand for its product. This work can be further extended in a number of ways. We can analyze alternative legal systems and allow the liability of the rm to depend on the level of producer care, consumer care, or both. This would give the parties an additional reason to take care, but the overall e ects are complicated by the presence of strategic interaction. We could also allow for heterogeneity in consumer susceptibility to product risks. Then, consumer participation in the market will be endogenously determined; that is, there will be a threshold level of susceptibility such that only consumers with lower levels of susceptibility would choose to participate in the market. We conjecture that the main results of this paper would continue to hold as long as the consumer response on the intensive margin (reading e ort) dominates their response on the extensive margin (decision to enter the market). It would also be interesting to study how litigation and regulation e orts interact when there is competition in the market. Among various policy tools, the design of the labels in term of the information content and display deserves special attention. According to the existing requirements for labeling allergens in packaged food, a rm needs to notify consumers about any amount of harmful ingredients in their products. Intuitively, a more informative label should promote consumer label-reading e orts and through the vigilance e ect increase producer care. Further analysis into the impact of a more precise disclosure rule tailored to the potential risk to consumers may prove fruitful in the design of labeling mandates aimed at reducing harm and increasing welfare. 21