It is assumed that, in the absence of a private standard, the market setting is one of incomplete vertical contracts, i.e., downstream food retailers and upstream food processors cannot sign enforceable contracts specifying the supply of a customized food product, the precise nature of the latter only being realized ex post (Hart and Moore, 1999)⁸. The possibility of a tort case is also ruled out by assumption, i.e., a food retailer is unable to determine that a tort (harm) has occurred, they cannot identify the tort-feasor (the food processor causing the harm), and as a result they do not sue in court for sanctions against a non-complying food processor⁹,¹⁰. This compares to the standard literature on suit settlement and trial where the existence of a tort and the identity of the tort-feasor are known, such that compensatory and possibly punitive damages can be assessed (see Kaplow and Shavell, 2002; Polinsky and Shavell, 1998) ¹¹.

To minimize post-contractual transactions costs, a private standard f i in industry i is established where a third-party certifier evaluates the product ex ante. Following Henson and Humphrey (2009), f i includes the following: (a) description of the production process firms in i must follow in order to comply with the standard; (b) verification of compliance with the standard through internal documentation by firms in i ; (c) mechanisms of internal audit so that firms in i can self-monitor their compliance; and (d) external audit of any firm in i by a certifier. It is also assumed that third-party certifiers are not subject to capture by the food processors they audit, and that their audits are random.

The private standard f i describes the product attribute q i food processors must comply with to satisfy the requirements of downstream food retailers, where q i = q f i , and q i ´ > 0 , i.e., product quality increases in the level of the standard. A food retailer’s reputation, indexed by the value of their brand equity B , is a function of upstream food processor(s) compliance with f i , where B = k q i and B ´ > 0 . Based on the earlier discussion of credence attributes, q i is drawn from a spectrum of final food consumer preferences for product characteristics, including food safety (pesticide residues/organic production), ethical production (animal welfare), right-to-know (GM ingredients), and sustainability (environmental/ eco-system impact). While consumers may suffer a loss of utility when food standards are not being met, the assumption here is that food retailers bear the aggregate cost of consumers boycotting their products/stores, along with the associated damage to their reputation.

Food processors in i , who are assumed risk-neutral, can choose whether to comply with the standard f i or not. If they fail to comply, economic damage d is incurred by food retailers in terms of the reduction in their brand equity B . By not complying with the standard, a food processor obtains a benefit b ∈ 0 , ∞ in terms of reduced processing and other costs, where b differs among firms in i , and has a positive continuous density g . with a cumulative distribution G . . For simplicity, the population of food processors in i is normalized to one.

Without self-reporting by food processors of their failure(s) to comply with f i , a certifier audits food processors with probability p , where the audit accurately establishes the private standard is or is not being met, each audit costing c . If found in non-compliance, the food processor is subject to a sanction s , i.e., it is not certified as meeting the private standard for a specific period, where the maximum level of the sanction s _ ≥ d . s is equal to the financial loss to the food processor of temporarily not being certified and being denied retail shelf space, plus any re-certification costs.

The certifier chooses the probability it will audit and the level of the sanction to maximize welfare, i.e., the sum of food processors’ benefits b minus the damage incurred by food retailers d due to non-compliance with the standard f i , plus the auditing costs c . A food processor will not comply with the standard if b ≥ p s , welfare being defined as:

W = ∫ P S ∞ b - d g b d b - p c (1)

where the first term are benefits of upstream non-compliance less the damage incurred by downstream retailer(s), and the second term is the auditing cost, the population of food processors (normalized to one) being examined with probability p , each audit costing c .

As in Becker (1968), the optimal sanction applied against any non-compliant food processor firm is s _ . In other words, if s * < s _ , the sanction imposed by the certifier could be increased and the probability of audit p lowered, such that the expected sanction p s remains constant, i.e., the level of deterrence is preserved, the first term in (1) being unchanged, while auditing costs, the second term in (1), are reduced, thereby increasing W .

Differentiating (1) with respect to W , and assuming s * = s _ :

d W d p = s _ d - p s _ g p s _ - c (2)

the optimal probability of audit p * being given by:

p * = d - c ∕ s _ g p * s _ s (3)

with the optimal expected sanction p * s _ determined as:

p * s _ = d - c s _ g p * s _ (4)

Interpretating Equation 4: the left-hand side is the economic loss due to deterring the marginal food processor, i.e., the firm would have gained b = p * s _ if they had not complied. The right-hand side of (4) is the net gain from deterring the marginal food processor, i.e., the damage d less the costs of deterrence.

There is already extensive literature on reasons why Becker’s (1968) result might not hold, including imperfect information about the probability of apprehension (Bebchuk and Kaplow, 1992)¹². However, it is worth noting here that the optimal sanction s * imposed by the certifier may be dependent on the extent of any remediation efforts by the food processor. For example, if a food processor fails to comply with a food safety standard, the damage d to a downstream retailer(s) could be mitigated by “clean-up” efforts on the part of the food processor e , e.g., shutting down a food production line in order to identify and remove hazards associated with food safety. For example, the US deli meat firm Boar’s Head recently shut down a plant in Virginia following a listeria outbreak that had resulted in nine deaths and consumers getting sick in 18 states (Jewett and Rosenbluth, 2024).

Necessarily, remediation efforts come at a cost c e , any remaining damages being d a , i.e., if there is no remediation, d > d a + c e . Following Innes (1999a), if there are large enough net benefits from remediation by food processors, d - d a - c e > 0 , the probability of auditing p should be raised, and the sanction reduced such that s < s _ , to secure the benefits of remediation. In other words, a higher probability of auditing increases the likelihood of remediation, generating benefits beyond those due to imposition of sanctions and deterrence of non-compliance¹³.

Malik (1990) and Innes (2001b) have also shown that when sanctions against non-compliance are increased, firms have an incentive to engage in costly activities to avoid being caught. For example, firms could lobby/bribe the auditor to turn a blind eye to their failing to meet the required standard. Specifically, when the sanction s rises, food processors not meeting the standard have a greater incentive to avoid being audited and caught. In this case, the optimal sanction should also be set below its maximum level, s < s _ and the probability of audit increased. The upper bound to the sanction s is one that equates the marginal benefit of increased deterrence with the marginal cost of increasing avoidance¹⁴.

If self-reporting of non-compliance by food processors is allowed, and no administrative costs are incurred by food processors through self-reporting, the sanction r imposed by the certifier should be no greater than the expected sanction applied to food processors that do not self-report, r ≤ p s . Therefore, food processors will report a breach to the certifier if and only if b ≥ m i n r , p s in which case, welfare becomes:

W = ∫ r ∞ b - d g b d b - p c G r (5)

The difference to Equation (1) being twofold: first, the lower limit of integration is r rather than the expected sanction of p s ; and second, the auditing cost is p c G r as opposed to p c , as only those food processors that do not self-report are audited, i.e., when r > p s . With a positive probability of auditing, p > 0 , a sanction of s , and r = p s , the same set of food processors fail to comply with the standard with or without the option of self-reporting, i.e., the integrals in Equations 1 and 5 are the same. Importantly, with self-reporting by food processors, certifier auditing costs will be lower by the amount 1 - p G p s p c . In this case the optimal auditing scheme is one where r = p s and s = s _ . If r > p s , there would be no incentive for food processors to self-report; and if r < p s , the probability of being audited could be lowered, resulting in individuals still self-reporting and paying r , but the cost of auditing would be reduced, thereby increasing W .

Substituting p s _ for r in (5), and differentiating with respect to p :

d W d p = s _ d - p s _ g p s - - p c s _ g p s _ - c G p s _ (6)

the optimal probability of audit p * being given by:

p * = d - c G p * s _ ∕ s _ g p * s _ s _ + c (7)

with the optimal expected sanction p * s _ determined as:

p * s _ = d - p * c - c G p * s _ s _ g p * s _ (8)

Like Equation 4: the left-hand side of (8) is the economic loss due to deterring the marginal food processor, i.e., the firm would have gained b = p * s _ = r * if they had not complied. The righthand side of (8) is the net gain from deterring the marginal food processor, i.e., the damage d less the costs of deterrence which now has two components: first, the expected cost p * c of exa-mining the marginal food processor who has been deterred from non-compliance, but is in the pool of food processors that could be audited; and second, the infra-marginal cost c G p * s _ ∕ s _ g p * s _ of examining food processors who do comply with a higher probability.

Again, drawing on Innes (1999a), if there are net benefits, d - d a - c e > 0 , to be gained from food processors remediating the damage resulting from their non-compliance, the sanction for self-reporting should be equal to the expected sanction from not self-reporting, r = p s , whereas for non-reporting food processors, the sanction for non-compliance should be set maximally at s _ .

The argument for this is straightforward: first, if r < p s , the probability of audit can be lowered without affecting the incentive to self-report, and undertake remediation efforts; and second, the Becker (1968) result holds, i.e., the sanction for non-reporting food processors is raised to s _ , while the probability of audit p is lowered, preserving the expected sanction for not reporting p s _ . As a result, the incentive for food processors to self-report is maintained, the benefits of damage remediation are realized, and certifier auditing costs are lower.

A similar result holds if food processing firms seek to avoid being apprehended for noncompliance with the private standard, i.e., the sanction for self-reporting should be equal to the expected sanction from not self-reporting, r = p s (Innes, 2001b). Essentially, self-reporting occurs before a food processor takes any avoidance action, ensuring that the costs incurred due to non-compliance, deterrence, and avoidance are all lowered. Importantly, compared to Malik’s (1990) earlier finding, with self-reporting, the sanction against non-reporting firms can be raised to its maximal level, without avoidance costs being incurred.

In the case of private food standards, certifiers typically apply a gradual system of sanctions against non-complying firms, starting with a warning, through removal of certification, to exclusion from the standard (Fuchs and Kalfagianni, 2010). In keeping with the literature on the economics on law enforcement, an additional sanction equivalent to “imprisonment” is also considered here. Specifically, it is argued that a certifier can threaten to publicly expose a food processor for failing to comply with a private standard, which could then result in a campaign by a nongovernmental organization (NGO) or activist(s), or both. Such a campaign would be designed to encourage consumers to boycott altogether the offending firm’s uncertified product(s), including where available through retailers who do not value compliance with the standard (Hatanaka et al. 2005). In addition, NGOs/activist groups are on record as advocating the use of third-party certifiers as a means of ensuring private standards are “…objective, transparent, and accessible to interested parties…” (Hatanaka, et al., 2005)¹⁵.

While there is an argument that firms in pursuing corporate social responsibility will seek to produce products that meet a higher standard, for which consumers are willing to pay a premium, firms are frequently pushed into taking such action by NGOs/activists¹⁶. For example, Starbucks, despite having established a reputation for corporate social responsibility in the 1990s, were threatened with a boycott in 2000 by the NGO Global Exchange if they did not sell and promote fair trade coffee (Argenti, 2004). There are other well-known examples of food processors and retailers responding to the threat of boycott over genetically modified (GM) content in their products and humane treatment of animals, including Heinz, Gerber, McDonalds and Burger King (Hudson and Lusk, 2004; Innes, 2006). Both Heinz and Gerber were targeted by Greenpeace in the late-1990s for their use of GM soy and corn products in their food/baby-food. In the case of MacDonalds and Burger King, the People for Ethical Treatment of Animals (PETA) targeted both in 2001 for purchasing eggs from producers operating with insufficiently large animal pens. Currently, 15 boycotts of food processing and retailing firms, including Coca-Cola and Kellogg’s are listed online (Ethical consumer, 2024).

The role of NGOs/activists in promoting boycotts of firms has been subject to analysis by economists. Fedderson and Gilligan (2001) examine the impact of an information-supplying activist on outcomes in a credence good market where consumers care about the operating practices of firms operating in a duopoly. Their model assumes that activists randomly monitor the specific operating characteristics of one firm, where these are either good or bad, neither being observable to consumers. Through monitoring, activists learn the quality-choice of that firm and then signal that knowledge to consumers who then make their purchasing decision. Activists can support an equilibrium where at least one firm supplies the high-quality good, even though consumers cannot observe quality even after consumption. In addition, depending on the degree of substitutability between goods, activists can support equilibria where either both firms supply high-quality, or low and high-quality goods are supplied. Therefore, activists may improve the workings of a credence goods market.

NGOs may also operate in a setting where the government is involved in standard-setting. Heyes and Maxwell (2004) examine the impact of an NGO in a competitive market where government sets a mandatory minimum standard and the NGO can confer a label on firms that voluntarily conform to their standard. Without third-party certification, only the low-quality good is supplied, the latter surviving in equilibrium if an NGO sets a voluntary standard. By comparison, a mandatory minimum standard ensures only a single quality can survive in equilibrium. It is shown that the voluntary label is more attractive to firms than the minimum standard, average quality being higher under the minimum standard. Given this result, Heyes and Maxwell (2004) show that a minimum standard is optimal when combined with a voluntary standard set by the NGO.

Alternatively, if the industry initially sets a standard and the NGO then pushes to increase the standard, Baron (2011) shows that the industry standard will be higher than in the absence of pressure from the NGO. In this model, firms can produce either a low-quality good, or a high-quality good with credence attributes, the standard being set by an industry credence organization and credibly certified by a third-party. The level of the standard is a function of the number of firms in the organization, and once collectively set, these firms compete in the high-quality segment of the market, while firms outside the organization sell the low-quality good. Preferences for the credence attribute are drawn from a uniform distribution of consumers. Baron (2011) models the problem as a four-stage game: first, the NGO demands the industry set a standard, after which the credence organization sets a standard; second, the NGO directs social pressure on the organization; third, the NGO and its target organization contest a campaign; fourth, given the outcome of the campaign, there is Cournot-Nash competition in the product market.

Suppose, therefore, that it is possible to stage a boycott of the product(s) sold by a food processor failing to comply with a private food standard, but such a sanction is costly to implement. The monetary sanction for damage inflicted on a food retailer is s 1 ≤ s _ 1 , and the sanction due to a boycott of the offending food processor is the monetary value of a permanent loss of retail shelf space for its product s 2 ≤ s _ 2 . The total cost of sanctions to a food processor is s = s 1 + s 2 , the cost of imposing a boycott being γ s 2 , where γ = η + π > 0 . The NGO/activists organizing a boycott of a specific food product incurs a cost h in terms of the lost opportunities for other boycott activities (Innes, 2006), and the food processor incurs costs π from contesting the boycott (Baron, 2011).

Following Baron (2011), the probability of a successful boycott campaign is given by ϱ = β η ∕ β η + π , where β > 0 reflects the public reputation of the food processor, the latter being more vulnerable to a boycott the higher is b . The NGO/activist A maximizes their expected utility E U A less their campaign costs:

E U A = ϱ q f i + 1 - ϱ g < f i - η (9)

where ϱ q f i is the probability that food product meets the private standard, and ϱ q < f i is the probability it only meets some publicly set minimum standard. In the case of the food processor, they choose the amount of resources π they expend on contesting the boycott in order to maximize their expected profits E Π :

E Π = ϱ Π q f i + 1 - ϱ Π q < f i - π (10)

The probability that an NGO/activist campaign succeeds is increasing in β , and decreasing in the costs c e of restoring food product quality to meet the private standard. In other words, the impact of a permanent loss of retail shelf space is a function of the value of the food processor’s reputation β , and how costly it is to restore that reputation.

From the Perrier water case, many food and drinks firms have learned how important it is to react promptly to safety and other issues relating to their products in order to preserve brand reputation. Following the discovery of traces of benzene in its mineral water, Perrier withdrew 160 million bottles of water from the market in 1990 at a cost of $150 million, their share price falling by 37 percent, the firm eventually being acquired by Nestlé, and taking more than five years to regain public trust in the brand (Caesar-Gordon, 2015). Perrier’s significant loss of market share and damaged reputation suggests that failure to manage a safety or other problem with either a branded food or drink product can have significant economic consequences (Kurzbard and Siomkos, 1992). Not surprisingly, Coca-Cola promptly and voluntarily recalled several Minute Maid drink products in late-2021 due to the potential presence of foreign matter with the potential for adverse health consequences (Shen, 2021).

Therefore, if a boycott is credible in the absence of self-reporting, then s = s _ 1 + s 2 , where s 2 > 0 , 0 < p < 1 , and with self-reporting r = p s . In addition to the reduction of auditing costs, food processors voluntarily reporting their non-compliance, the social costs of initiating boycotts can also be reduced. With self-reporting, the total sanction r = r 1 + r 2 = p s = p s _ 1 + s _ 2 , where r 1 and r 2 are the monetary and boycott sanctions respectively. If r < s _ 1 , the certifier should set r 1 = r 2 , and r 2 = 0 , i.e., no boycott(s) will be implemented, generating social cost savings of p γ s 2 . If instead r > s _ 1 , the certifier should set r 1 = s _ 1 , and set r 2 = p s _ 1 + s _ 2 - s _ 1 , the savings in social costs being 1 - p γ s 2 . In other words, the monetary costs of non-compliance, s _ 1 are applied with certainty rather than with probability of p , and the threat of boycott is reduced by 1 - p s _ 1 .

Essentially, a given level of deterrence due to the sanction r , can be achieved at lower cost, with a lower probability of audit, because food processors who do not self-report non-compliance face a greater sanction through boycott of their product(s). If there are benefits from remediation by food processors, the certain sanction applied to food processors that self-report should be set at r 1 < s _ 1 , while the maximal sanction should be applied to firms that do not self-report, i.e., they are subject to a boycott.