Regulating through disclosure: the case of food hygiene barometer ratings in China

Abstract This paper provides the first assessment of China’s twenty-year experiment with food hygiene barometer rating systems, originally developed in the West for publicly communicating the grades awarded by food safety inspectors to individual businesses. This approach to regulating through disclosure is often celebrated for efficiently ‘nudging’ improved business compliance by empowering consumers to make ‘better’ decisions, but little is known about disclosure-based regulation in China or other low- and middle-income countries. Combining policy document and quantitative social media analysis with key informant interviews (n = 35), we show that barometers have failed to improve hygiene in China’s rapidly expanding private food sector: more than 75% of restaurants in four diverse case-study localities remain merely ‘Adequate’ with many of those unable in practice to meet basic safety standards. This is because regulatory implementation has been hesitant and unreliable; consumers ignore or distrust barometers; and food businesses lack the capacity and competitive incentive to improve. That failure to empower consumer sovereignty and leverage business behaviour change, however, also reflects how barometers – despite their liberal individualist conceit - were adapted to China’s revolutionary ‘mass line’ traditions of societal supervision of government regulators as much as food businesses. We conclude that barometers – far from being a ‘light-touch’ alternative to command-and-control regulation- require significant governance capacity, which may be lacking in low- and middle-income countries that struggle to conduct even basic regulatory oversight. Disclosure-based regulation also requires high levels of economic development, formalisation and trust to inculcate consumer and business responsiveness to information disclosures. Finally, our paper contributes to debates about risk communication and regulation by drawing the novel conclusion that the conceits underpinning seemingly universal tools of regulating through disclosure, get adapted to national state traditions and norms in ways that are far removed from their origins.


Introduction
Walk into a restaurant in many countries across the world and you are increasingly likely to see that amongst the TripAdvisor-style signs on doors, walls, and counters, there is one that uses smiley faces, stars, or numerical scores to communicate the outcome of the restaurant's latest food safety inspection. Often called 'food hygiene barometers' , these signs are the public-face of official risk-rating systems that are designed to improve food safety and public health by using disclosure to 'nudge' businesses and consumers into making 'better' decisions (cf Thaler and Sunstein 2008). Barometers are expected to achieve those goals by helping customers make informed decisions about where to eat and creating reputational incentives for poorly scoring businesses to improve their performance, as well as freeing up time and resources for inspectors to concentrate on the worst performing businesses. Indeed, having spread across the world over the last decade or so -from Britain to Bangladesh -barometers have become something of a 'best practice' global norm.
While there is much international enthusiasm for barometers as tools of risk governance, relatively little is known about their adoption, implementation, and impacts across different country contexts. Limited research on their use within developed western economies suggests that these processes depend on nationally varying factors, such as the design and credibility of barometer systems, national regulatory styles, public willingness and capacity to exercise consumer sovereignty, and the power of market mechanisms to incentivise behaviour change (Fleetwood 2019;Jin and Leslie 2009;Patel and Rietveld 2021;Self and Rothstein 2021). Almost nothing is known, however, about barometers in low-and middle-income countries, where possibilities for leveraging consumer choice to improve business compliance are likely to be harder to realise than in advanced western liberal market economies and even the ideas that underpin such approaches may have little fit or relevance to national governance and societal norms.
In that context, we use the case of food safety barometers in China to explore the adoption, implementation, and impacts of regulating through disclosure in developing middle-income countries. Although China was a strikingly early and enthusiastic adopter of barometers to address long-standing problems with the "integrity, quality and safety awareness" of its food businesses (State Council 2017), our research shows their implementation has been partial, hesitant, and unreliable, and they have had little influence on consumer decision-making and business behaviour. Those findings draw on policy document and quantitative social media analysis and in-depth interviews with key state and business informants in four purposely selected case-study localities from across China. We detail those mixed methods below after a review of the international literature on the use of barometers. We then set out three sections of empirical findings about the history of Chinese policy experiments with barometers, consumer responses, and business behaviours. That is followed by a more general discussion of how barometers fit the distinctive cultural and institutional contexts of the People's Republic of China. We conclude by drawing out some wider implications for regulating through disclosure in other low-and middle-income countries.

Existing international research on hygiene barometers and food safety regulation in China
Food hygiene barometers are part of a long-term international trend towards using information disclosure to improve risk governance, ranging from the public rating of the quality of local hospitals to mandatory reporting of corporate greenhouse gas emissions (Griffiths et al. 2017;Tang and Demeritt 2018). These developments have been driven by demands for the disclosure of previously confidential regulatory compliance information to enhance the public legitimacy and deter the corruption of regulatory processes -following Judge Louis Brandeis' aphorism, 'sunlight is the best disinfectant' . As light-touch alternatives to command-and-control regulation, barometers also have much in common with Thaler and Sunstein's (2008) fashionable theory of 'libertarian paternalism' . That theory advocates the use of 'nudge'-style governance interventions, such as information disclosure, as light-touch ways of structuring and leveraging the choices of 'informed citizens' to achieve societally desirable outcomes, such as improving public welfare and combatting market failure.
The case for using food hygiene barometers to leverage change in the catering sector is simple enough. Catering is a heterogenous mixed-compliance sector, which even the most advanced economies struggle to govern. In the uk, for example, there are half a million cases of food poisoning annually with more than half of all foodborne disease outbreaks associated with catering establishments (Jones et al. 2017). The situation is substantially worse in low-and middle-income countries (Jaffee et al. 2019).
Food hygiene barometers were likely first invented within uk local government in 1997, when Norwich City Council disclosed the hitherto confidential food hygiene ratings they awarded to food businesses (Stanton, Burton, and Gooding 2008). While that experiment was quickly abandoned for fear of commercial sensitivities, over the next decade or so, barometers were reintroduced in the uk and taken up by other countries in Europe, North America, Oceania, and Asia (Fleetwood 2019). As they have spread across the world, barometers have assumed a bewildering variety of forms including numeric scores, letter grades, smiley faces, traffic-lights, and textual descriptions, all of which may be displayed in-situ and/or posted online with varying degrees of statutory mandate both across and within countries.
Some studies suggest that food hygiene barometers can shape consumer behaviours, increase business compliance, and decrease foodborne illness. Multiple uS studies, for example, have argued that consumers use barometers to make choices about where to eat and that barometers have improved businesses compliance (Cruz, Powell, and Fielding 2013;Leslie 2003, 2009;Mckelvey, Wong, and Matis 2015;Patel and Rietveld 2021;Rothbart et al. 2019;Wong et al. 2015). Likewise, studies in Finland and Singapore suggest barometers influence consumer choice and promote business compliance (Aik et al. 2018;kaskela, Sund, and Lundén 2021), while in Toronto, the introduction of barometers reduced critical hygiene violations (Thompson, de Burger, and kadri 2005).
However, other research questions the benefits of barometers. For example, studies in the uk, Canada, and New Zealand found that most consumers were neither aware of, let alone used, food hygiene barometers to choose restaurants (Barysheva 2020;Filion and Powell 2011;Food Standards Agency 2016). Moreover, one study of over 700,000 inspections in San Diego and New York concluded that increases in average ratings reflected grade inflation rather than improved hygiene and that barometers appeared to have no impact on foodborne illness rates (Ho 2012). Indeed, there is evidence that barometers based on past performance neither predict future compliance nor have much influence on the worst performing food business organisations (FBOs), while disputes over ratings perversely shift regulatory resources to reinspecting better performing FBOs (kaskela, Sund, and Lundén 2021; Patel and Rietveld 2021; Self and Rothstein 2021;Wong et al. 2015).
These conflicting conclusions should not be surprising since the impacts of barometers are likely to depend on a number of factors. For example, research shows that consumer responses are shaped, inter alia, by barometer design and mode of display, public trust in official information, and highly varied consumer experiences, demands, and expectations (Barysheva 2020;Filion and Powell 2011;Henson et al. 2006;Patel and Rietveld 2021;Wong et al. 2015). Likewise, business responses to barometers depend on their understandings, capacities, and/or willingness to improve within often highly segmented catering markets (Bavorova, Fietz, and Hirschauer 2017;Jin and Leslie 2009). One might also expect wider governance styles and norms to shape the application and impact of barometers, such as their underlying rationales or complementarity with other policies (Self and Rothstein 2021;ukSTC 2011).
With research to date focused largely on advanced western economies, this article focuses on China, which was a strikingly early adopter of food hygiene barometers in 2002, just five years after their first attempted introduction in England (MoH 2002). Now the world's largest economy by purchasing power parity, China's move from state ownership to a market economy has fuelled explosive growth in the private catering sector, which by offering relatively low regulatory and economic barriers to entry is now thought to account for 5% of GDP (CHA 2020).
That rapid growth has severely tested governance capacities and strategies, illustrated by numerous food safety scandals that have called the legitimacy of the party-state into question. Indeed, President Xi (2013) has suggested that persistent failures will invite "people to ask whether our party is fit to rule China".
Over the last 20 years, China has sought to solve food safety problems by repeatedly reforming its regulatory regime and adopting western 'best-practices' such as HACCP and risk-based regulation (L. Wang, Demeritt, and Rothstein 2022). While little is known about the adoption, implementation, and impact of food hygiene barometers, we might expect several factors to shape those processes.
First, disclosure-based regulation in China may face resource and capacity constraints common to low-and middle-income countries. While China's authoritarian state can mobilise vast fiscal and policing powers to achieve the top priorities of its leaders in ways their peers can only dream about , the efficiency of local public administration is undermined by endemic corruption (Wedeman 2022) and bureaucratic fragmentation within various lines and levels of the party-state (Lieberthal and Lampton 1992). Despite its impressive police powers, there are questions about whether the People's Republic has the distinctive administrative capacities "needed to effectively deploy state power through an information-based regulatory framework" (Bowen and Panagiotopoulos 2020, p. 205). Disclosure-based regulation requires auditing and assurance, information management, and risk communication skills beyond the traditional repertoire of regulatory enforcement agencies relying chiefly on the territorial logics of sovereign power rather than the more indirect mechanisms of disciplinary power by which advanced liberal societies are governed through the expression of free choice (Rose 1999).
Beyond state capacity, market structure may present other challenges for disclosure-based regulation in China. Like many low-and middle-income countries, China has a large informal sector, which provides essential employment and other services for millions of migrant workers that have poured out of the countryside since decollectivization (Huang 2009). Operating outside normal regulatory oversight, FBOs in the informal sector are typically small and economically marginal, competing more on cost than quality. As such they have limited capacity or incentive to meet higher regulatory standards of hygiene and safety (Blackmore, Alonso, and Grace 2015). Such problems are not confined to the informal sector. China's State Council (2017) acknowledges that most of the country's 11.8 million licenced food businesses employ less than 10 people with limited training and poor hygiene resulting in chronic food safety risks, but the sheer number of small FBOs, combined with their vital role in local economies and the ease with which operators can shift into the unregulated informal sector, makes strict enforcement difficult.
There are also questions about the willingness and capacity of Chinese consumers to respond to food hygiene barometers in ways that incentivize FBOs to improve their ratings. A wealth of research on consumers in the Global South highlights the importance of poverty as a limiting factor in consumer responsiveness to front-of-package labelling and other nutrition information (koen, Wentzel-Viljoen, and Blaauw 2018; Mandle et al. 2015) and that may also apply to mandatory disclosures by FBOs. While low incomes may limit the potential of food hygiene barometers in China to leverage consumer choice, the well-documented sensitivity of Chinese consumers to food safety concerns might be expected to increase the effectiveness of disclosure-based instruments (R. D. Liu, Pieniak, and Verbeke 2014;Peng et al. 2015). Research on Chinese consumers has shown them to be willing to go to great lengths to allay food safety concerns by seeking out information from food labels and other sources (kendall et al. 2019;A. Liu and Niyongira 2017;Soon and Liu 2020), building closer relationships with producer networks (J. Y. Zhang 2018), preferring larger and more reputable retailers (A. Liu and Niyongira 2017;Soon and Liu 2020), and paying more for certified food (R. F. Liu et al. 2020). Although such studies have focused largely on the wealthier and better educated consumers in China's large cities, they do suggest there is an audience in China for the kind of information communicated by hygiene barometers.
At the same time, public responses to barometers might also be shaped by distinctively Chinese patterns of trust in government. While citizens in many countries trust local governments more than national ones (Wu and Wilkes 2018), the reverse is true in China. Chinese public concerns about local government competence and corruption are widespread and historically engrained (L. Li 2004;Shangguan 2015), reflecting both the difficulties of policy implementation in a large fragmented state as well as deeply entrenched Confucian traditions that celebrate the Emperor's virtues while blaming failures on feckless local officials. One consequence of this "hierarchical trust" (L. Li 2004) might be that while national information disclosure initiatives may enjoy public support, policy success is likely to depend on the extent to which local government is trusted to provide reliable information.
Finally, the effectiveness of transparency as a regulatory tool is likely to depend on familiarity and fit with established governance traditions. Research suggests that transparency about the limits of safety can be difficult for states with strongly entrenched norms of public protection, where governments struggle to explicitly tolerate potential adverse outcomes as acceptable risks (Rothstein, Borraz, and Huber 2013). While transparency is not conventionally associated with China, the party-state has nevertheless introduced a number of transparency measures since the opening-up and liberalisation reforms in 1978 (Yang 2006). For example, the 2007 Government Information Disclosure Regulations expanded citizens' rights to access state-held information to enhance public legitimacy and internal accountability (A. C. Liu 2016), while other disclosure programmes were designed to help monitor local government, promote public participation, and improve regulatory performance (Stromseth et al. 2017;Zhang, Mol, and He 2016). Transparency initiatives in the People's Republic also draw on its revolutionary 'mass line' (qunzhongluxian 群众路线) traditions of collective consciousness-raising and mobilisation through public information for achieving party goals. Examples include Mao's Great Leap Forward, the 'Chinese Quality Long March' (zhongguo zhiliang wanli xing 中国质量万里行) campaign in the 1990s to 'name and shame' rogue traders, and more recently the "all-out people's war" on Covid-19 (W. Li and Yang 2005;Rothstein et al. 2022). At the same time, concerns for party legitimacy and public order have also led to selective and quickly reversible information disclosure policies in China (Stromseth et al. 2017). In contrast to western liberal ideals of individual rights to know and consumer sovereignty, the Confucian norms underpinning China's "transparency without democracy" (Tan 2014) emphasise paternalist state protection and beneficence coupled with public deference and collective duties (Perry 2008).

Research methods and data
To investigate the case of food hygiene barometers in China, this paper adopts a multi-method approach based on source and method triangulation of findings generated through policy document analysis, semi-structured interviews, and quantitative content analysis of news and social media from China's "contentious public sphere" (Lei, 2018).
We traced the introduction and evolution of food hygiene barometers and their relationship to wider Chinese government policies towards food safety and transparency by scrutinising official policy documents, interviewing five provincial-and national-level officials, and a media analysis. The media analysis used the CNkI (Chinese National knowledge Infrastructure) database to trace coverage of barometers and of food safety issues more generally in the People's Daily (人民日报), the official newspaper of the Central Committee of the Communist Party of China and largest newspaper group in China, and Guangming Daily (光明日报), another official organ of the party aimed particularly at China's educated elite. Searching for articles published from 2001 to 2021 containing the word 'food' (shipin 食品) captured 659 articles in the People's Daily and 413 in Guangming Daily, which were then screened to extract those focused on food safety incidents or problems. This set (People's Daily: n = 209, Guangming Daily: n = 130) was then thematically coded by the first author to trace the changing frequency of news themes over time following the iterative method of Escobar and Demeritt (2014).
To understand wider public discourse about food hygiene barometers, we also analysed social media commentary on China's dominant Weibo (微博) microblogging platform, a Chinese version of Twitter created in 2009 with 500 million active users (Sina, 2019). We used Boolean logic to combine search terms developed iteratively through trial and error to maximise specificity and differentiate posts about barometers from the oceans of general discussion of food standards and hygiene. Applying those terms (Table 1) to the Weibo archive between 30.03.2011 and 8.11.2021 produced 689 posts, which were individually reviewed to remove duplicates and record the number and type of 'likes' , reposts, and replies they attracted. The resulting set of 634 unique posts was then thematically coded by the first author using codes for their source, sentiment and content, whose internal and external validity was checked by the other authors.
To assess local implementation and impacts of food hygiene barometers, we purposely selected four case-study localities from across China to reflect its differing food cultures and levels of urbanisation and economic development ( Table 2). Within each locality we then randomly selected one case-study station for market regulation (SMR): the lowest administrative level for food safety regulatory inspection. We interviewed the Chief Inspector of each case-study SMR, as well as their superiors at district/county or prefectural level and the Chief Inspectors of five other SMRs in those localities about the adoption, implementation, and impacts of hygiene barometers. To encourage candour, interviewees were promised anonymity, and all interviews were conducted following a protocol approved by the king's College London Research Ethics Committee (REPF LRS-18/19-11247). In total, we interviewed 22 government officials, whose comments were triangulated against other sources. To understand consumer and business responses, we constructed a representative sample of restaurants, randomly selecting 10 points within each SMR and then focusing on the 3 restaurants closest to each point, as geo-located on Baidu. For each restaurant in this set of 120, we collected on 26-28.11.2021 the 20 most recent consumer reviews posted to Dianping (literally 'public reviews' 大众点评), the most popular restaurant review platform in China with over 600 million users in 2021 (Dianping, 2021). In total we collected 2004 consumer comments, which the first author categorized using consumer sentiment and business responsiveness codes, developed iteratively and checked for internal and external validity by the other authors. We also solicited interviews with the owner or manager of every third FBO in our set, successfully interviewing 13 FBOs (response rate = 32.5%) about their experience of the scheme. These materials from consumers and business were then triangulated against each other and then against the findings from government collected through interviews, policy, and news media analysis.

National adoption and local implementation
China first introduced food hygiene barometers in 2002 to help solve the problems arising from the explosive growth of private food businesses in its newly liberalised economy. China had joined the WTO in 2001, but with 146 food poisoning fatalities in that year alone (Gao, Zhang, and Zhang 2002), food safety was becoming a publicly "salient issue" for the Ministry of Health (MoH 2002), which defied its traditional governance tools of reactive enforcement and revolutionary-style special campaigns. Having learnt about risk-based regulatory inspection from uk Food Standards Agency delegates at a FAO/WHO meeting in January 2002 (Ye 2003), the MoH (2002) rushed out its own risk-scoring system just a few months later for prioritising food safety inspections across processing, distribution, and catering services. The MoH system drew on a mix of readily scorable indicators emphasising administrative processes and structural features as much as actual food hygiene, to grade those FBOs that passed as -A-Excellent; B-Good; C-Adequate. Further to that "reforming and innovative spirit" (MoH 2002), the MoH (2003) also mandated public disclosure of FBO ratings via "multiple channels", including potentially the news media and internet.
The MoH (2002) hoped disclosure would incentivise businesses to improve their hygiene practices, explicitly stating that "consumers have the right to know about the hygiene status of food businesses." That notable use of the language of 'rights' , which appeared to reflect western liberal-democratic norms, was complemented by the MoH's vision of how information disclosure could contribute to the national project of improving food safety through collective moral suasion and policing. As the MoH (2002) declared, "Openness and transparency can facilitate the societal supervision by consumers so as to promote and enhance the work on food hygiene." Local government officials were repeatedly exhorted to "extensively mobilise the society to participate. Let consumers understand food hygiene ratings in a timely, convenient way and make informed choices to protect their rights and interests and drive out poorly compliant businesses" (MoH 2003).
While the form of disclosure was initially left to the discretion of local government, a senior inspector recalled to us how the Minister for Health had been so impressed during a uS visit by New York City's store-front displays of restaurant inspection ratings that local regulatory offices were immediately ordered to "actively explore disclosure by signage systems" for all food processing, distribution, and catering businesses (MoH 2003). The MoH (2004) followed up that order in 2004 by standardising the signage system and requiring local regulatory authorities to "publish the results for the public on a regular basis". Food processors, over whom MoH shared regulatory authority with the Administration for Quality Supervision, Inspection, & Quarantine, were given the option of displaying their rating as part of front-of-pack labelling ( Figure 1a). However, caterers for whom MoH was solely responsible, were all required to post in some "eye-catching place" their food hygiene license and a sign (Figure 1b) showing their rating alongside the name of the local rating agency to promote "the societal supervision on food hygiene regulation" (MoH 2004). In this way, the signage requirements sought to mobilise disclosure for 'societal supervision' not just of food businesses but also of local inspectors.
Despite the Ministry's ambition to see hygiene barometer systems used across all food sectors, implementation by local government was slow and halting, going little beyond a few pilots in some localities. Then in 2008, experimentation with barometers in the processing and distribution sectors was halted after the government reorganised the entire regulatory regime. It stripped MoH of its shared responsibilities for those sectors in a bid to end fragmentation and increase agency accountability in the wake of the Sanlu milk crisis in which six babies died and 52,000 were hospitalized across China after consuming infant formula adulterated with melamine (Snyder 2016).
Oversight of catering inspections was given to the MoH-affiliated State Food and Drug Administration (SFDA), which, in 2012, made some minor revisions to the risk-scoring tool and redesigned the signage (SFDA 2012). The signage still recorded the averaged grade from the previous year as A, B, or C, but was supplemented by one of three 'smiley-face' icons to communicate the most recent inspection finding, all in a 'traffic light' colour scheme (Figure 2). While the addition of the face icons -inspired by Denmark's 2001 Smiley Scheme (MoFAF 2021) -made the scheme more sensitive to existing rather than just historic compliance, it also complicated interpretation of the signage. At the same time, the SFDA (2012) did little to clarify how FBOs should display their signs beyond being in an "eye-catching place, such as entrances or lobbies". Moreover, despite the number of Chinese internet users increasing by 9-fold since 2002 to 564 million in 2012 (CINIC 2003(CINIC , 2014, the SFDA did not require that local inspectorates make ratings available online. In rationalising the revised scheme, the SFDA (2012) no longer appealed to consumer rights-to-know and individual choice, but rather emphasised barometers as instruments for collective moral suasion and policing by "mobilising all sectors of the society to participate in the supervision by timely disclosure of the food hygiene ratings". Indeed, the SFDA required displays to include its national telephone hotline for reporting poor business compliance and local regulatory incompetence or corruption. That ideal of societal supervision was institutionalised by the Food Safety Law 2009, which created the right for affected consumers to "demand reparation worth ten times the price of [the substandard food product], in addition to compensation for losses" (art. 96). In turn, these shifts in the underlying philosophy and design of barometers as tools of societal supervision reinforced an already entrenched culture of reactive inspection, in which as the Chief Inspector of SMR II explained, "consumer complaints steer our attention much more than laboratory testing". He did observe that consumer complaints and the rise of professional 'fake fighters' , who made money from reporting low risk issues, added to the burden of local inspectorates (cf. Yee and Liu 2020). However, that Chief Inspector still saw value in complementing "government supervision and industry self-discipline [with] supervision from consumers and the media… It's very difficult to ensure food safety if you only count on us." This use of transparency as a whole-society system of social control was endorsed in 2015 by Bi Jingquan, the Director General of the Chinese Food and Drug Administration (CFDA), which had replaced the SFDA as a full ministerial agency overseeing the whole food supply chain following yet another regime reorganisation. Referring to the newly developed 'social credit system' for tracking misconduct by businesses and individuals, Bi (2015) declared, "Disclosing the information of substandard food is the best protection to consumers, the biggest deterrence to law breakers, the most powerful check on regulators, the most active guidance for public opinion, as well as the greatest contribution to the construction of the social credit system. " However, there was less enthusiasm for disclosure at lower levels of government. Interviews with all nine street-level inspectors suggested broad but shallow support for the basic idea of disclosing business hygiene ratings, tempered by significant doubts about the efficacy of the barometer scheme. As the Chief Inspector of SMR III said, "Theoretically, it allows customers to make free choices… That is a good goal and we want it to succeed so that restaurants will pay attention." However, he went on to say that, in practice, the scheme had done little to improve FBO compliance. As he complained, "…for years I have observed that consumers don't visit catering businesses according to whether they are A, B, or C, and businesses are not interested in gaining an A-grade… So [disclosure] is very superficial in its impact". After almost two decades of posting their ratings publicly, the vast majority of FBOs in our 4 SMRs were only graded as C-Adequate, with one SMR reporting no A-grade businesses at all (Figure 3).
Interviews with other chief inspectors suggest this pattern is typical across China and that actual FBO hygiene is often even worse than published ratings suggest. As the Chief of SMR II confided, "the actual situation-not the data reported to my superior-[is that] at least 60% of FBOs should not even be classified as C-class". He explained that the problem was that inspectors were reluctant to fail FBOs because of concerns about the social unrest that might follow from revoking licences or imposing inflexibly large fines.
Managerial officials at county-and prefectural-levels were even more ambivalent. They did acknowledge that the scheme might mobilise public supervision of businesses. As one prefectural-level official explained, it can help overcome our current weak capacities by changing our regulatory philosophy. We are moving from the old situation where government is the sole supervisor to engaging more actors and creating a co-governance situation.
They were concerned, however, that disclosure also exposed their performance to potentially uncomfortable scrutiny from the public and higher-level officials. As one county-level official said, It is also a way of monitoring our inspectors… First, have you inspected? Second, did you investigate the right problems? … For example, experts can investigate whether our inspectors are identifying existing issues in FBOs.
That ambivalence was particularly evident in relation to extending disclosure to online platforms, which has been limited despite the spectacular rise of internet usage in China. In response to increasing consumer complaints about poor quality online take-away services, the CFDA (2017) mandated such businesses display their ratings on online take-away platforms, following the model adopted for in-situ signs used by bricks and mortar restaurants. However, it was silent on whether lower-level governments should put all catering services inspection ratings online. Online listings of FBO ratings were available in just one of our four SMRs, but even then only for the handful of its A-graded FBOs.
One reason for that hesitant approach to online disclosure is that behind the scenes officials have been growing increasingly anxious about public confidence in food safety in the wake of the 2008 Sanlu milk scandal (Snyder 2016). As our analysis of food safety issues in official newspapers shows, coverage dramatically increased following that scandal but noticeably declined after 2012 (Figure 4). That pattern likely reflected a response to President Xi's (2013) warning that food safety had become a major test for the legitimacy of China's party-state. With food safety estimated to account for 45% of all rumours on social media (Guangming Daily 2016), newspapers were increasingly urging the public to work "hand-in-hand" with Despite their doubts about its effectiveness, some street-level inspectors told us they would welcome expanding online disclosure to give them more leverage over FBOs to improve their hygiene. As one inspector suggested "the national bureau could establish a unified online system, disclosing all the inspection ratings to the public" following the practices of many other regulators around the world.
However, managerial officials at county-and prefectural-levels were more hesitant. The views of this county-level regulatory official were typical: Because the media and society have significant concerns… more transparency could create problems for both businesses and government… The media just report problems without fully understanding them. There is a tension between promoting FBOs compliance and preventing social unrest, so I believe disclosure needs to be selective… We will only disclose if the centre requires it.
Like his peers, this official felt no pressure to go beyond minimum disclosure requirements and, indeed, only saw considerable institutional risk in doing so if full online disclosure prompted media criticism or public controversy for which he could be sanctioned under the cadre responsibility system (Peter 2015).

Consumer responses
In addition to official hesitancy towards disclosure, multiple lines of evidence suggest food hygiene barometers have limited influence on Chinese consumers. When asked about consumer responses to barometers, FBOs and inspectors consistently reported that consumers hardly seemed to notice them. The Chief Inspector of SMR III was typical in bemoaning that "although the ratings are displayed in every restaurant, customers don't notice them at all". A restaurateur with over twenty-years experience concurred: "[customers] ignore the sign. They usually just glance at the dining environment".
Social media analysis reinforces that sense of low consumer awareness of the food hygiene barometer system, which has also been noted in survey research on Chinese consumer perceptions of food safety in restaurants (Bai et al. 2019). Neither the barometers nor the official hygiene inspection ratings they publicly communicated were mentioned in any of the just over 2,000 individual restaurant reviews we analysed on the market-leading Dianping platform, which otherwise included many comments about restaurant hygiene such as "Old female cooks not wearing masks and quarrelling with each other as they cooked…I was really worried that they would contaminate my food." Similarly, searching the last 10 years of China's dominant Weibo micro-blogging archive found just 777 posts, reposts, and comments about food hygiene barometers from across the whole country. Since a single food safety incident could trigger more than 767,000 daily posts on Weibo (Hong et al. 2015), the relatively tiny volume of discussion about barometers suggests low public awareness. That conclusion was reinforced by the paucity of 'likes' or replies (less than 2 on average) attracted by the 634 unique Weibo posts about barometers, the vast majority (81%) of which were from government agencies (63%) or official media (18%) issuing or repeating official policy announcements about the food hygiene barometer system that failed to 'go viral' or attract much public attention.
Content analysis of the 157 Weibo posts and comments from individuals about food hygiene barometers highlights several other reasons for their limited influence on Chinese consumers ( Figure 5). Mistrust was by far the most common negative sentiment. One source of mistrust was the potential for corruption: "What's the point? It just gives officials a chance to take bribes. " Others doubted their reliability, questioning whether posted grades were up-to-date or highlighting mismatches with personal experience of dubious hygiene. For example, one customer described "waiting for my dumpling soup in a breakfast bar near my home… The elderly employees didn't wear uniforms. They reused empty food steamers without washing them… Loud sounds of coughing came from the kitchen. But the restaurant was awarded a green smiley-face [A-grade]. How can this be?" Doubts about the effectiveness of barometers were the second most common category of negative Weibo sentiment. Netizens directly challenged the policy premise that displaying inspection grades would prompt consumers to demand better hygiene. Ordinary people, explained one Weibo commentator, "won't pay attention. Who will trust it? The smiley face system just adds a symbol. But the public is really concerned about ingredient safety. And how can you ensure food safety without ensuring the raw materials are safe. It's useless!" Another said, "it's not just about smiley faces. You also need to integrate the performance into the [social] credit system." With the vast majority of FBOs graded merely as 'C-Adequate' (cf. Figure  3), other comments highlighted the lack of consumer choice as another factor limiting the impact of hygiene barometers ( Figure 5).
Consumer impact was also limited by basic communication design flaws. With over a billion internet users, a persistent theme of Weibo discussion was about the government's failure to post ratings online. Responding to a government announcement about the hygiene barometers, one person asked, "Can't you put the full list online?", while another added, "Government should disclose [ratings] on online third-party platforms…to suit contemporary citizens' habits". That concern was echoed by two Chief Inspectors we interviewed, one of whom noted, "when people search for restaurants, they prefer using online third-party apps, such as Dianping." Evidence about whether people notice and understand the standardised design of the physical signage was more mixed. Although a third of Weibo posts and comments from individuals indicated that they had noticed the physical signs, some also wondered whether people would be confused by the difference between the letter grade and smiley face (cf. Figure 2). Other communication design critics were snarky about the green smiley face indicating excellence, asking "Why are restaurants with excellent hygiene labelled as aliens " or "are you sure the hygiene is good when your face turns green?" Chief Inspectors we interviewed expressed consistent concerns about whether "lay people can easily understand" the barometer iconology or indeed appreciate their health implications.
There is strong evidence that food hygiene ratings had little influence on consumer decisions. The Chief Inspector of SMR II spoke for many of his colleagues when he explained, "consumers only care about whether food tastes good… everyone knows a restaurant that has poor hygiene, but is still popular because the food is delicious." That view is supported by our social media analysis. The most 'liked' Weibo post in our dataset, endorsed by 149 people, noted how "few consumers notice the grades and even fewer pay attention to it when deciding about whether to visit a restaurant." Similarly, consumer restaurant reviews on the Dianping platform showed consumers to be consistently more concerned with taste, value for money, service, and ambience than with hygiene, which was mentioned less frequently across all our case-study SMRs (Figure 6).
The apparently low priority accorded to hygiene, however, may not reflect indifference as much as public beliefs that restaurants licenced to operate by the government should be safe. That universal expectation was reinforced in the Chinese context by top leaders repeatedly describing food safety as a "sacred political duty" (Consulate General of the PRC 2016), given the state's constitutional responsibility under Article 21 to "protect public health". Although 10% of Weibo comments from individuals supported the basic idea of barometers, a slightly higher number (11%) objected to them in principle, often quite vociferously. As one exclaimed, "You classify for food safety? Seriously? Manage food safety properly!". Twelve Weibo posts specifically condemned barometers as a "face project" (mianzi gongcheng 面子工程); a colloquialism capturing purely performative activities that do little to address public problems. As one person asked, "is it true that even the red face means businesses meet the lowest standard for operation? If it isn't, why are those businesses allowed to stay open?" Another commented, "I thought food safety is a must." Such sentiments suggest many Chinese consumers were fundamentally ill-disposed to shopping around based on hygiene ratings.

Business responsiveness
With hesitant implementation and consumer indifference, businesses were not very responsive to the public hygiene rating system. As one FBO manager said of the posted ratings, "nobody pays attention. Who even looks at them?" Our social media analysis found frequent anecdotal reports on Weibo of businesses being careless about displaying their signs, for example, obscuring them behind large plants. One consumer in an eastern city recounted how "recently, I visited Y business and to my surprise, the smiley faces were covered by other colour printed figures that made them hard to see". Another consumer in a southwestern city complained that on one popular street of food outlets she frequently visited, "some restaurants cover the red faces with posters and advertisements… but people still go to them." Such practices were 'strictly prohibited' by the SFDA (2012), whose revised inspection system required inspectors to assess explicitly whether FBOs were displaying their hygiene ratings in an "eye-catching place" (SFDA 2012). However, as just one indicator in the scoring system, it made little difference to inspection outcomes or business responsiveness to them. As the Chief Inspector of SMR III observed, "because customers don't look for the display, businesses don't feel any impact on their profits, so they don't care".
Even if consumers actively looked out for the barometers, businesses felt little competitive pressure to respond. With the overwhelming majority of FBOs graded as 'C-Adequate' (cf. Figure  3), consumers had little effective choice and businesses had no incentive to improve their ratings. For example, in SMR III, over 360 out of 400 FBOs achieved just a bare pass, with only one restaurant securing the top A-grade. Our Weibo analysis found evidence of consumer frustration with the lack of choice (cf. Figure 5), with posts from across the country highlighting, for example, "All the businesses along X street are Cs" or "I searched [the online take-away] ratings one by one, but they were all Cs." Interviews suggested that FBOs were conscious that their competitors were all "at the same level" and so did not regard a C-grade as commercially disadvantageous. The Chief Inspector of SMR II agreed that FBOs saw no need to improve their ratings, arguing, "managers regard the inspection rating system as formulaic paperwork; they seldom ask about how they're scored. It doesn't have a big influence on their business." In the absence of market or moral pressures, he said, "there are few C-rated FBOs asking how to upgrade to B or A…" That is not to say that FBOs were immune to customer pressure via social media, which put a spotlight on poor hygiene and allowed FBOs to stand-out and win custom. Our analysis showed that in all four case-study SMRs many FBOs were responsive to customer comments posted to the Dianping 'Tripadvisor-style' review platform (Figure 7). Typical here was the speed with which one business in SMR I responded to a customer's concerns about dirty tableware by saying "we immediately communicated your complaints to our kitchen staff and told them to improve, " although it is of course hard to verify whether practices did in fact improve.
Lack of business capacity and willingness to improve was a final reason the barometers failed to live up to MoH (2003) hopes that disclosure would "induce FBOs to increase their investment in food hygiene 'software' and 'hardware'". It has been a condition of licensing since 1989 for all FBO owners or their managers across China to undertake at least 20 or 50 hours respectively of training, pass an exam, supervise the training of all their staff involved in food handling, and since 2011 take 40 hours of refresher training every year (MoH 1989;SFDA 2011). However, the inspectors we interviewed explained that knowledge in the sector was still woefully inadequate, not least given the relatively low entry barriers to small entrepreneurs with little education and skills. Many small FBOs often start out without even a mandatory food hygiene license until they are caught by an inspector, perhaps as a result of occasional national special enforcement campaigns, and few put much effort into training. In SMR I, where 90% of catering FBOs employ less than 10 people, the Chief Inspector explained, Indeed, inspectors frequently told us that FBOs simply did not understand the standards underpinning the rating system nor did they ask about them, and we found that some businesses paid such little attention that they were unable to tell us their own rating without looking at the sign themselves. As one FBO manager told us, "I don't know.
[Inspectors] have their standards, but how can I know about them?" All FBOs were interested in was whether they were going to pass and avoid fines, or worse, closure. Recognising that "the foundation of the food industry is weak" (State Council 2017), government has tried to expand training provision to raise standards and safety. Inspectors, however, despaired that it was almost impossible to educate and persuade most small FBOs to improve their food safety practices. As the Chief Inspector of SMR III put it, "They seldom do what you tell them to do. They just don't care… They carry on just the same, even if you tell them one hundred times!" Even those FBO owners who want to improve can find it prohibitively costly to secure a higher grade. Structural features-such as the physical fabric of the premises, facilities, and equipment -account for over 20% of the total inspection score for catering businesses (SFDA 2012), but their improvement can be beyond the reach of businesses that are often only at the margins of financial viability. As the Chief Inspector of SMR I explained, "there are many FBOs along our streets with small kitchens that are just long-narrow areas… They don't have fly-killer lamps let alone disinfection cabinets. …Even if they want to upgrade, the architecture prevents them. Is it realistic to demand that they knock-down the building and erect a new one?" Faced with these structural constraints, many FBOs were unable to respond to the information nudge as proponents of regulation through disclosure imagined that they would and should.

Discussion
China's strikingly early embrace of hygiene barometers may seem puzzling given governance transparency is not often associated with the People's Republic (Mol 2014), but it does reflect a wider experimental policy style that has been "a pervasive feature of China's economic transformation" in the post-Mao 'Reform and Opening up' era (Heilmann 2008, p. 1). However, despite initial enthusiasm within the Department of Health to borrow from, and indeed, go further than, international best practice by rolling-out hygiene barometer ratings right across the food chain, that ambition was frustrated by repeated regulatory restructuring in response to chronic safety problems in China's rapidly expanding food sector. Barometers have only been implemented in catering and even here the experience has been disappointing. Barometers have not empowered consumers to express their sovereign safety preferences in the market and leverage improvements in catering hygiene practices. Indeed, there is little evidence that Chinese consumers pay attention to barometers in deciding where to eat out, and FBOs feel little pressure to improve their grades. Twenty years after barometers were first adopted, most restaurants in our four, representative case-study localities across China have remained merely 'C-Adequate' , with many likely failing to deliver acceptable hygiene in practice.
One reason for this disappointing outcome are the compromises that have been made over health and safety in China's rapidly developing economy. With all levels of government prioritising employment and economic development (knight 2014), there has been spectacular growth in the catering sector, but it rests on "weak foundations" as the State Council (2017) itself acknowledges. understanding of, and capacities to improve, food hygiene are poor across the sector. Moreover, poor practice is informally tolerated by local inspection officials. They routinely grade FBOs that deserve to fail as 'C-Adequate' in order to avoid the social unrest and blame from top leaders that could come with job losses in the almost 12 million catering businesses that are often barely financially viable. As a result, most businesses compete on price and other non-safety related factors. These findings from China echo those in the wider literature suggesting that barometers have at best made good businesses better, but not bad businesses good  and are likely to be typical of many low-and middle-income economy contexts with limited will and capacity to enforce standards.
A second reason why disclosure of hygiene grades has failed to nudge hygiene improvements is that Chinese consumers tend to ignore them when deciding where to eat, despite their well-established tendency to actively seek out assurance from labels and other sources of information about the quality of all foods they purchase (kendall et al. 2019;Y. Zhu et al. 2020). Chinese consumers are not alone in ignoring hygiene barometers as research elsewhere has shown (Barysheva 2020;Filion and Powell 2011;Food Standards Agency 2016). It did not help that China's physical barometer signs were poorly designed and often difficult for consumers to find in the absence of clear and well-enforced rules on their display, factors that are likely to be common in other low-and middle-income countries that struggle with policy implementation. However, our analysis suggests that even if the signs were better displayed, consumer responsiveness to barometers would be inhibited by patterns of trust and expectations of the state that are distinctive to China. Historically ingrained Chinese public concerns about the competence, capacity, and trustworthiness of local government officials undermined, with some justice, consumer trust in the information barometers communicated. At the same time, our sentiment analysis of consumer reviews on the popular online Dianping platform also suggested that hygiene was less salient for consumers compared to taste, value for money, service and ambience. However, it is unclear whether that finding reflects consumers' limited willingness-to-pay for safety or their expectation that the state should assure safety rather than leaving them to choose how much risk to accept. Though state duties of care vary even across advanced liberal economies (Borraz et al. 2022), in China public expectations of state protection have a particular resonance both with the revolutionary collective welfare promises of the Party in the People's Republic and ancient paternalistic Confucian norms.
A third, and closely related, reason for the failure of barometers to drive improvement concerns the way in which barometers have drifted away from their liberal individualist origins. It is well known that as policy tools are transferred around the world, their motivating ideas, principles and practices mutate in sometimes contradictory ways (Borraz et al. 2022;Demeritt et al. 2015). Barometers are no exception. While they were initially wrapped in the language of consumer 'rights to know' , that conceit of facilitating 'consumer sovereignty' was soon filtered through Chinese revolutionary traditions of the mass line for "mobilising all sectors of the society to participate in the supervision" (SFDA 2012) of not just businesses, but local officials as well. That transformation expressed a Confucian emphasis on the collective duties of the public to enhance the solidarity of the 'state-family' (guojia 国家), rather than improve safety per se. Shifting the spotlight of barometers onto local supervision made officials uneasy, however, and they resisted making them more accessible by putting them online for fear of exposing their own performance to wider scrutiny and possible punishment under the cadre responsibility system. Central government also became concerned that online disclosure of widespread poor performance could undermine public confidence. That helps explain why government rolled back on initial promises to go beyond requirements for physical signage and restricted internet disclosure to online-delivery platforms.

Conclusions and policy implications
We draw three main conclusions from our research on China's experiments with regulating through disclosure of food hygiene barometer ratings. First, barometers have not empowered consumers with the ability to express their sovereign safety preferences in the market and leverage hygiene improvements in China's rapidly expanding private food sector. This is because regulatory implementation has been hesitant, weak, and unreliable; consumers ignore or distrust barometers; and FBOs lack the resources, skills, and competitive incentive to improve.
Second, beyond the case of China, we can also make some more general predictions about the prospects for food hygiene barometers and other disclosure-based tools of nudge-style governance in other low-and middle-income countries. Leveraging improved business compliance and safety through the disclosure of inspection outcomes depends on effective supervision to generate reliable information about the safety risks posed by food businesses. If China's powerful party-state has struggled to effectively mobilise the appetite and administrative capacities needed to gather and communicate such information, other, less capable states that struggle to conduct even basic oversight will find it all but impossible. In addition to governance capacity, disclosure-based regulation also requires certain levels of economic development and formalisation to inculcate business responsiveness to information disclosures. While low entry costs make catering an important vehicle for employment and endogenous development in low-and middle-income countries, such businesses often run on a shoestring with limited capacity to supply better hygiene, not least in the face of often large informal sectors offering cheaper alternatives. In turn, consumer responsiveness to barometers in low-and middle-income countries can be blunted by limited ability to pay for improved standards, and expectations -at least in countries with paternalistic state traditions -that it is for the state to assure safety, rather than for consumers to choose acceptable levels of risk.
Finally, a third, and more novel, conclusion is that the uses to which food hygiene barometers are put, the meanings associated with disclosure, and indeed, the criteria for policy success and failure, are nationally-specific. In the liberal democracies where barometers were first developed, these tools are designed to correct the information asymmetries causing market failure by nudging individual consumers to take hygiene into account when deciding where to eat. In China, by contrast, barometers reflect distinctively Confucian ideals of collective responsibility. Thus, for the party-state the value of barometers turns as much on their symbolic role in performing the norms of mass line mobilisation, public vigilance, and government concern as on their instrumental impacts on consumer choice, business behaviour, and food safety.
Our research suggests three avenues for further research. First, the generalisability of our empirical findings about food hygiene barometers could be tested in other localities across China. Second, there is also scope to test the external validity of our analysis for regulating other risks in China through disclosure, such as environmental pollution or product safety. Finally, we could also compare the domestication of disclosure-based regulation in China with other low-and middle-income countries, such as Vietnam with its similar Confucian and revolutionary mass line inheritance, or India with its very different state traditions of English common law, constitutional democracy, and Hindutva populism.

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

Ethics approval
Ethical approval has been granted by Research Ethics Office of king's College London (Reference No. LRS-18/19-11247).

Data availability statement
The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.