Does Corporate Social Responsibility Make a Difference on Labor Conditions? a Case Study in Southern China

Does corporate Social obligation make a difference on crunch conditions? A shift Study in gray China Yiu Por Chen Assistant professor Public Services ammonia alum Program DePaul University & IZA e-mail emailprotected edu Ph one 312/362-8441 telecommunicate 312/362-5506 1 Abstract The corporate favorable responsibility code of illume a bun in the oven (CSRC) has been the symboliseative of corporate organization (CG) consumption by multinational corporations (MNCs) for their suppliers.Using the act price perspective, this breeding argues the CSRC system whitethorn reduce just aboutwhat execution be for MNCs in planning and holding corporate social responsibility (CSR). However, the CSRC whitethorn suffer from self-centeredness at the supplier take, which whitethorn undermine CSRC outcomes. By utilizing a linear breaker point rejoinder example with a randomize muckle of pulverization proletarians of 12 roleplay manufacturers in Southern China, this deal provides a unified verse to evaluate the gistuates of CSRCs on comprehend conditions from workplaceers voices.Further much, this eng duration offers fresh empirical evidence to guide the ways expedience whitethorn undermine CSRC effects on elbow grease relations. JEL code M14, M23, J28, J50, J81, J83, J88 Keywords Corporate Social province, Multinational Corporations, Outsourcing, condom, fagManagement Relations, Working Conditions, Workers Rights, fatigue Market Policy 2 on that point ar lots of firms and lots of changes in diligence movement intrusts among them, so t here(predicate) is no lack of selective information.To judge whether figure outal roil marketplace rules cook worse employment outcomes, as claimed by orthodox analysts, one could contrast employment between firms with more or less exacting internal rules. ? Richard Freeman (2005 19-20), suggesting that a microanalysis of workers and firms will be a travel plan forward to cons u nbenting the effect of press institutions. 1. Introduction Corporate social responsibility (CSR) has been rapidly institutionalized and now covers the vast major(ip)ity of industrial sectors approximately the globe.Donaldson (1996) high uplights that 90% of on the whole Fortune 500 companies collapse ceremonious codes of view, which be by far the close ordinary mode of CSR utilize by corporations today. Furthermore, 51% of German firms claim to waste codes of conduct, comp ard to 41% in the UK, and 30% in France (Schneider and Barsoux, 1997). Brytting (1997) in each case establish that 52% of the larger Swedish companies had codes of conduct. In fact, CSR is largely the response of multinational corporations (MNCs) to the accusation of being the major actors in new exploitations of less split uped countries (LDCs) (Chan, 2001). For the outsourcing MNCs, one of the important aspects of CSR is to visualise that workplaces and get the fit conditions in poor nations are non too grave (even if it is worse than in the workplaces of developed countries). Theoretic altogethery speaking, the overlord arise to CSR suggested in the literature is a s vexholder approach. The stakeholder approach advocates that firms should non only maximize attain entirely in addition be cast off as right-hand(a) citizens of the community and take account of the needs of some former(a)wise stakeholders that whitethorn be affected by the firms proceeds.This approach to CSR involves on the whole relate stakeholders in the development and capital punishment of the CSR (Freeman, 1984). In reality, the corporate brass 1 CSR in general is closely conjugate with principles of sustainable development, that is, enterprises should be obliged to make conclusions based not only on the financial/economic factors but also on the social and environsal consequences of their activities. In toll of application condition issues related to the outsourcing process, there are numerous reports and news or so the exploitation of parturiency.The interest reader push aside visit some struggle NGOs websites, for example, www. AMRC. org, China outwear Watch, ILO, etc. 3 (CG) approach (or firm-centered approach), which uses corporate social responsibility codes of conduct (CSRC) to regulate the suppliers operations in LDCs, is more prevalent. 2 From an institutional speculative perspective, this article argues that the high movement live incurred prevents the stakeholder approach from bringing all stakeholders together and developing strategies to deal with the externality produced by the outsourcing of drudgery.Instead, the current CSR practice has reduced the original stakeholder approach from the CG approach to CSR, the CSRC. In practice, composition the CSRC whitethorn get around the achievement cost of getting all stakeholders together and reduce potential conflicts that may arise during the CSRC planning and implementation process, this CSRC may be vulnerable to egocentrism (or agency puzzle) at the supplier directs (Jiang, 2009). As the codes are externally imposed on the suppliers, they heap exploit the knowledge asymmetry to disregard some of the codes.Because factory inter pick ups and perspectives for suppliers counsel could hardly notify the true faces of apprehend conditions, an effective measure to reveal workers voices at suppliers aim is called for. However, systematic survey data are needed to verify the harshness of CSRC to repel conditions. In addition, there is no unified method acting to understand the effect of CSRC on labor conditions. 3 This article provides a metaphysical argument and a unified empirical approach to analyze labor conditions under the CSRC and the egoism effect at the supplier side.Utilizing a unidimensional degree response theory (UIRT) impersonate, this study looks at the CSRC effect on 16 labor conditions in three dimensions labor dicker and d defenseles s rights, violation of truth, and guard and animate environment. These labor conditions data are generated from a randomized survey of 12 toy factories in Southern China, including 10 CSRC suppliers for McDonalds, Mattel, Disney, and Wal-Mart, and 2 non-CSRC factories as a comparison convention. 2In particular, some MNCs and crinkle knowledges would use Corporate Social Responsibility Codes Certifications (CSRCC) as measures to regulate suppliers conduct, such(prenominal) as Nike and the like. 3 For example, the Office of peasant struggle, obligate Labor, and Human Trafficking at the say-so of worldwide Labor Affairs, U. S. Department of Labor mentioned in the Chinas country report that Because of Chinas repressive political system, it was not possible to view as whatsoever information directly from China.There are no Chinese non-governmental organizations (NGOs) sprightly in this area, and foreign NGOs do not have access. Accordingly, it was impossible to keep going or reject allegations by various labor and human rights organizations that baby birdren are working in export industries which produce for the United States market (Office of Child Labor, Forced Labor, and Human Trafficking, Bureau of global Labor Affairs). 4 The UIRT manikin effectively examines the relative sincereness of a labor chore with the labor kernel fuss, which is the some serious labor chore found in this study.It guides that, low gear, the CSRC may have a positive effect on labor conditions. CSRC suppliers have less serious labor condition lines than factories with no CSRC. Second, because diametric CSRCs may have incompatible coverage and supervise intensity, contrary CSRCs may offspring in different labor condition issues. Non-CSRC factories are found to have six-spot serious labor puzzles McDonalds, Mattel, Disney, and Wal-Mart have four, 2, one, and zero in serious tasks, reckonively.Third, suppliers working with more than one CSRC may reduc e the possibility of opportunism at the supplier aim. temporary hookup suppliers with 1 CSRC would have 2 occupations as serious as the labor coupling riddle, suppliers with 2 CSRCs would have no such problem. However, the CG approach to CSRC may suffer from opportunism. This study documented some typical suppliers measures to disconcert CSRC auditing fervidness drill, skilled motion, self-aggrandizing a holiday, change the factory, and refuge drill.In addition, the subsequent analysis shows that different suppliers measures to fence in the CSRC may represent a diverse aim of opportunism. The subsequent percentages are organized as follows the second section explores the effect of the stakeholder approach and CG approach to CSRC and their effects on labor conditions the tercet section describes the survey, basic observation from the data, and the methodology the quartern section provides the estimation declarations the last section discusses the conclusions and dra ws some conclusions. 2.Stakeholder versus corporate governance approaches to CSR and labor conditions A transaction cost perspective With the globalization of production in recent years, MNCs from developed countries migrated their manufacturing (high labor cost) operation to LDCs for the sake of cost minimization or to be closer to the consumer market. Due to the regulations of the host government, their production is usually outsourced to topical anesthetic suppliers. The intense competition among suppliers may induce overthrow bidding price for outsourcing deals from MNCs.This low bidding price might, in turn, result in the downward pressure of labor cost, refuge equipments (Jiang, 2009 Pun, 2005a b). As a result, the globalization of production may even make working conditions weaker than it was out front the market opened ( lee side, 5 1995 Chan, 2003). Such worker conditions have been describe in mass media, generating concern from consumers, NGOs, policymakers, and acad emics. For example, in China, labor relations have been changing since the market reform began (Zhu and Warner 2004 a b). Workers usually suffered from new exploitations (Chan, 2001). In response to the recent surge of rushs from NGOs, media, and consumer movements, outsourcing MNCs corporate social responsibility (CSR) has been knowing to meet the consumer deal for a clearer product with fewer exploitations and/or a wagerer working environment and safety for the LDC worker. 5 In fact, the original stakeholder theory was mapped out by Freeman (1984). The theory attempts to ascertain which groups are stakeholders in a corporation and olibanum deserve management attention. A lodges stakeholders are therefore all those who are influenced by, or dope influence, a companys decisions and actions.This theory intends to address the commandment of Who or What Really Counts. A firm, as a member of the stakeholder, butt jointnot detain without fetching into account the needs of th e other stakeholders (Freeman, 1984). Figure 1 master copy stakeholder exercise (Freeman, 1984) 4 Richard B. Freeman (1994) conducted one of the earliest surveys on labor stocks under globalization of production. For further CSR code description, occupy go to http//www. csreurope. org/ 5 Bjorner (2004) shows that consumers would prefer to pay more for a clean product in the market.See also Harrison and Scorses (2003) study on how globalization impacts compliance with labor timeworns. 6 The CSR problem may be viewed from the institutional perspective when a firms operations create externality, the stakeholders, those who have been affected by the externality, could bargain with the firm and achieve welfare improvement as bulky as the property rights of each party are sound delineate and the information cost for negociate is low (Coase, 1960). on a lower floor these conditions, an agreement that guarantees the interests of stakeholders could be made.In the case where the ext ernality is produced by MNCs outsourcing, the CSR is therefore an institution (or the rule of game) that may help to remedy past and potential conflicts of interest among stakeholders and produce a more sustainable future development. However, the stakeholder approach of CSR may not be sustainable due to the compl run lowy of the agreement, the high cost of property rights identification, and the high information cost of bargaining. First, there is question about the willingness of MNCs to de bespeak a valid CSR practice that would help alleviate workers conditions at the supplier take (AMRC, 200410).Second, it is sometimes difficult to identify all the stakeholders because they are often topically opinionated. The transaction cost may be significant as stakeholders who can work together are needed to conceptualize and implement CSR. 6 Third, even when one can draw a rough express about potential stakeholders, the question is how these really different usually polarized stake holder views can successfully collaborate from the drafting the CSR to its implementation and admonishering (Dubinsky, 2002).The stakeholder approach to CSR is thus considered to have a higher transaction cost than the CG approach because it has to find out all stakeholders and bring them together to develop the CSR design. In general, the CG approach to CSR is an grammatical construction use to describe what some see as a companys obligation to be sensitive to the needs of all stakeholders in its business operations (Hill and Jones, 1992). 7 This is a top-down approach that aligns with the MNCs fashion (Husted, 2003).However, in the case of outsourcing decision, agree to Williamson (2008), the choice of form of governance is undefended to transaction costs due to asset specificity, uncertainty, frequency and their transaction costs, among others. These transaction costs would consequently determine the form of outsourcing (market or vertical integration). 8 The CG approach to CSR thus coincides with the outsourcing form and uses external social auditors instead of relying on stakeholders to proctor the behaviors of suppliers. 6 Murray (2002) suggests that even when companies have elegant maintainments of principle, the complexity is the account skill.Local-level appointment is needed and involves a complex regime of stakeholders (Murray, 2002 41). 7 Most of the CSR has uniform objectives. See, for example, Leipziger (2001) for SA8000, one of the guides to the new CSR code. 7 Figure 2 Corporate governance approach the firm is at the center. Source Fassin (2008) Figure 5. In practice, the CG approach to CSR usually employs the CSRC as a tights to accountant the behavior of suppliers. 9 From the MNCs point of view, the CSRC may have less uncertainty and may be a more cost-effective market solution. First, MNCs can use their subsisting internal system to reminder the suppliers.Second, MNCs can employ external specialists such as auditing firms inst ead of the stakeholders to plan and monitor the implementation of CSRC at the supplier level. 10 One of the potential pitfalls of the CG approach to CSR is that even when the MNCs aim to implement the CSRC for these suppliers/suppliers, monitoring issues prevail and are subject to opportunism (Brown, 2002). 11 In some cases, the CG approach to CSRC has been criticized as a lame duck by labor NGOs, even if the MNCs are indeed willing to safeguard better working conditions for 8Williamson (2008) has extended the transaction cost economics approach to outsourcing decision of a firm. The contractual schema suggested that key attributes of transactions such as asset specificity, uncertainty, and frequency and their transaction costs will determine the form of outsourcing (market or vertical integration) (Williamson, 2008 8). The schema assumed the outsourcing form would be determined and stabilized as long as the property rights are well defined and reliably enforced by courts. 9 See Kot ler and Lee (2005) for their suggestions on victimization CSR from a corporate governance point of view and from a marketing perspective. 0 See, for example, the social auditing of Pricewaterhouse. Also see the judicature of SA 8000 by Social Accountability International (SAI) as a saucily formed labor standard that responds to the need of MNCs code of conduct. Their website http//www. sa-intl. org/index. cfm? fuseaction=Page. viewPage=473 11 Dara ORourke (2001, 2004), an urban planning professor at U C Berkeley, has shown that there has been a problem in implementing the CSR code to the workshop level and the codes impacts on workshops working condition and safety may be not as reckoned. 8 uppliers workers. This is because of the existing monitoring problems at the implementation level . 12 There may be devil basic problems at the supplier level. First, the suppliers may not want to accede with the procedure that the MNCs require (AMRC, 2004, 65 Jiang, 2009). There are many mea sures from suppliers to MNCs CSRCs. Second, the social auditing monitoring procedures may be biased and possibly diluted by suppliers measures (ORourke, 2002). 13 canvassors sometimes forego out some important factors during their probes (See ORourke, 2000, 2001, 2002). 4 This leads to agency problems associated with CG approach to the CSRC (Goodpaster, 1991 Winn, 2001). In the CG approach, there is high monitoring cost and information asymmetry among stakeholders, especially between local anaesthetic suppliers and the MNCs. 15 In this regard, the CSR movement also encountered challenges about its asperity at the divinatory and implementation levels. 16 To validate the theoretical discussion, the riseable hypotheses are as follows H1 If the CSRC is good for workplace conditions, consequently we should see suppliers with CSRC that have better labor conditions than no CSRC factories.H2 In addition, if a different CSRC content has a different effect, we should expect a systemat ic difference among CSRC firms. H3 Moreover, if the CSRC suffers from opportunism, different measures to get around the CSRC at the subcontract level may affect CSRC effects. 12 In fact, a count of social auditors have been created in response to the new demand from MNC. However, the incentive for these auditors to reveal the true situation at the workshop level is in doubt (ORourke, 2000). 13 The Designs of Codes vary greatly across companies and industries.For example, the mankind Worldwide Responsible Apparel Production (WRAP) certification requested that the autarkical monitors be paid by the company under investigation. That causes serious doubt of the independency of auditor. Also, there is no requirement to hearing workers during the factory investigation (Jenkins, 2002 26). 14 ORourke (2002) investigates the social auditing process of Pricewaterhouse Coopers in a plume factory in Shanghai. Pricewaterhouse Coopers used a standard factory monitor process for factory manag ement, but not for workers.He also shows that there may be management bias and f jurisprudence reporting on the auditing process. The management bias implys failures to collect information from workers and failures to access restraints on license of affiliation. 15 Dubinsky (2002) documented the garment worker of suppliers in L. A. for supposition. While the CSRC in GUESS was intended to enhance workers condition, the workers interviewed reveal no material improvement in their working conditions and were threatened and penalized when they speak out. 6 For more information of the consultants for CSR code, please visit http//www. thecsrgroup. com/ 9 H4 Finally, if CSRC can increase transaction cost of opportunism at the supplier side, more CSRC would reduce the effect of opportunism. To assess the validity of these hypotheses, this study employs data generated from a random survey of workers from 12 suppliers of mark toy MNCs in Shenzhen and evaluates the CSRC effect using a unid imensional item response theory sit down. The following section describes the survey and the methodology of this study. . muckle and Methodology This section will discuss the survey and the methodology used this study. Section 3. 1 provides details on the survey design and the consume frame. Section 3. 2 compares the CSR code of conduct of branded companies with the mandate of the International Council of represent Industries (ICTI) and International Labor geological formation (ILO) conventions. Section 3. 3 provides a rank of CSRC t communicates in terms of level of hindrance. Section 3. 4 demonstrates the methodology used in this study. 3. The survey The survey was conducted at the Shenzhen and Dongguan areas in Southern China, both highly concentrated with labor-intensive industries, in particular, toy factories. 17 The survey, which took place from June 2006 to March 2007, sharpened on the toy industrys CSRC. To systematically examine the impact of CSRC on labor condition s, this survey for the first time used the industrial phone book and place a sample distribution frame before the randomized selection of 12 toy factories in Shenzhen and Dongguan areas at Southern China.The selection of the sample is as follows Their size should be comparable (usually around snow0 workers in each factory). This study identified 10 code factories in spite of appearance which there were eight suppliers with single CSR (Mattel, McDonalds, Disney, Wal-Mart) and two suppliers with two CSRCs (Disney & Wal-Mart, Mattel, and Wal-Mart). The survey also included two suppliers with similar technologies that produce only for the domestic market as a comparison group. 20 17 Please see Appendix 1 for the background of the setting, toy industry in China, and the reasons for choosing the sites in this study. 0 workers at the entry level were interviewed in each factory, and all the workers were interviewed anonymously. 18 A total of 240 workers were surveyed in this study. 19 3. 2 A Comparison of Corporate Social Responsibility Codes of Conduct put over 1 shows a comparison between the masterys of the study firms CSRC, the International Council of Toy Industries (ICTI) and the International Labor Organizations conventions according to publicly available information. The first editorial shows ILO conventions that are real particular and cover the labor conditions mentioned.However, the ILO conventions operate at the state level. That is, unless a state signed the conventions, ILO convention restrictions would not be imposed at the local level. The ICTI codes at the stand level and other firms CSRC followed the ILO conventions, but with less detail. 20 tabular array 1 is divided into the following sections violation of law, bargaining rights and immunity of standstill, workshop safety and living environment, and monitoring procedures. In terms of violation of law, all the codes indicated that there should no child labor under 14 years of age and no forced labor.However, with the barion of the ICTI code, no soulfulness CSRC mentions maternity leave. As for the section on labor bargaining rights and freedom of association, Disney shows its respect for rights for association and collective bargaining, while the ICTI and Wal-Mart show respect for rights of association only. With regard to safety and living environment, Mattels CSRC shows the most elaborated description, followed by the ICTI Disney and McDonalds CSRCs show similar content and are less detailed than those of Mattel and ICTI. Wal-Marts CSRC shows the to the lowest degree detail in this aspect.For the monitoring mechanism, the ICTI CSRC mentioned auditing at the sign stage and in production, as well as a fulfil auditing during the production cycle. The ICTI also mentioned that the factory audit could reexamine the employment record, books, and interview worker mysticly and the audit could be unexpected to reduce the opportunism. While no individual CSRC f orgets such detailed monitoring mechanism description, Disney makes it clear that they can review records and books, and suppliers cannot subcontract their work further. McDonalds mentions that the suppliers management 18 on a lower floorstanding that suppliers may have pressure to workers not to reveal the true factory information to strangers, I have instructed the surveyors to talk to perspective workers and develop trust between each other. Surveyors would explain the survey purpose and ask the agreement of perspective answerers. The survey is conducted outside of the factory areas, usually at local restaurant during workers lunch time, to eliminate influence from suppliers. 19 The survey questionnaire can be provided upon request. 20 More detailed descriptions of the CSRC for ICTI can be found in its website http//www. toyicti. rg/info/codeofbusinesspractices. html. 11 should put an causal agent to monitor the CSRC standard, and show that the auditor can enter workers dormitor y and interview the workers. Among other CSRCs, only Wal-Mart shows that it has increased efforts to monitor the suppliers in both announced and unannounced factory audits. Mattel, however, does not show any monitoring information to public. In a nutshell, individual firms CSRC could be truly different and their monitoring efforts may differ from earlier observations. The next section offers basic observations about the rank of the hassle CSRC confinements from a survey of 12 factories. 12Table 1 A comparision of Toy Industries Corporate Social Responsibility Code of Codes with ILO convent International Labor Organization (ILO) convent International Council of Toy Industries (ICTI) trespass of Law child labor (15 yrs) (C59) minimum age of 14 * under age labor (16 yrs) No gestation leave (C3) maternity benefits as provided by law black Over term work (C1) (C6) ( C41) (171) by law Forced Labor (C29, C105) Labor dicker Rights and Freedom of experience dicker Rights (C98) negoc iate Channel Employee Association (C87) Compliant to Client Labor summation (C87) Labor incision subprogram embodied Bargain (C154) learn work natural rubber nd Living environs assoil synthetic rubber (C155) workshop Safety (C155) victuals problem decease environment problem other problem no discrimination (C111) compensation Coercion and Harassment notification to employee monitor and inspection Unannouned Factory Audit no forced or prison house labor is employed Disney 15 yrs (but 14 yrs if law throw overboarded) Mattel specified by countrys law McDonald 15 yrs (but 14 yrs if law allowed) yes, only if no educate 60hr/week, 1 day/week off no forced labor Wal-Mart not allow 14 yrs old 48/week +12 hr OT, 1/week day off but special business circumstances no forced labor overtime moldiness be voluntary and documented, rest days with local law payroll deductions must comply with law and mattel requirment 72hr/6days, 14/24hrs no forced labor freel employee bureau by local law upplier need to respect right for association lawful wedding activities supplier need to respect right for Collective Bargain lawful collective bargain in lawful and peasceful manner, without penalty, interference supplier need to respect right for association emergency exits, adroit emergency evacuation machinery safeguards = local laws and employees trained facilities or appropriate comestible for meals and other breaks housing, rooms, sanitary facilities meet basic needs ventilated, combustion safety by local laws toilet facilities meet local hygiene requirements , and are properly maintained standards set by law apprise safety adequate lighting and ventilation drinkable drinkng water, sanitary facilites, health and safety housing envriomental friendly race, color, gender, religion, disability with the local law or local standard, choose the higher one rogram and system for fires, spills and natural disasters, emergency exit unlocked have programs to address health an d safety, first aid, medical interference canteens must be safe, sanitary, meet basic needs dormitories must be safe, sanitary, meet basic needs minimize environmental impact fire exit esential safety equipment, first aid kit, emergency care potable drinkng water, adequate, clean restrooms, appropriately-lit work station race, color, gender, religion, disability with the local law no unhealth and hazardous enviroment nontoxic material, envrionmental friendly race, color, gender, religion, disability with the local law licitly mandated rates eed to have local lanuage post the CSR terms to employee 1 to 2 factory manager shall be monitor the compliance of standard yes yes, review employment record and books, yes, but no public figure specified, 26% of 16000 in 2006 (6% increase) can go to housing accommodations can go to housing accommodations others Audit inculdes Initial , Follow-up, In-Production. no further subcontractor can be used. can private interviews with employees 16000 a udit for 8873 suppliers in 2006 Source International Labor Organization (ILO) conventions, International Council of Toy Industries (ICTI), The Walt Disney Company Code of Conduct for Manufactures, Mattel, Inc. Global Manufacturing Principles, McDonalds Code of Coduct for Suppliers, Wal-Mart Stores, Inc.Standard for Suppliers *but notwithstanding the foregoing, that C138 Minimum Age Convention (1973) and C182 Worst Forms of Child Labor Convention (1999) of the International Labor Organization yield Code post or available for all employees in local language. need to have local lanuage post the CSR terms to employee 13 3. 3 Ranking the fuss of CSRC confinements Before going into the methodological analysis, let us examine the level of the labor problem from the survey data. This survey considers each labor problem as a business for the supplier to pass on. The seriousness of each labor problem can be considered as the level of difficulty to accomplish the task. Table 2 shows the pe rcentage of all responsives saying the task is a problem = 1 and the task is not a problem = 0. 21 The larger the mean in the variable, the greater the number of respondents who answered 1, and the more problem is perceived in that task. The variations in these variable promoter are large. all told the respondents cited 1 in response to northward, which had a mean of 1. This is considered to be the most difficult problem. The two other most difficult problems are rush to thickening and labor association, with 99% and 98%, respectively, with 1 on average. As only few respondents raise other problems, its mean is the utmost at 2% on average, which indicates that suppliers consider this the easiest problem. Table 2 Corporate Social Responsibility Tasks Diffculty level tip Obs Mean Std. Dev.Min Labor Bargaining and Association Rights Labor Union 240 1. 00 0. 00 1 0. 99 malady to Client 240 0. 11 0 Employee Association 240 0. 98 0. 14 0 Bargaining Rights 238 0. 45 0. 50 0 Barga ining Channel 240 0. 73 0. 45 0 Labor department Use 240 0. 48 0. 50 0 Collective Bargain 240 0. 96 0. 19 0 Strike Use 240 0. 62 0. 49 0 Violation of Law child labor (14 yrs) 238 0. 11 0. 31 0 under age labor (16 yrs) 238 0. 34 0. 47 0 No Maternity leave 238 0. 82 0. 39 0 guilty Over Time work 238 0. 85 0. 36 0 Safety and Living Environment Fire Safety 235 0. 03 0. 18 0 Workshop Safety 235 0. 26 0. 44 0 Food problem 235 0. 83 0. 38 0 Live environment problem 235 0. 29 0. 46 0 0. 02 other problem 235 0. 3 0 Note the respond 1 means the task is a problem, while 0 means the task is not a problem in respondents factory. labor partnership has 100% respond 1 and that mean it is the hardest problem for suppliers. other problem has 2% respond with 0 and is considered to be the most easiest problem for suppliers. Max 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 21 Please see Appendix 2 for a detailed dislocation of the labor condition distribution. 14 3. 4 Methodology Table 2 presents the distribut ion of the CSRC task and the problem perceived. However, both the respondents knowledge of the problem and the suppliers characteristics may bias the analysis of CSRC effect on labor conditions.The respondents knowledge and the suppliers effort to implement CSRC are, however, unperceivable. To correct these unobservable biases and obtain a consistent musical theme of the likelihood of suppliers problem, this methodology section proposes to use the unidimensional item response theory (UIRT) (or the Rasch) personate. In general, the item response theory (IRT) is a consistence of theories describing the application of numeric models to data from questionnaires and trial runs as a tail for measuring abilities, attitudes, or other variables. 22 The IRT model is based on the idea that the prospect of getting an item correct is a function of a latent trait or ability. The UIRT model is a member of IRT family which applies to divided data. 3 The UIRT model is usually used in test analysis, which can analyze the relative difficulty level of an examination question by removing the individual (the examinees) unobservable influence on the answer of the question. It is used for statistical analysis and development of assessments, often for high-stake tests such as the Graduate Record Examination (GRE). 24 Using the GRE as example, a individual with higher intelligence would more presumable be able to the right way respond to a question in an intelligence test. This study introduces UIRT model to evaluate the impacts of CSRC on labor conditions. The UIRT model in this study portends the opportunity of a worker getting a positive response in a list of the ifferent tasks by taking into account each tasks difficulty level and the different abilities, knowledge, 22 IRT models apply mathematical functions that specify the probability of a discrete outcome, such as a correct response to an item, in terms of person and item parameters. psyche parameters may, for example, represent the ability of a student or the strength of a persons attitude. souvenir parameters include difficulty (location), discrimination (slope or correlation), and pseudoguessing (lower asymptote). contingents may be questions that have incorrect or correct responses, as well as statements that allow respondents to indicate the level of agreement. 3 In general, the UIRT model estimates these outcomes using two types of predictors a persons ability and the test items difficulty level. A persons ability and the test items difficulty level are tending(p) as X in the right side of the equation. Y is a persons response to a test item (or a survey item) and is given on the left(a) side of the equation. 24 Among other methods, IRT provides a root word for evaluating how well assessments and individual questions on assessment work. In education, psychometricians apply IRT to achieve tasks as developing and refining exams, maintaining banks of items for exams, and equating the difficulties of successive versions of exams (for example, to allow comparisons between results over time). 15 nd willingness level of each respondent and individual supplier. The UIRT model is particularly useful in this study because this model can effectively take into account an individuals knowledge of the problem and each suppliers willingness to carry out the task. With the estimated outcomes on hand, one can compare the relative difficulty level of any task by using the easiest task (or the hardest task) that suppliers could fulfill. The last section shows that the other problem is the easiest problem among others this study will therefore use the other problem as the basis for the relative difficulty in rank problems in the statistical analysis. 25In our context, the UIRT model is defined as follows allow us consider a sample of individuals who answer the survey as if taking a test, and assume the probability, Pij , that worker is answer yes (or 1) responds to task j and depends only on a parameter, ai , representing the respondents ability (which can represent the joint outcome of a worker level understanding of the firm and other factors such as suppliers characteristics) and another parameter, b j , representing the difficulty level of the task attempted. hook on further that for some monotone transformation, M. M ( P ) = ai ? b j ij (1) for all i and j. The odds of getting a task right decreases with tasks difficulty (and thus the minus sign before b j ). That is, an additive representation is postulated for the suitably modify probabilities, Pij . When M is the inverse logistic transformation,M ( Pij ) = ln( Pij ) = ai ? b j 1 ? Pij (2) (ln is the natural logarithm), we have UIRT model. 25 We can also use union as a base to obtain a more punctilious estimation of the probability of getting a task right as the responses are all zeros. As this article focuses on the relative difficulty level of tasks, it is well-founded to use the easiest task as the base. 16 The UIRT model has certain very desirable statistical properties for the estimation of these parameters. With the UIRT model, this grouping method can be approximated to estimate parameters. 26 We can then compare the estimated M ( Pij ) to determine the relative difficulty level of each task. robability of getting a particular task with 1 can also be calculated The as exp( M ( Pij )) /(1 + exp( M ( Pij ))) . 27 With these desirable properties of the UIRT model, this study can allow different groupings of CSRC and evaluate their effects on labor conditions by comparing the parameters. 4. Results 4. 1 Basic Result Table 3 shows the benchmark model that pooled all the factories together to investigate the overall picture of labor conditions using the UIRT model and the other problem as the basis of comparisons. As shown in Table 2, response 0 means the worker thinks that there is no problem with a particular task, while 1 means the task has problem.Therefore, usi ng the estimated coefficients, the union coefficients serve as the upper destined and the other problem as the lowest bound the interpretation of the coefficient is therefore an assessment of the relative signification of the problem against union and other problem. That is, when a task has similar significance as the union, it means that the task is very difficult for suppliers, while a task that is not significantly different from the other problem, it means that the task is also an easy problem for suppliers. Table 3 shows (from left), the rigid effect model, the fixed effect with probability weighted, and the random effect model, respectively. All the models have passed their test of specification. Comparing 26Specifically, it can be shown (Lord & Novick, 1968, p. 429) that a respondents raw score (number of tasks correct) is a minimal sufficient statistic for his/her ability. This leads to a practical implementation of the model in that statistical estimates of abilities a nd item parameters can be obtained by proceeding as if everyone with the similar raw score has incisively the comparable ability. The parameter estimates are commonly computed using conditional maximum likelihood CMLE and yield consistent estimates of item parameters (Andersen, 1973). 27 Obviously, the Pij is unobservable, as are ability and item parameters. Estimates of the Pij (and, hence, the ai and b j could be obtained if it were possible to give a respondent the same question about a task on repeated occasions, and his/her responses were independent over trials. Clearly, this is not possible. Alternatively, the probabilities could be estimated if a worker with similar ability could be identified. Assuming a group of people with the same ability, as individuals respond to items independently, the observed proportion of individuals within the group who respond positively to task j is an estimate of the probability that any given person from that group passes the task. 17 the fixed effect model and the model with probability weighted, the ranking of variables coefficients is in the same order. There is no qualitative difference between the two models.In fact, the coefficient value and the ranking are identical between the fixed effect model in the first column and the random effect model in the trey column, suggesting that heterogeneity is not a problem and the random effect model is not necessary. 28 Therefore, this study will be based on the fixed effect model for extended models in the subsequent sections. Union, complaint to client, and employee association are the top three difficult tasks for suppliers. The estimated probability of union, complaint to client, and employee association in the fixed effect model is 1, 0. 9998, and 0. 9996, respectively, suggesting a 100%, 99. 98%, and 99. 96% chance of getting these tasks with a problem. These estimated probabilities are very consistent with the statistics in Table 2. This reflects that CSRC may not be very helpful for labor union and freedom of association activities.All the tasks in the model are positively and significantly different from the other problem, suggesting that these problems warrant our attention, except the fire safety problem. The reason why the fire safety problem is not frequently cited as a problem may be attributed to the fact that after a fatal factory fire in Shenzhen area in the early 1990s, a workshop-level fire safety law has been introduced. Since then, labor NGOs and the local government have paid attention to fire safety in workplaces. 29 After a brief overview about the problem with the pooled data, the next section tests hypotheses about the effectiveness of the different CSRCs and the problem of opportunism at the supplier level. 28When reading across the columns, one should vitiate directly comparing the coefficients among different models as the assumptions are different. However, we can focus on the relative importance of variables in each m odel. 29 For a detailed discussion about the labor movement and the history about the event, read AMRC (2004 41-82). 18 Table 3 Unidimensional Item repartee Model for Labor Conditions Fixed Effect Item Coef. Labor Bargaining and Association Rights Union 22. 71 (1. 82) Complaint to Client 8. 37 (0. 9) Employee Association 7. 85 (0. 76) Bargaining Rights 3. 87 (0. 61) Bargaining Channel 5. 03 (0. 59) Labor department Use 3. 99 (0. 58) Collective Bargain 7. 25 (0. 73) Strike Use 4. 54 (0. ) Violation of Law Child labor (14 yrs) 1. 96 (0. 6) Under age labor (16 yrs) 3. 38 (0. 61) Maternity leave 5. 51 (0. 62) Illegal Over Time work 5. 74 (0. 63) Safety and Living Environment Fire Safety 0. 71 (0. 7) Workshop Safety 3. 01 (0. 61) Food problem 5. 58 (0. 65) Live environment problem 3. 18 (0. 61) constent bet of obs F( 16, 239) Prob F Pseudo R2 Linktest _hat 4045. 00 10551. 86 0. 00 0. 49 0. 00 1. 00 Pz 0. 00 0. 00 0. 00 0. 00 0. 00 0. 00 0. 00 0. 00 Fixed Effect with p-weighted Coef. P z 28. 30 (0. 59) 10. 70 (0. 92) 10. 52 (0. 78) 4. 72 (0. 61) 6. 14 (0. 61) 5. 09 (0. 6) 9. 39 (0. 75) 5. 33 (0. 61) 2. 56 (0. 66) 4. 03 (0. 61) 6. 2 (0. 62) 7. 99 (0. 67) 1. 58 (0. 78) 4. 47 (0. 61) 7. 28 (0. 65) 4. 74 (0. 62) 0. 00 0. 00 0. 00 0. 00 0. 00 0. 00 0. 00 0. 00 Random Effect Coef. 29. 19 (9. 73) 8. 44 (0. 9) 7. 92 (0. 76) 3. 88 (0. 61) 5. 05 (0. 6) 3. 99 (0. 58) 7. 31 (0. 72) 4. 56 (0. 6) 1. 96 (0. 6) 3. 38 (0. 61) 5. 55 (0. 62) 5. 79 (0. 63) 0. 71 (0. 7) 3. 01 (0. 61) 5. 62 (0. 65) 3. 18 (0. 61) 4. 06 (0. 58) 4045. 00 F( 16, 239) Prob F /lnsig2u Pz 0. 00 0. 00 0. 00 0. 00 0. 00 0. 00 0. 00 0. 00 0. 00 0. 00 0. 00 0. 00 0. 00 0. 00 0. 00 0. 00 0. 00 0. 00 0. 00 0. 00 0. 31 0. 00 0. 00 0. 00 0. 04 0. 00 0. 00 0. 00 0. 31 0. 00 0. 00 0. 00 0. 00 4045. 00 88251. 79 0. 0 0. 55 1. 00 (0. 13) 0. 00 (0. 01) 0. 00 1. 00 1. 00 (0. 13) _hatsq 0. 00 (0. 01) Other Problem is the basis for the comparison Jacknife standard break in parenthesis 60. 68 0. 00 -4. 55 (0. 29) sigma_u 0. 10 (0. 02) Prob = chibar2 0. 40 Likelihood-ratio test 0. 08 19 4. 2 Which CSR code is better? As shown in Table 1, different branded companies may have differences in the CSRC statement. This section aims to evaluate the effect of an individual brand companys CSRC on labor conditions and to understand the relationships between companys CSRC statement and CSRC outcomes. Table 4 shows the UIRT model for labor conditions by different CSRCs.As there are suppliers with more than one CSRC, this study will pool suppliers with same CSRC to estimate the difficulty level of relative tasks. For example, as the survey has one supplier with Disney only but two suppliers with Disney & Wal-Mart, this study combines these three supplier data into the UIRT model. As discussed in the methodology section, the UIRT model can then basically extract the common information from these Disney suppliers and investigate the aggregated Disney CSRC effect. Table 4 ranks the CSRC according to the number of ta sks similar to the difficulty level of union, descending from left (the easiest task) to right (the most difficult task).In Table 4, columns 1 to 5 show suppliers with a larger number of problems and the lowest number of problems similar to the difficulty level of the problem union. They are non-CSRC suppliers (with 6 problems), McDonalds (5 problems), Mattel (2 problems), Disney (2 problems), and Wal-Mart (with 0 problem). The first observation is that the most serious problems are concentrated on labor bargaining and association rights. When centering on individual CSRC in this section, the non-CSRC suppliers in cited in the first column show that five tasks have the same difficulty level as union. In contrast, there are no such tasks for the Wal-Mart suppliers, suggesting that they may commit less serious labor bargaining and association rights problems.While we focus on the tasks with the lowest bound of difficulty, Disney suppliers may have more (6 tasks) tasks that are insig nificantly different from the other problem. This means that Disneys suppliers may be less likely to have labor bargaining and association problems. why so? Can the foreign mission statements of CSRC give some useful information? As Table 1 shows, Disneys mission statement provides a more detailed description than other CSRCs with respect to the labor bargaining and association problems. However, while there is no hardcore statement about labor bargaining and association problems in the McDonalds CSRC, its suppliers are more likely to commit violations over other CSRCs, as shown in the second column of Table 4. On 20 he whole, it seems that the CSRC could have some positive impacts on labor bargaining and association problems, and the effects would be dependent on the details of the CSRC statements. 30 Let us now discuss the violation of laws. While the non-CSRC suppliers may have serious problems regarding maternity leave, McDonalds suppliers may have serious problems with smu ggled overtime work when compared with the labor union problem. This is consistent with McDonalds CSRC statement the most concise among other CSRCs on illegal overtime work. When focusing on the insignificant problem, an interesting pattern emerged the tasks that are not as serious as other problems are inversely distributed along the columns, except for Wal-Mart suppliers.Disney suppliers continually outperform other CSRCs with no tasks considered significant, while Wal-Mart suppliers show significance in all the tasks. While Table 1 does not show any significant difference between Wal-Marts CSRC compared to the other CSRC in this issue, the results may be affected by the fact that these suppliers do not produce goods only for Wal-Mart. The next subsection analyzes the number of CSRCs and its the impact on labor conditions. In terms of safety and living environment issues, all suppliers under this study do not indicate significant problems on fire safety. However, the non-CSRC su ppliers are consistently worse than other suppliers in terms of workshop safety and food problem.Indeed, food seems to be a common problem among suppliers, except for Disney suppliers. Consistent with results in labor bargaining and association rights, Disney suppliers outperform other CSRC suppliers because no task is significantly different from the other problem. In all the three sections, Disney suppliers continually outperform other CSRCs. This result seems to be consistent with Disneys CSRC statement, shown in Table 1. Some suppliers work for more than one CSRC, which may mean more transaction costs in avoiding the CSRC. It is interesting to explore how the number of CSRCs affect firms. 30 Certainly, we can just observe the CSRC statement, but efforts to enforce the statement terms are unobservable.As discussed in the methodology section, the UIRT model is a tool to remove unobservable influences and obtain consistent estimates of the difficulty level of tasks. 21 Table 4 Uni dimensional Item reply Model for Labor Conditions by different CSR code No CSR code McDonalds Mattel Item Coef. Pt Coef. Pt Coef. Pt Labor Bargaining and Association Rights Union 42. 38 0. 00 41. 63 0. 00 19. 98 0. 00 (4. 33) (4. 15) (3. 01) Complaint to Client 42. 38 0. 00 41. 63 0. 00 5. 53 0. 00 (4. 33) (4. 15) (0. 94) 42. 38 41. 63 19. 98 Employee Association 0. 00 0. 00 0. 00 (4. 33) (4. 15) (3. 01) Bargaining Rights 42. 38 0. 00 -0. 03 0. 26 0. 28 0. 77 (4. 33) (0. 03) (0. 3) Bargaining Channel 42. 38 0. 00 21. 30 0. 00 19. 98 0. 00 (4. 33) (2. 27) (3. 01) Labor department Use 0. 00 0. 87 20. 67 0. 00 2. 69 0. 00 (0. 00) (2. 01) (0. 68) Collective Bargain 42. 38 0. 00 41. 63 0. 00 4. 80 0. 00 (4. 33) (4. 15) (0. 89) Strike Use 20. 77 0. 00 41. 63 0. 00 1. 48 0. 05 (2. 30) (4. 15) (0. 74) Violation of Law Child labor (14 yrs) Under age labor (16 yrs) Maternity leave Illegal Over Time work 0. 00 (0. 00) 19. 92 (1. 77) 42. 38 (4. 33) 0. 87 0. 00 0. 00 0. 00 -0. 03 (0. 03) -0. 03 (0. 03) 21. 92 (2. 44) 41. 63 (4. 15) 0. 00 (0. 00) 0. 00 (0. 00) 21. 84 (2. 43) 0. 26 0. 26 0. 00 0. 00 0. 28 (0. 93) 0. 51 (0. 9) 3. 83 (0. 77) 5. 23 (0. 99) 0. 30 (0. 93) 2. 29 (0. 78) 4. 01 (0. 88) 0. 77 0. 57 0. 00 0. 00 Disney Coef. 23. 12 (13. 27) 23. 12 (13. 27) 6. 32 (13. 94) 4. 63 (14. 02) 3. 24 (14. 00) 4. 87 (13. 94) 8. 00 (19. 73) 23. 12 (13. 27) 2. 94 (13. 92) 4. 48 (14. 02) 6. 49 (14. 24) 5. 35 (14. 05) 1. 13 (14. 02) 3. 27 (14. 01) 7. 97 (28. 15) Pt 0. 09 0. 09 0. 65 0. 74 0. 82 0. 73 0. 69 0. 09 0. 83 0. 75 0. 65 0. 71 21. 01 (1. 94) Safety and Living Environment Fire Safety Workshop Safety Food problem 0. 00 (0. 00) 21. 01 (1. 94) 20. 89 (1. 91) 0. 87 0. 00 0. 00 0. 34 0. 31 0. 00 0. 75 0. 01 0. 00 0. 94 0. 82 22 0. 78 4. 3 Is more CSRC better?This section focuses on the number of CSRCs if more CSRCs increase transaction costs for suppliers, which in turn results in violations of the CSRC, then there should be less violation with regard to labor conditions as the number of CSRCs increases. Table 5 shows the UIRT model for labor conditions according to the number of CSR code. The nonCSRC supplier column is the same as in Table 4, while the 1 CSRC column has a pool of suppliers with a single CSRC and 2 CSRC suppliers include Mattel Wal-Mart and Disney Wal-Mart. The pattern of violations in labor conditions in Table 8 is consistent with the hypothesis that more CSRCs is better to control opportunism at the suppliers level the suppliers of 2 CSRCs constantly outperform the 1 CSRC suppliers and the non-CSRC suppliers on the control of labor conditions.For example, in terms of labor bargaining and association rights, the number of tasks with a similar significant coefficient as union is 5 in No CSRC suppliers, 2 in 1 CSRC supplier, and none in 2 CSRC suppliers. All the estimation results in violation of law and safety and living environment show that the tasks relative to coefficients are reduced as the number of CSRC increases. These results ar e consistent with the perspective that an increase in the number of CSRCs may increase the transaction costs of opportunism at the suppliers level. However, suppliers may also have measures to reduce governance effects from CSRC. The effects of different measures of opportunism are investigated in the next section. 23 24Table 5 Unidimensional Item Response Model for Labor Conditions by number of CSR code No CSR code 1 CSR code 2 CSR codes Item Coef. Pt Coef. Pt Coef. Labor Bargaining and Association Rights Union 42. 38 0. 00 39. 19 0. 00 19. 40 (4. 33) (0. 31) (3. 12) Complaint to Client 42. 38 0. 00 39. 19 0. 00 6. 11 (4. 33) (0. 31) (0. 89) Employee Association 42. 38 0. 00 39. 19 0. 00 5. 57 (4. 33) (0. 31) (0. 76) Bargaining Rights 42. 38 0. 00 19. 10 0. 00 2. 02 (4. 33) (0. 26) (0. 67) Bargaining Channel 42. 38 0. 00 20. 14 0. 00 3. 92 (4. 33) (0. 20) (0. 64) Labor department Use 0. 00 0. 87 19. 73 0. 00 3. 51 (0. 00) (0. 25) (0. 58) Collective Bargain 42. 38 0. 00 23. 61 0. 00 5. 21 (4. 33) (0. 98) (0. 6) Strike Use 20. 77 0. 00 20. 24 0. 00 3. 51 (2. 30) (0. 26) (0. 64) Violation of Law Child labor (14 yrs) Under age labor (16 yrs) Maternity leave Illegal Over Time work Safety and Living Environment Fire Safety Workshop Safety Food problem Live environment problem 0. 00 (0. 00) 21. 01 (1. 94) 20. 89 (1. 91) 0. 87 0. 00 0. 00 0. 87 15. 47 (0. 87) 14. 77 (14. 65) 22. 30 (0. 42) 18. 70 (0. 23) 2030. 00 0. 64 0. 00 0. 32 0. 00 0. 00 0. 43 (0. 76) 2. 98 (0. 68) 4. 48 (0. 79) 2. 72 (0. 67) 1335. 00 0. 40 0. 00 (0. 00) 19. 92 (1. 77) 42. 38 (4. 33) 21. 01 (1. 94) 0. 87 0. 00 0. 00 0. 00 17. 05 (0. 41) 19. 27 (0. 24) 20. 85 (0. 24) 24. 31 (15. 01) 0. 0 0. 00 0. 00 0. 11 1. 61 (0. 64) 1. 76 (0. 67) 4. 07 (0. 67) 4. 30 (0. 70) Pt 0. 00 0. 00 0. 00 0. 00 0. 00 0. 00 0. 00 0. 00 0. 01 0. 01 0. 00 0. 00 0. 57 0. 00 0. 00 0. 00 0. 00 (0. 00) upshot of obs 680. 00 Pseudo R2 0. 84 Other Problem is the basis for the comparison Jacknife standard error in parenthesis 4. 4 Suppliers Measures to circumvent CSRC auditing 25 When considering other transaction costs of governing with the CSRC, suppliers may apply different strategies to get around the CSRC auditing. Because most CSRC auditing would give advance notice to suppliers, suppliers would have room for measures to circumvent the auditing.This study has documented some of the common supplier measures to circumvent CSRC auditing fire drill, trained question, giving a holiday, cleaning the factory, and safety drill. Certainly, suppliers may use more than one measure to get around the CSRC auditing it is likely that suppliers with more labor problems would use more measures to cover up their problems. This section first looks at the number of measures that may be associated with labor problems. We then investigate how the different measures affect labor conditions. Table 6 shows the UIRT model for the number of measures to circumvent CSRC auditing. The first column of Table 6 shows the result of no measure. A response of no measure to circumvent CSRC auditing is associated with no significant measures that are different from other problem, suggesting that a supplier without any measure for CSRC auditing is performing relatively well in terms of labor conditions. The second column, i. e. , one measure to circumvent CSRC auditing, shows a very different pattern. Although no task is as difficult as the union problem, all tasks show a significant coefficient, except the fire safety problem. The third column, two or more measures, indicates an even stronger coefficient in all the tasks than the column with one measure. Furthermore, the task complaint to client is as strong as union in this column.When comparing the three columns in Table 6, it appears that the more the measures suppliers use, the higher the likelihood they commit more labor problems. This is consistent with the transaction cost hypothesis that opportunism may increase the transaction cost of governing with the CSR C. It is not necessary that all measures have an equal effect on labor problems. The next section shows the effect of individual measure of CSRC auditing on labor problems. 26 Table 6 Unidimensional Item Response Model for Labor Conditions by Number of Suppliers Measure No Measure One Measure Two or More Measures Item Labor Bargaining and Association Rights Union 19. 88 0. 22 20. 47 0. 0 34. 57 0. 00 (15. 58) (3. 66) (4. 30) Complaint to Client 5. 35 0. 79 7. 34 0. 00 34. 57 0. 00 (19. 88) (1. 20) (4. 30) Employee Association 5. 35 0. 79 7. 34 0. 00 20. 80 0. 00 (20. 03) (1. 21) (2. 85) Bargaining Rights 1. 64 0. 92 2. 40 0. 00 17. 16 0. 00 (15. 09) (0. 77) (2. 19) Bargaining Channel 3. 47 0. 82 4. 24 0. 00 17. 84 0. 00 (14. 91) (0. 74) (2. 13) Labor department Use 3. 20 0. 83 3. 85 0. 00 17. 42 0. 00 (14. 89) (0. 71) (2. 17) Collective Bargain 19. 88 0. 22 6. 63 0. 00 19. 85 0. 00 (15. 58) (1. 03) (2. 30) Strike Use 2. 94 0. 85 3. 78 0. 00 18. 47 0. 00 (15. 15) (0. 73) (2. 18) Viol ation of Law Child labor (14 yrs) -14. 1 0. 34 1. 26 0. 09 15. 87 0. 00 (14. 61) (0. 74) (2. 18) Under age labor (16 yrs) 2. 24 0. 88 2. 59 0. 00 16. 91 0. 00 (15. 11) (0. 77) (2. 12) Maternity leave 2. 77 0. 86 4. 60 0. 00 18. 89 0. 00 (15. 14) (0. 76) (2. 20) Illegal Over Time work 5. 23 0. 80 6. 39 0. 00 19. 09 0. 00 (19. 94) (1. 06) (2. 26) Safety and Living Environment Fire Safety 0. 00 1. 00 0. 73 0. 40 12. 91 0. 32 (22. 56) (0. 86) (12. 82) Workshop Safety 0. 00 1. 00 2. 03 0. 01 16. 24 0. 00 (22. 56) (0. 78) (2. 09) Food problem 4. 13 0. 79 5. 10 0. 00 21. 50 0. 16 (15. 39) (0. 86) (15. 24) Live environment problem 2. 19 0. 89 2. 67 0. 00 16. 88 0. 00 (15. 10) (0. 77) (2. 2) Number of obs 267. 00 1777. 00 1321. 00 Pseudo R2 0. 52 0. 51 0. 52 Other Problem is the basis for the comparison Jacknife standard error in parenthesis Suppliers measures to circumvent CSRC auditing include fire drill, trained question, giving a holiday, Cleaning the Factory, and Safety Drill. 27 4. 4 S uppliers Measures to circumvent CSRC auditing Table 7 shows the UIRT model for individual measures to circumvent CSRC auditing. The measures in this study include fire drill, safety drill, giving a holiday, trained question, and cleaning the factory. As the safety drill estimation is not converged, Table 10 shows only the four other supplier measures.Table 7 ranks the measures according to the number of the most serious tasks (compared to union) from the left, namely, fire drill (6 tasks), giving a holiday (4 tasks), trained question (0 task), and cleaning the factory (0 task). Why do fire drill suppliers have the most tasks similar to the union coefficient? As discussed in the last subsection, fire safety is one of the most serious areas that the Chinese local authorities are concerned with, and it is the most frequent task to be checked during a factory visit. If a supplier has not even met the basic standard of the fire safety, it is reasonable to infer that other tasks are likel y to be a problem as well. This is similar in the case of holidays for workers meaning that the supplier tries to blackball workers from revealing this fact.As giving holidays for a busy factory is costly, suppliers are likely to create more than one problem in the task. Trained question is one of the most common ways that suppliers try to get around CSRC auditing during the worker interview. Suppliers try to train workers with model answers that fit the CSRC auditing interview questions. The trained question effect may be very limited as only trained workers model their answers to the interview question, and this cannot cover other the problems in the factory. It is clear that the trained question may have fewer problems than fire drill and giving a holiday. Cleaning the factory was found the least serious measure.Only suppliers who are confident with their labor conditions would use cleaning the factory to polish their workshop. 28 Table 7Unidimensional Item Response Model for L abor Conditions by Suppliers Measure Fire Drill curb Holiday Trained question Cleaning Factory Item Coef. Pt Coef. Pt Coef. Pt Coef. Pt Labor Bargaining and Association Rights Union 40. 28 0. 00 39. 74 0. 00 35. 79 0. 00 21. 80 0. 00 (1. 05) (3. 83) (5. 39) (2. 87) Complaint to Client 40. 28 0. 00 39. 74 0. 00 21. 55 0. 00 8. 79 0. 49 (1. 05) (3. 83) (2. 80) (12. 70) Employee Association 40. 28 0. 00 22. 22 0. 24 21. 55 0. 00 7. 68 0. 00 (1. 05) (18. 58) (2. 98) (1. 00) Bargaining Rights 18. 38 0. 29 0. 00 0. 75 17. 88 0. 00 3. 5 0. 00 (14. 33) (0. 00) (2. 78) (0. 75) Bargaining Channel 18. 38 0. 00 39. 74 0. 00 18. 36 0. 00 4. 47 0. 00 (0. 45) (3. 83) (2. 62) (0. 73) Labor department Use 21. 11 0. 00 16. 38 0. 32 18. 07 0. 00 4. 38 0. 00 (0. 61) (16. 06) (2. 69) (0. 71) Collective Bargain 40. 28 0. 00 39. 74 0. 00 20. 60 0. 00 6. 54 0. 00 (1. 05) (3. 83) (2. 81) (0. 85) Strike Use 40. 28 0. 00 17. 62 0. 00 19. 30 0. 00 4. 85 0. 00 (1. 05) (1. 31) (2. 88) (0. 74) Violation of Law C hild labor (14 yrs) 0. 00 0. 45 18. 35 0. 00 16. 41 0. 00 2. 01 0. 01 (0. 00) (1. 70) (2. 61) (0. 73) Under age labor (16 yrs) 18. 38 0. 29 18. 00 0. 00 17. 51 0. 00 3. 33 0. 00 (14. 3) (2. 04) (2. 60) (0. 75) Maternity leave 21. 11 0. 25 22. 22 0. 27 19. 09 0. 00 5. 22 0. 00 (14. 97) (19. 80) (2. 72) (0. 74) Illegal Over Time work 40. 28 0. 00 22. 22 0. 27 19. 85 0. 00 6. 14 0. 00 (1. 05) (19. 80) (2. 72) (0. 85) Safety and Living Environment Fire Safety 0. 00 0. 39 0. 00 0. 75 14. 16 0. 00 0. 72 0. 40 (0. 00) (0. 00) (2. 54) (0. 85) Workshop Safety 0. 00 0. 38 0. 00 0. 75 16. 77 0. 00 2. 84 0. 00 (0. 00) (0. 00) (2. 70) (0. 75) Food problem 40. 28 0. 00 39. 74 0. 00 20. 59 0. 00 6. 11 0. 00 (1. 05) (3. 83) (2. 87) (0. 89) Live environment problem 18. 38 0. 00 21. 51 0. 00 17. 53 0. 00 2. 88 0. 00 (0. 45) (2. 15) (2. 72) (0. 5) Number of obs 85. 00 408. 00 1440. 00 2452. 00 Pseudo R2 0. 83 0. 88 0. 49 0. 50 Other Problem is the basis for the comparison Jacknife standard error in pa renthesis except the fire estimation, which indicates standard error in parenthesis. Safety Drill is not converged in estimation. 29 Robustness period For convenience, the estimation of the UIRT model in Table 7 uses the other problem as a base. However, as the other problem is not exactly along zero in the data, there may be a problem in estimating the true probability of the task. More importantly, there may be problem of the UIRT assumption about the zero covariance among items in the estimation.This validness check uses the union as a base in Table 8, which reruns the UIRT model in Table 5. If the zero covariance assumption is violated, using the upper bound of the problem, the union, as the base of estimation will give very different results. However, if the zero covariance assumption is not violated, as this estimation uses union as a base, the coefficient will be interpreted exactly diametrical to that given in Table 5 those problems with larger and significant coefficient s given in Table 5 should now be close to zero and insignificant in Table 8 and vice versa. Most of the coefficients here are negatively signed, suggesting that they are weaker than the union problem.Non-CSRC suppliers are seen to have six problems (complaint to client, employee association, bargaining rights, bargaining channel, collective bargain, and maternity leave), 1 CSRC has two problems (complaint to client and emplo