Not Walking the Walk: How Dual Attitudes Influence Behavioral Outcomes in Ethical Consumption
不言而行：双重态度如何影响道德消费中的行为结果

Received: 22 November 2016/ Accepted: 10 April 2017/Published online: 28 April 2017
收到日期：2016 年 11 月 22 日/接受日期：2017 年 4 月 10 日/在线发布日期：2017 年 4 月 28 日

(C) Springer Science+Business Media Dordrecht 2017
(C) 斯普林格科学+商业传媒多德雷赫特 2017

by the nature of the stimuli. On the other hand, implicit attitudes are relatively unaffected by the nature of the stimuli present and remain relatively constant. Based on the findings, implicit attitudes guide behavior and determine an individual's preferences. Even though explicit attitudes react to the stimuli presented, our findings suggest they have no impact on the choice of consumers. These findings improve the understanding of ethical consumption, provide a reason as to why the attitude-behavior gap exists, provide a foundation for future researchers and help firms better understand the impact of perceived business on creating a behavioral shift in ethical consumption.
根据刺激的性质。另一方面，隐性态度相对不受所呈现刺激的性质影响，保持相对恒定。根据研究结果，隐性态度指导行为并决定个体的偏好。尽管显性态度对所呈现的刺激做出反应，但我们的研究结果表明它们对消费者选择没有影响。这些发现提高了对道德消费的理解，为态度与行为之间的差距存在的原因提供了理由，为未来研究者提供了基础，并帮助企业更好地理解感知业务对创造道德消费行为转变的影响。

Keywords Attitude-behavior gap Ethical consumption Dual attitudes Implicit association test
关键词态度-行为差距 道德消费 双重态度 内隐联想测试

EA Explicit attitude EA 明确态度

IA Implicit attitude IA 隐性态度

CPE Consumers' perceived ethicality
CPE 消费者的道德感

PREF Preference PREF 首选

OLS Ordinary least squares
OLS 普通最小二乘法

IAT Implicit association test
IAT 隐性联想测试

TP Time period TP 时间段

Recent years have seen an increase in the consumption of ethical products. This is reflected in the 2013-2014 annual report of Fairtrade World Wide, commonly recognized as the world's leading ethical label, which sells over 30,000 products in 125 countries. Shoppers spent billion on Fairtrade products in 2014, an increase of from 2012 and a distant billion in 2004 (Fairtrade International
近年来，道德产品的消费量有所增加。这体现在公认为世界领先的道德标签的 Fairtrade World Wide 的 2013-2014 年度报告中，该标签在 125 个国家销售超过 30,000 种产品。购物者在 2014 年购买了 亿美元的 Fairtrade 产品，比 2012 年增加了 ，远远超过了 2004 年的 亿美元（Fairtrade International）。

2014; Revenues of Fairtrade International Products 2016). Major multinationals are also attempting to align their business ethics strategies to increase activity in the ethical food sector. For example, McDonald's Australia sources and serves only sustainable coffee (Clarke 2012) and Nestle launched a unique Fairtrade-certified instant coffee for its British consumers in 2005 (Clarke 2012). For countries as a whole, in the UK, sales of ethical foods accounted for of food sales and in the USA, they accounted for (Carr et al. 2013). At the heart of this trend lies the assumption that there is a change in consumers' norms and values (Caruana 2007) and businesses need to shift their ethics strategies to reflect this change.
2014 年；公平贸易国际产品 2016 年收入。主要跨国公司也在努力调整其商业道德战略，以增加在道德食品领域的活动。例如，麦当劳澳大利亚仅采购和供应可持续咖啡（Clarke 2012 年），雀巢于 2005 年为其英国消费者推出了一款独特的公平贸易认证即溶咖啡（Clarke 2012 年）。就整个国家而言，在英国，道德食品销售占食品销售的 ，在美国，占 （Carr 等，2013 年）。这一趋势的核心是假设消费者的规范和价值观发生了变化（Caruana 2007 年），企业需要调整其道德战略以反映这种变化。

While in absolute terms some of these trends might seem encouraging, in relative terms they are not as promising, considering the overall size of the markets. In addition, other than the food sector, the growth in ethical consumption is small (Eckhardt et al. 2010). Further, the adoption rates for ethical products are not as high and on a global scale, ethical products still only represent a niche market. In much of Asia, that is, China, Japan and India, ethical foods only account for or less of total food and beverage sales (Clarke 2012). The numbers are even lower in South America and Africa.
尽管从绝对值来看，这些趋势中的一些可能看起来令人鼓舞，但从相对值来看，考虑到市场的整体规模，它们并不那么令人鼓舞。此外，除了食品行业之外，道德消费的增长是有限的（Eckhardt 等人，2010 年）。此外，道德产品的采纳率并不高，在全球范围内，道德产品仍然只代表一个小众市场。在亚洲大部分地区，即中国、日本和印度，道德食品仅占总食品饮料销售额的 或更少（Clarke，2012 年）。在南美洲和非洲，这个数字甚至更低。

These low consumption percentages are despite the fact that consumers vehemently report their insistence for demanding ethical products. A study of European consumers by Cowe and Williams (2000) found that while approximately of consumers reported caring about ethical consumption, only of purchases reflected this ideal. Similarly, 25% of New Zealanders reported addressing ethical challenges through their purchase decisions (Clarke 2012), but the actual sales were closer to . Coined the '30:3 phenomenon,' this discrepancy between attitudes and behavior is labeled as the attitude-behavior gap (Caruana et al. 2016; Papaoikonomou et al. 2011b; Boulstridge and Carrigan 2000; Kollmuss and Agyeman 2002; Cowe and Williams 2000). Evidently, while there appears to be a reported demand for ethical products, it is not being appropriately reflected in sales.
这些低消费百分比是尽管消费者强烈报告他们坚持要求道德产品。 Cowe 和 Williams（2000）对欧洲消费者的研究发现，大约 的消费者表示关心道德消费，但只有 的购买反映了这一理想。同样，25%的新西兰人报告通过他们的购买决策解决道德挑战（Clarke 2012），但实际销售更接近 。这种态度和行为之间的差距被称为态度-行为差距（Caruana 等人 2016 年；Papaoikonomou 等人 2011b；Boulstridge 和 Carrigan 2000；Kollmuss 和 Agyeman 2002；Cowe 和 Williams 2000）的“30:3 现象”。显然，虽然似乎有对道德产品的需求报告，但这种需求并没有得到适当地反映在销售中。

This gap in self-reported attitude and actual behavior poses an important research problem for business ethics (Caruana et al. 2016) and has been studied across various outcomes such as behavior (Papaoikonomou et al. 2011a), corporate responsibility (Boulstridge and Carrigan 2000) and environmental responsibility (Kollmuss and Agyeman 2002). As suggested by Papaoikonomou et al. (2011b), to contribute to the underlying reason behind the gap would be a significant study, as the assumption that attitudes predict behavior cannot be taken for granted. Thus, a major motivation behind this research is to better understand why this gap exists. In doing so, we go beyond the methodological and contextual attributions contributing to the gap.
自我报告的态度与实际行为之间的差距为商业道德（Caruana 等人，2016 年）提出了一个重要的研究问题，并已在各种结果中进行了研究，如行为（Papaoikonomou 等人，2011a 年）、企业责任（Boulstridge 和 Carrigan，2000 年）和环境责任（Kollmuss 和 Agyeman，2002 年）。正如 Papaoikonomou 等人（2011b 年）所建议的那样，为了对这种差距背后的原因做出贡献将是一项重要的研究，因为不能想当然地认为态度可以预测行为。因此，这项研究背后的一个主要动机是更好地理解为什么存在这种差距。在这样做的过程中，我们超越了对导致这种差距的方法论和背景属性的贡献。
Instead, we focus on the distinction between two types of attitudes (explicit and implicit). In addition, from a managerial perspective, there is a need to understand why consumers do not behave in the manner that they predict, in order to develop marketing and communication strategies for ethical/sustainable products.
相反，我们专注于两种态度（显性和隐性）之间的区别。此外，从管理的角度来看，有必要了解消费者为什么不按照他们预测的方式行事，以便为道德/可持续产品制定营销和沟通策略。

The primary aims of this research are, therefore, threefold. First, instead of studying attitude as a unidimensional identifier, as in prior research, we propose and test a more holistic understanding of consumer attitude and its gap with behavior. Toward this, we utilize the theory of Dual Model of Attitudes (Wilson et al. 2000) to underpin our arguments. This model suggests that individuals have two distinct attitudes-explicit and implicit. Explicit attitudes are easier to influence and may drive behavior in the shorter term. It is these attitudes that are measured by selfreported tasks. Implicit attitudes, however, are harder to change and more likely to drive behavior in the longer term. It is likely that implicit attitudes therefore are the true drivers of purchase of ethical products, while explicit attitudes drive socially expressed intentions, especially where a brand's ethical reputation might be in question. In our research, we contribute to the field of business ethics by allowing the manager focusing on ethical products to identify strategies that will allow them to attenuate the gap between intention and behavior. We do this by teasing apart the two attitude constructs that might more especially lead in non-congruent directions and have dissimilar influences on behavior. We, thus, study this effect longitudinally to enable us to tease out the differential effects of implicit and explicit attitudes on ethical consumption.
这项研究的主要目标有三个。首先，与以往研究中将态度作为单一标识符进行研究不同，我们提出并测试了对消费者态度及其与行为之间差距的更全面的理解。为此，我们利用了《态度双重模型》（Wilson 等人，2000 年）来支持我们的论点。该模型表明，个体具有两种不同的态度-显性和隐性。显性态度更容易受到影响，可能在短期内驱动行为。这些态度是通过自我报告任务来衡量的。然而，隐性态度更难改变，更有可能在长期内驱动行为。因此，隐性态度很可能是道德产品购买的真正驱动因素，而显性态度则驱动社会表达意图，特别是在品牌的道德声誉可能受到质疑的情况下。在我们的研究中，我们通过让管理者专注于道德产品，帮助他们确定策略，以减少意图与行为之间的差距，从而为商业道德领域做出贡献。 我们通过分析可能导致不一致方向并对行为产生不同影响的两种态度构建来做到这一点。因此，我们从纵向研究这种效应，以使我们能够分析隐性和显性态度对道德消费的差异影响。

Second, one of the main concerns for research in ethics is the strong potential of eliciting socially desirable responses in self-reported measures of explicit attitudes (Petty et al. 2003). From the theories of planned behavior and reasoned action (Bagozzi 1992), we recognize that attitude (operationalized as implicit and explicit), intention (operationalized as brand preference) and behavior (operationalized as choice) are three separate but primarily sequential constructs. Toward addressing the first part of the sequence, i.e., attitudes, we contribute by using another measure (in addition to self-reporting) of attitudes that is not susceptible to the social desirability bias. Specifically, we utilize an implicit association test to measure participants' implicit or 'hidden' attitudes toward a target brand. This enables us to (a) more accurately measure attitudes that a consumer holds without the potential confounding effect of social desirability; (b) better understand the differential impact that information has on implicit and explicit attitudes; and (c) understand the unique influence of implicit attitudes on behavior.
研究伦理学的一个主要关注点是在自我报告的明示态度测量中引发社会期望回应的强大潜力（Petty 等，2003 年）。从计划行为理论和理性行为理论（Bagozzi，1992 年）的角度，我们认识到态度（作为内隐和明示的操作化）、意向（作为品牌偏好的操作化）和行为（作为选择的操作化）是三个分开但主要是顺序的构建。为了解决序列的第一部分，即态度，我们通过使用另一种不易受社会期望偏见影响的态度测量（除了自我报告）来做出贡献。具体来说，我们利用内隐关联测试来测量参与者对目标品牌的内隐或“隐藏”态度。这使我们能够（a）更准确地测量消费者持有的态度，而不受社会期望的潜在混淆影响；（b）更好地理解信息对内隐和明示态度的差异影响；以及（c）理解内隐态度对行为的独特影响。

Finally, existing research largely focuses on self-reported intention to purchase. Although an important
最后，现有研究主要集中在自我报告的购买意向上。尽管这是一个重要的
construct, based on existing theories of contemporary attitudes, literature points to the fact that intention and behavior are not overlapping constructs. This is problematic because most research assumes that intention will directly determine individual's behavior, which is often not the case (Carrington et al. 2010). In order to address this, in addition to studying self-reported preferences, we also measure actual consumer choice to better understand the likelihood of purchase based on the two types of attitudes. Thus, all three measures that appear in contemporary attitudes, i.e., attitude, intention and behavior, are addressed and explored in our studies.
根据现代态度的现有理论构建，文献指出意图和行为并不是重叠的构建。这是有问题的，因为大多数研究假设意图会直接决定个体的行为，而这往往并非如此（Carrington 等人，2010 年）。为了解决这个问题，除了研究自我报告的偏好外，我们还测量实际消费者选择，以更好地了解基于这两种态度的购买可能性。因此，在我们的研究中，现代态度中出现的所有三种测量，即态度、意图和行为，都得到了探讨和研究。

Overall, the main motivation behind this research is to better understand why the attitude-behavior gap exists in ethical consumerism and how a business manager can attenuate this gap. We therefore need to understand how and why attitudes change and the dissimilar way in which different attitudes impact behavior. This will allow business managers of ethical products to design and implement the necessary steps that need to be taken to ensure an actual change in behavior. Importantly, the domain of business ethics will obtain insights toward furthering ethical consumption.
总的来说，这项研究的主要动机是更好地理解为什么伦理消费中存在态度与行为之间的差距，以及企业管理者如何减轻这种差距。因此，我们需要了解态度如何以及为什么会改变，以及不同态度对行为产生影响的方式。这将使伦理产品的企业管理者能够设计和实施必要的步骤，以确保行为上的实际变化。重要的是，商业道德领域将获得进一步推动伦理消费的见解。

The rest of the paper is structured as follows. In the next section, we review existing literature to provide a background on the attitude-behavior gap in particular and its implications on ethical consumption. The section also provides an explanation of the dual model of attitudes, its various components, and its relevance to the current research. Following this, we develop the hypotheses. We then present an experimental overview of the two longitudinal studies followed by their details. The latter section also discusses the analyses and results for each study. We conclude with a discussion of the contributions of the current research, the implications of its findings, and an overview of its limitations and avenues for further research.
本文的其余部分结构如下。在下一节中，我们回顾现有文献，以提供有关特定态度-行为差距及其对道德消费的影响的背景。该部分还解释了态度的双重模型、其各个组成部分以及与当前研究的相关性。在此之后，我们提出了假设。然后，我们介绍了两项纵向研究的实验概述，然后是它们的详细信息。后一部分还讨论了每项研究的分析和结果。最后，我们总结了当前研究的贡献、其发现的影响以及其局限性和进一步研究的途径。

Ethical consumption is the purchase of a product that takes into consideration a particular ethical issue and is chosen freely by the individual consumer (Doane 2001; De Pelsmacker et al. 2005). For example, if the issue of animal welfare (e.g., cage-free eggs) is in question, certain consumers might be more willing to affiliate themselves with a particular brand and, hence, be more willing to choose that brand over others (Doane 2001; De Pelsmacker et al. 2005). At the same time, they might punish companies, products or brands that digress from ethical conduct according to them (considering the issue) by boycotting their products or by wanting to pay a lower price (Doane 2001).
道德消费是考虑特定道德问题并由个体消费者自由选择的产品购买（Doane 2001; De Pelsmacker 等，2005 年）。例如，如果动物福利（例如，无笼鸡蛋）是一个问题，某些消费者可能更愿意与特定品牌联系在一起，因此更愿意选择该品牌而不是其他品牌（Doane 2001; De Pelsmacker 等，2005 年）。同时，他们可能会惩罚那些从他们（考虑到问题）的道德行为偏离的公司、产品或品牌，通过抵制他们的产品或希望支付更低的价格（Doane 2001）。

Multiple studies have investigated whether customers value the ethical behavior displayed by a firm and have consistently found a preference for ethical products and the desire to reward firms that are ethical (e.g., Cowe and Williams 2000; De Pelsmacker et al. 2005). For example, Creyer and Ross (1997) measured the: (1) importance of ethicality of firm behavior; (2) willingness to reward ethical firm via purchasing behavior; (3) willingness to punish an unethical firm via (non) purchasing behavior; and (4) expectations regarding the ethicality of corporate behavior in society. Their results found that consumers expect firms to behave ethically and there is a desire to reward those that do. More recent studies (e.g., Singh et al. 2012; Markovic et al. 2015) have found a link between the ethical perception of a corporate brand and consumer outcomes such as brand loyalty and brand equity, in both goods and service sectors. These studies suggest that companies that engage in ethical practices (or are perceived as ethical by consumers) can benefit, while those that fail to do so can be negatively impacted.
多项研究已经调查了客户是否重视公司展示的道德行为，并一直发现客户偏好道德产品并希望奖励那些道德的公司（例如，Cowe 和 Williams 2000; De Pelsmacker 等人 2005 年）。例如，Creyer 和 Ross（1997 年）衡量了：（1）公司行为的道德性重要性；（2）通过购买行为奖励道德公司的意愿；（3）通过（非）购买行为惩罚不道德公司的意愿；以及（4）对社会中企业行为的道德性的期望。他们的研究结果发现，消费者期望公司行为道德，希望奖励那些做到这一点的公司。更近期的研究（例如，Singh 等人 2012 年；Markovic 等人 2015 年）发现了公司品牌的道德感知与消费者结果之间的联系，如品牌忠诚度和品牌资产，在商品和服务领域都是如此。这些研究表明，从事道德实践的公司（或被消费者认为是道德的公司）可以受益，而那些未能这样做的公司可能会受到负面影响。

Moreover, it is found that the nature of information presented impacts behavior (Folkes and Kamins 1999). Folkes and Kamins (1999) conducted a series of experiments to understand the effect of ethical/unethical information on attitudes. Participants were presented with information regarding firm behavior (unethical vs. ethical behavior) and product attribute information (superior vs. inferior attribute). The unethical behavior in question was child labor. Results found that unethical firm behavior elicited negative attitudes toward the firm. In fact, firms that actively helped victims were rated more positively than both firms that conducted unethical behavior and those that refrained from unethical behavior. It is evident that information regarding ethical/unethical behavior has a strong influence on consumer attitudes and we can conclude that a positive or negative piece of information about a brand will influence consumers' attitudes toward that brand.
此外，研究发现，所呈现的信息性质会影响行为（Folkes 和 Kamins，1999 年）。Folkes 和 Kamins（1999 年）进行了一系列实验，以了解道德/不道德信息对态度的影响。参与者被呈现了有关公司行为（不道德 vs. 道德行为）和产品属性信息（优越 vs. 劣质属性）的信息。所涉及的不道德行为是童工。结果发现，不道德的公司行为引发了对公司的负面态度。事实上，那些积极帮助受害者的公司被评价得比那些从事不道德行为和那些避免不道德行为的公司更积极。显然，有关道德/不道德行为的信息对消费者态度有很强的影响，我们可以得出结论，关于品牌的积极或消极信息将影响消费者对该品牌的态度。

The research question that the current research addresses is that consumers' attitude toward ethical/unethical brands, however, does not match their purchase behavior. The question which arises here is why, if attitudes are an indication of behavior, do consumers who have a negative attitude toward unethical business practices still affiliate themselves with unethical companies, even though they report that they intend to shop ethically?
当前研究所探讨的研究问题是消费者对道德/不道德品牌的态度，然而，并不符合他们的购买行为。这里出现的问题是，如果态度是行为的指示，为什么对不道德商业行为持负面态度的消费者仍然会与不道德公司联系在一起，即使他们声称他们打算进行道德购物？

Research demonstrates that consumers are willing to 'reward' those companies that operate in an ethical manner
研究表明，消费者愿意“奖励”那些以道德方式运营的公司
and 'punish' those that partake in unethical behavior (Creyer and Ross 1997). Due to increasing demand for ethical products and a rise in ethical consumerism, companies are increasingly interested in promoting their CSR activities and highlighting ethical business decisions, in the hope that profits can be positively impacted (Singh et al. 2012). However, while most consumers reveal the desire to partake in ethical consumption, very few actually do. This results in a gap arising from the difference in consumer attitudes toward ethical products and their purchase behaviors. Researchers have studied the gap and tried to understand why it exists. However, current findings are still relatively inconclusive.
对那些参与不道德行为的人进行“惩罚”（Creyer 和 Ross 1997）。由于对道德产品的需求增加以及道德消费主义的兴起，公司越来越有兴趣推广他们的企业社会责任活动，并强调道德商业决策，希望利润能够受到积极影响（Singh 等，2012）。然而，虽然大多数消费者表达了参与道德消费的愿望，但实际上很少有人这样做。这导致了消费者对道德产品的态度与购买行为之间存在差距。研究人员已经研究了这种差距，并试图理解它为什么存在。然而，目前的研究结果仍然相对不确定。

One possible explanation for this gap is the fact that in attitude research experimenters often employ respondent self-report methods where participants are likely to give socially desirable answers (Bray et al. 2011; De Pelsmacker et al. 2005; King and Bruner 2000). This is especially the case when participants want to adhere to expected norms and make a positive impression on either the researcher or other participants. The role of social desirability bias in obscuring connections between independent and dependent variables is well established in psychological literature (King and Bruner 2000). In terms of ethical consumerism, participants are likely to over report ethical actions to give the 'right' answer (Bray et al. 2011). As an example, Chung and Monroe (2003) conducted a study to examine social desirability bias in the context of ethical-decision. Participants were presented with five vignettes relating to business ethics situations and were required to answer a survey to measure ethical evaluation and intentions at the end of each vignette and found that the more unethical a situation was, the greater the social desirability bias. Therefore, when studying the impact of attitudes on behavior, it is not enough to measure explicit attitudes through self-report tasks (De Pelsmacker et al. 2005). While self-report measures have been identified as a problem, researchers still primarily use self-report measures to understand consumer attitudes, failing to account for an understanding of consumers' attitude gap.
这种差距的一个可能解释是，在态度研究中，实验者经常采用受访者自我报告方法，参与者很可能会给出社会上希望听到的答案（Bray 等，2011 年；De Pelsmacker 等，2005 年；King 和 Bruner，2000 年）。当参与者希望遵守预期规范并给研究者或其他参与者留下积极印象时，情况尤其如此。社会上希望性偏见在模糊自变量和因变量之间的联系方面在心理学文献中得到了充分确认（King 和 Bruner，2000 年）。在伦理消费主义方面，参与者很可能会过度报告道德行为以给出“正确”的答案（Bray 等，2011 年）。例如，Chung 和 Monroe（2003 年）进行了一项研究，以检验伦理决策背景下的社会上希望性偏见。参与者被呈现了与商业道德情境相关的五个小插图，并被要求在每个小插图结束时回答一份调查问卷，以衡量道德评估和意图，结果发现，情境越不道德，社会上希望性偏见就越大。 因此，在研究态度对行为的影响时，仅通过自我报告任务来测量明示态度是不够的（De Pelsmacker 等人，2005 年）。尽管自我报告测量被确定为一个问题，研究人员仍然主要使用自我报告测量来了解消费者的态度，未能考虑消费者态度差距的理解。

Our proposed solution to the above issue is the use of implicit attitudes as an additional measure which will allow for the measurement of attitudes without the influence of a social desirability bias. Implicit attitudes are immune to social desirability bias but have been largely untapped as an explanatory factor in ethical research. Based on the model of dual attitudes, we argue that explicit and implicit attitudes coexist and are distinct measures of attitudes. The former, while easy to collect, is susceptible to social desirability bias, whereas the latter is more difficult to measure. We would like to reiterate here that although explicit attitudes might be influenced by implicit attitudes (Dijksterhuis 2010), the two are distinct constructs which independently represent attitudes. Further, the discrepancy between explicit and implicit attitudes can be attributed also be attributed to the fact that they can compete with each and be different in nature or direction (Dijksterhuis 2010; Wilson et al. 2000). In this paper, we argue that the attitude-behavior gap exists in ethical consumerism due to the different effects implicit and explicit consumer attitudes have on behavior.
我们对上述问题的提出的解决方案是使用隐性态度作为额外的衡量标准，这将允许在没有社会期望偏见的影响下测量态度。隐性态度对社会期望偏见具有免疫力，但在道德研究中作为解释因素的利用程度较低。基于双重态度模型，我们认为显性态度和隐性态度共存且是态度的不同衡量标准。前者虽然易于收集，但容易受到社会期望偏见的影响，而后者更难以衡量。我们在此重申，尽管显性态度可能受到隐性态度的影响（Dijksterhuis 2010），但两者是独立代表态度的不同构建。此外，显性态度和隐性态度之间的差异也可以归因于它们之间可能存在竞争，并且在性质或方向上有所不同（Dijksterhuis 2010；Wilson 等人 2000）。在本文中，我们认为道德消费主义中态度与行为之间存在差距，这是因为隐性和显性消费者态度对行为产生不同的影响。

Explicit attitude has long been considered to be a key variable that influences behavior (Perugini 2005). In fact, traditional models of behavior, such as Ajzen's Theory of Planned Behavior (1985), consider explicit attitudes to be the key determinant of behavior. The term explicit attitude refers to attitudes commonly discussed and measured in social psychology; they commonly manifest as a response to a stimuli and are easily expressed through written and verbal self-report measures. When it comes to understanding ethical consumption, explicit attitudes have been widely used. For example, De Pelsmacker et al. (2005) surveyed participants to investigate their willingness to pay for Fairtrade coffee, Carrigan and Attalla (2001) conducted focus group interviews to understand whether ethics matter in purchase behavior, and Singh et al. (2012) conducted a survey to understand whether consumer perceived ethicality influences brand trust, affect and loyalty.
明示态度长期以来被认为是影响行为的关键变量（Perugini 2005）。事实上，传统的行为模型，如 Ajzen 的计划行为理论（1985 年），认为明示态度是行为的关键决定因素。明示态度一词指的是社会心理学中常被讨论和测量的态度；它们通常表现为对刺激的反应，并且可以通过书面和口头的自我报告措施轻松表达。在理解道德消费方面，明示态度已被广泛使用。例如，De Pelsmacker 等人（2005 年）调查参与者，以调查他们愿意为公平贸易咖啡支付多少，Carrigan 和 Attalla（2001 年）进行了焦点小组访谈，以了解道德是否影响购买行为，Singh 等人（2012 年）进行了一项调查，以了解消费者感知的道德性是否影响品牌信任、情感和忠诚度。

While a significant amount of research has cited social desirability bias and self-report issues as contributors to the gap between attitudes and behavior (King and Bruner 2000), most research still uses self-reported surveys, interviews and focus groups to obtain information on consumer attitudes toward ethical products (e.g., Shaw et al. 2006; Creyer and Ross 1997; Carrigan and Attalla 2001). Thus, in spite of research interest in understanding the drivers of the attitude-behavior gap, existing research in ethics has predominantly focused on using explicit attitudes to study ethical behaviors including consumption of ethical products, with minimal attention on implicit attitudes.
尽管大量研究已经指出社会期望偏差和自我报告问题是导致态度和行为之间差距的因素（King 和 Bruner，2000），但大多数研究仍然使用自我报告调查、访谈和焦点小组来获取有关消费者对道德产品态度的信息（例如，Shaw 等人，2006 年；Creyer 和 Ross，1997 年；Carrigan 和 Attalla，2001 年）。因此，尽管研究对理解态度与行为之间差距的驱动因素很感兴趣，但现有的伦理研究主要集中在使用明示态度来研究包括消费道德产品在内的伦理行为，对隐含态度的关注较少。

Implicit attitudes (a) manifest as actions or judgments; (b) are automatically activated; (c) influence responses which cannot be controlled (Greenwald et al. 1998). Prior research that has measured implicit attitudes has studied associations between gender, race and ethnicity, just to name a few (Greenwald et al. 1998). In general, implicit attitudes have been widely used in research domains such as psychology and social psychology. For example, Campbell et al. (2007) used the implicit attitude test (IAT henceforth) to understand narcissists' implicit attitudes about themselves; McCarthy and Thompsen (2006) used it
隐性态度表现为行为或判断；自动激活；影响无法控制的反应（Greenwald 等人，1998 年）。先前测量隐性态度的研究已经研究了性别、种族和族裔之间的关联，仅举几例（Greenwald 等人，1998 年）。总的来说，隐性态度已广泛应用于心理学和社会心理学等研究领域。例如，Campbell 等人（2007 年）使用隐性态度测试（以下简称 IAT）来了解自恋者对自己的隐性态度；McCarthy 和 Thompsen（2006 年）使用它。
to understand alcohol and smoking expectations and behavior; and Reynolds et al. (2010) utilized the IAT to understand individuals' implicit assumptions regarding business morality. However, despite an exhaustive literature review, we found no evidence of the relationship between implicit attitudes and ethical judgement and behavior. In fact, we did not find any studies relating implicit attitudes to either consumer behavior or intentions. Studying the impact of implicit attitude on unethical behavior could unearth a reason as to why explicit attitudes toward ethical consumption are not always transformed into purchase behavior.
理解酒精和吸烟的期望和行为；Reynolds 等人（2010 年）利用 IAT 来理解个人对商业道德的内在假设。然而，尽管进行了详尽的文献回顾，我们并未发现内在态度与道德判断和行为之间的关系的证据。事实上，我们没有发现任何研究将内在态度与消费者行为或意图联系起来。研究内在态度对不道德行为的影响可能会揭示明确态度对道德消费的购买行为并非总是转化的原因。

Wilson et al. (2000) aim to reconcile the coexistence of implicit and explicit attitudes through their Model of Dual Attitudes. While many attitude models assume that when people change their attitude, the previous attitude is replaced by the new one, the dual-attitude model suggests that individuals have both an automatic implicit attitude and an explicit attitude. These two attitudes can mutually coexist, even when they are incompatible (Wilson et al. 2000). This can occur through one of four mechanismsrepression, independent systems, motivated-overriding and automated overriding. The attitude which guides behavior is dependent on the cognitive capacity of the individual. That is, it is dependent on whether the explicit attitude can be retrieved and whether the explicit attitude overrides the implicit one (Wilson et al. 2000).
Wilson 等人（2000 年）旨在通过他们的“双重态度模型”调和内隐态度和外显态度的共存。虽然许多态度模型假设当人们改变态度时，先前的态度会被新的取代，但双重态度模型表明个体既有自动的内隐态度又有外显态度。这两种态度可以相互共存，即使它们是不兼容的（Wilson 等人，2000 年）。这可以通过四种机制之一来实现：压抑、独立系统、动机性覆盖和自动覆盖。指导行为的态度取决于个体的认知能力。也就是说，它取决于外显态度是否能被检索以及外显态度是否覆盖内隐态度（Wilson 等人，2000 年）。

They suggest that implicit attitudes are automatic, making them the default response when individuals do not have the motivation to retrieve explicit attitudes-especially through external cues. Due to the nature of attitudes, explicit attitudes change more easily compared to implicit attitudes. Accordingly, individuals can often adopt new explicit attitudes, even as their implicit attitude remains unchanged. Eventually, the new attitude can 'wear off' and the implicit attitude once again becomes influential (Wilson et al. 2000).
他们建议，内隐态度是自动的，使其成为个体在没有动机检索显性态度时的默认反应-尤其是通过外部线索。由于态度的性质，显性态度相对于内隐态度更容易改变。因此，个体通常可以采纳新的显性态度，即使他们的内隐态度保持不变。最终，新的态度可能会“消退”，内隐态度再次变得有影响力（Wilson 等人，2000 年）。

Based on the previous discussion of the influence of positive or negative information on consumer attitudes and the model of dual-attitude, we can infer that explicit attitudes are more easily influenced by the information presented. Accordingly, it is hypothesized that when exposed to information regarding ethical (unethical) business practices, consumers' explicit attitude toward the brand becomes more positive (negative). These effects, as explained in the dual effects model, are instantaneous. In the long run, however, the updated explicit attitudes tend to wear off.
根据先前讨论的积极或消极信息对消费者态度的影响以及双态度模型，我们可以推断显性态度更容易受到所呈现信息的影响。因此，假设当暴露于有关道德（不道德）商业实践的信息时，消费者对品牌的显性态度会变得更积极（消极）。正如双效应模型所解释的那样，这些影响是瞬时的。然而，从长远来看，更新的显性态度往往会逐渐消退。

In contrast to explicit attitudes, it is more difficult to change implicit attitudes as they are automated, default responses. The dual-attitude model suggests that implicit attitudes remain relatively constant in their evaluation of the attitude object (Rydell and McConnell 2006). However, this is based on the power as well as the valence of the information as highly positive or negative information could change both implicit and explicit attitudes. Accordingly, we propose that:
与明确态度相比，改变内隐态度更加困难，因为它们是自动化的、默认的反应。双态度模型表明，内隐态度在对态度对象的评价上保持相对恒定（Rydell 和 McConnell 2006）。然而，这是基于信息的力量以及价值观，高度积极或消极的信息都可能改变内隐和明确态度。因此，我们提出：

H1 A single incidence of exposure to positive (negative) information will increase (decrease) positive explicit attitude but not implicit attitude toward a brand in the short run, but not in the long run.
H1 一次暴露于积极（消极）信息会在短期内增加（减少）对品牌的积极显性态度，但不会在长期内增加（减少）对品牌的隐性态度。

Previous research has shown that an ethical perception of the firm positively influences consumers' intentions to buy (preference) (Carrigan and Attalla 2001; Luchs et al. 2010; Sen and Bhattacharya 2001). In trying to understand the disconnect between attitudes and behavior, researchers have tried to explain ethical consumerism by drawing on models which explain ethical consumer behavior. In particular, the 'Theory of Planned Behavior' (Ajzen 1991) and 'The General Theory of Marketing Ethics' (Hunt and Vitell 1986) have been used as key frameworks to understand ethical consumer behavior. Ajzen's (1985) theory of planned behavior is a framework, which underpins much of the research in ethical consumerism today. This framework suggests intentions are a motivator of behavior; the stronger the intention to perform, the more likely the intention will turn into action (Ajzen 1991).
先前的研究表明，公司的道德感知积极影响消费者购买意向（偏好）（Carrigan 和 Attalla 2001；Luchs 等人 2010；Sen 和 Bhattacharya 2001）。为了理解态度和行为之间的脱节，研究人员试图通过借鉴解释道德消费行为的模型来解释道德消费主义。特别是，“计划行为理论”（Ajzen 1991）和“营销伦理学一般理论”（Hunt 和 Vitell 1986）已被用作理解道德消费行为的关键框架。Ajzen（1985）的计划行为理论是今天道德消费主义研究的基础，该框架表明意图是行为的动机；意图越强烈，意图转化为行动的可能性就越大（Ajzen 1991）。

Another model widely used to understand individual decision making in the context of ethical consumption is Hunt and Vitell's General theory of Marketing Ethics (1986). The model is based on the principles of deontology (consumers focus on obligations or rules) and teleology (consumers are guided by consequences of the action) and suggests that consumers undertake assessments of alternative behaviors based on deontology or teleology to arrive at an overall ethical judgment (Hunt and Vitell 1986; Bray et al. 2011). Hunt and Vitell (1986) postulate that once individuals perceive a situation as having ethical content, they then devise a set of alternatives or actions in order to resolve the ethical problem. Once these alternatives or actions are formed, both a deontological and a teleological evaluation will occur (Hunt and Vitell 1986). Individuals' ethical judgment is a function of both of these evaluations.
另一个广泛用于理解伦理消费背景下个体决策的模型是亨特和维特尔的《营销伦理通论》（1986 年）。该模型基于义务论（消费者关注义务或规则）和目的论（消费者受行为后果引导），并建议消费者根据义务论或目的论对替代行为进行评估，以得出整体伦理判断（亨特和维特尔，1986 年；Bray 等，2011 年）。亨特和维特尔（1986 年）假设一旦个体认为某种情况具有伦理内容，他们就会设计一组替代方案或行动来解决伦理问题。一旦形成这些替代方案或行动，将会发生义务论和目的论评估（亨特和维特尔，1986 年）。个体的伦理判断是这两种评估的功能。

These models suggest that the evaluation of available information/situation (related to ethicality of products/ brands/companies in our case) could influence consumers' attitudes (Ajzen 1985) and judgments (Hunt and Vitell 1986) which, in turn, influence their intentions to behave. However, research suggests that intentions are useful for predicting a single action (e.g., I will buy that hat) rather than multiple 'single actions' (i.e., I will keep on buying that hat) (Sheeran 2002). To observe a change in ethical
这些模型表明，对可用信息/情况（与我们案例中的产品/品牌/公司的道德性有关）的评估可能会影响消费者的态度（Ajzen 1985）和判断（Hunt 和 Vitell 1986），进而影响他们的行为意图。然而，研究表明，意图对于预测单个行为（例如，我会买那顶帽子）是有用的，而不是多个“单个行为”（即，我会继续购买那顶帽子）（Sheeran 2002）。要观察道德变化
consumerism, consumers need to undertake multiple 'single actions.' Accordingly, intention might not be an accurate predictor of purchase behavior which is sustained over a longer time period. This is the reason we study behavior separately in the current research, in addition to preference. Further, self-reported intention (measured as brand preference in the current research) is more closely related to explicit attitudes and thus tends to dissipate over time. Hence, we propose the following hypothesis:
消费主义，消费者需要承担多个“单一行动”。因此，意图可能不是购买行为的准确预测因素，这种行为会持续较长时间。这就是为什么我们在当前研究中单独研究行为的原因，除了偏好。此外，自我报告的意图（在当前研究中以品牌偏好来衡量）与明确态度更密切相关，因此随着时间的推移往往会消失。因此，我们提出以下假设：

H2 A single incidence of exposure to positive (negative) information will increase (decrease) brand preference for a brand in the short run, but not in the long run.
H2 一次暴露于积极（消极）信息会在短期内增加（减少）品牌偏好，但在长期内不会。

The mechanism that we propose behind the above predictions is that brand preference will primarily be affected by explicit attitudes. As discussed, implicit attitudes which are more stable will not change in the short or long run, due to a single exposure to positive or negative information and will thus not lead to changes in brand preference. This implies that explicit attitudes will drive the changes in brand preference in the short (but not long) run. We, therefore, hypothesize that:
我们提出的上述预测背后的机制是，品牌偏好主要受明示态度的影响。正如讨论的那样，更稳定的内隐态度不会因为对正面或负面信息的单次暴露而在短期或长期内发生变化，因此不会导致品牌偏好的变化。这意味着明示态度将在短期内推动品牌偏好的变化（但不是长期）。因此，我们假设：

H3 For a single incidence of exposure to positive (negative) information, explicit attitudes will increase (decrease) the brand preference in the short run, but not in the long run.
H3 对于一次暴露于积极（消极）信息的情况，明示态度会在短期内增加（减少）品牌偏好，但在长期内不会增加（减少）。

The previously mentioned behavior models lay the foundation for much of the research in ethical consumerism. However, they do make the leap of linking intention to actual behavior because intentions are viewed as motivations to perform behavior (Sheeran 2002). Based on this assumption, research in ethical consumerism adopts the approach of measuring purchase intention without considering purchase behavior or variables that more closely mimic actual behavior. The issue with this, however, is that intentions have repeatedly been shown to be poor predictors of actual behavior (e.g., Cowe and Williams 2000; Carrigan and Attalla 2001). Recent research in ethical consumerism while aiming to understand the underlying mechanism behind the intention-behavior gap (e.g., Carrington et al. 2014) has called for more empirical research in this area in order to better quantify the magnitude of this gap (Hassan et al. 2015).
先前提到的行为模型为伦理消费研究奠定了基础。然而，它们确实进行了从意图到实际行为的跳跃，因为意图被视为执行行为的动机（Sheeran 2002）。基于这一假设，伦理消费研究采用了衡量购买意向的方法，而不考虑购买行为或更接近实际行为的变量。然而，问题在于意图一再被证明是实际行为的较差预测因素（例如，Cowe 和 Williams 2000；Carrigan 和 Attalla 2001）。最近的伦理消费研究旨在理解意图与行为之间差距的潜在机制（例如，Carrington 等人 2014 年），呼吁在这一领域进行更多的实证研究，以更好地量化这一差距的大小（Hassan 等人 2015 年）。

Existing research suggests that consumers do not always disclose their true attitudes about ethical products as most attitude measures are self-reported tasks which lead to socially desirable answers (King and Bruner 2000; De Pelsmacker et al. 2005). In addition, research also suggests that extraneous factors such as contextual information could affect the attitude, making consumers change their attitude toward a brand at the moment of choice/action (Carrington et al. 2010). The authors note that it is possible for ethical judgment to differ from purchase intention, as an alternative may be chosen due to proffered consequences such as benefit to oneself. When ethical judgment is inconsistent with behavior and intention, guilt can occur, while consistency between judgment, behavior and intention can result in greater satisfaction (Hunt and Vitell 1986; Bray et al. 2011).
现有研究表明，消费者并不总是透露他们对道德产品的真实态度，因为大多数态度测量是自我报告任务，导致社会上希望的答案（King 和 Bruner 2000; De Pelsmacker 等人 2005）。此外，研究还表明，诸如情境信息之类的外部因素可能影响态度，使消费者在选择/行动时改变对品牌的态度（Carrington 等人 2010）。作者指出，道德判断可能与购买意向不同，因为可能会选择另一种选择，因为其带来的后果，比如对自己的利益。当道德判断与行为和意图不一致时，可能会产生内疚感，而判断、行为和意图之间的一致性可能会导致更大的满足感（Hunt 和 Vitell 1986; Bray 等人 2011）。

As can be surmised, the differences in implicit and explicit attitude retrieval can have direct relevance for the behavioral patterns that follow (Dimofte 2010). By focusing on explicit attitudes alone, the previous inquiry into ethical consumption studies may have exaggerated the ease with which (a) people change their attitude and (b) people change their behavior. This is because even if an explicit attitude changes, an implicit attitude can remain constant leading to different evaluations of the same attitude object (Wilson et al. 2000). For example, recent reports indicated that the clothing company Zara was using questionable labor practices in its Argentinean factories (Osborne 2013). For a Zara customer, this report of company's unethical behavior could result in a negative adjustment of explicit attitudes toward the company. However, consumer's implicit attitudes may retain their positive brand associations and get evoked as the individual passes by a Zara store and remembers an occasion when a Zara outfit was worn. We thus propose that actual behavior will be influenced by implicit attitudes and not by explicit attitudes. Specifically, we hypothesize that: 

H4 For a single incidence of exposure to positive (negative) information, implicit attitudes (but not explicit attitudes) will predict the actual brand choice.
H4 对于一次暴露于积极（消极）信息的情况，内隐态度（而不是明示态度）将预测实际品牌选择。

Two longitudinal studies were conducted to examine the extent to which implicit and explicit attitudes influence choice behavior of ethical products, given the nature of the information presented.
进行了两项纵向研究，以检验内隐和外显态度在伦理产品选择行为中的影响程度，考虑到所呈现信息的性质。

Fifty students from a large university were recruited to participate in the pretest in order to develop and test the manipulation used in the main studies. Each participant was randomly assigned to one of the two conditions. The stimuli in the pretest were tested to establish that the valence of the information presented on the brand (CocaCola) was perceived as intended. Participants were presented with either positive or negative information about the company. Positive information was sourced from an article outlining an instance of the company's ethical
从一所大学招募了五十名学生参加预测试，以开发和测试主要研究中使用的操作方法。每位参与者被随机分配到两种条件之一。预测试中的刺激被测试以确保品牌（可口可乐）上呈现的信息的情感被认为是预期的。参与者被呈现了关于公司的正面或负面信息。正面信息来自一篇概述公司道德实例的文章。
behavior (Oxfam International 2013). Negative information was sourced from a news article and outlined an instance of the company's unethical behavior (Merrill 2013). Coca-Cola was selected as the company because participants were expected to have fairly well formed implicit as well as explicit attitudes toward its products. Data on three other well-known companies-Apple, Sony and Nestle-were also collected, but based on the strength of the manipulation, Coca-Cola was retained for further analysis.
行为（Oxfam International 2013）。负面信息来自一篇新闻文章，概述了该公司不道德行为的一个实例（Merrill 2013）。可口可乐被选为研究对象，因为预计参与者对其产品有相当明确的内隐和外显态度。还收集了另外三家知名公司的数据-苹果、索尼和雀巢，但基于操纵的强度，可口可乐被保留用于进一步分析。

Once the articles had been read, participants completed a short survey in which they indicated whether the information depicted the company in a favorable/unfavorable light; whether the actions outlined were good/bad and whether they agreed/disagreed with the actions outlined. Participants were also asked to indicate their attitude toward Coca-Cola in terms of the brand (a) respecting moral norms, (b) adhering to the law, (c) being a socially responsible brand, (d) avoiding damaging behavior at all costs and (e) overall favorability. These items were based on Brunk's (2012) consumer perceived ethicality (CPE) scale. CPE has been successfully adapted in research to understand consumer's explicit attitudes toward business practices (e.g., Singh et al. 2012). Participants were asked to respond on a 7 -point Likert scale 'Strongly Disagree'; 'Strongly Agree'). The five measures displayed an acceptable level of reliability . 

Two responses were incomplete and were dropped, resulting in a final sample of 48 participants (females . Twenty-six participants were exposed to the positive information condition, and 22 to the negative information condition. Independent tests were conducted to determine the difference in the ratings of CocaCola between the two conditions. Results reveal a significant difference between the two information conditions on the averaged, composite CPE score across the conditions There was also a significant difference in whether participants agreed with the actions outlined in the article and whether they saw the actions as 'good' . Thus, the pretest establishes that the manipulation is successful in influencing the brand's perceived ethicality, as intended.
两个回应不完整，被删除，最终样本为 48 名参与者（女性 。26 名参与者暴露于积极信息条件，22 名参与者暴露于消极信息条件。进行了独立 检验以确定两种条件下对可口可乐评分的差异。结果显示两种信息条件在跨条件的平均复合 CPE 得分上存在显著差异 。参与者是否同意文章中概述的行动 以及他们是否认为这些行动是“好的” 也存在显著差异。因此，预实验表明操纵成功地影响了品牌的道德感知，正如预期的那样。

The first objective of Study 1 was to understand whether the valence of the information influences implicit and explicit attitudes in a dissimilar way. The second was to
研究 1 的第一个目标是了解信息的情感价值是否以不同的方式影响内隐和外显态度。第二个目标是

investigate whether shifts in attitudes (implicit/explicit) lead to changes in the preference for the brand. Third, it examined whether attitude changes and preference decisions are sustained over a period of time.
调查态度（内隐/外显）的转变是否会导致对品牌偏好的改变。第三，它检查了态度的变化和偏好决策是否能够持续一段时间。

To test these objectives, the study implemented a onefactor, two-level (information: positive, negative) betweensubjects design where explicit attitudes, implicit attitudes and brand preference were the dependent variables of interest. One hundred and two participants were recruited from the undergraduate population of a large university in exchange for financial remuneration. After dropping 17 invalid and incomplete responses across the two time periods, data from 85 participants were utilized in the analysis. The data dropped were primarily due to errors in responses on the 'Implicit Attitude Test'. The study design is illustrated in Fig. 1a.
为了测试这些目标，研究实施了一个单因素、两水平（信息：积极、消极）的受试者设计，其中明示态度、隐含态度和品牌偏好是感兴趣的因变量。从一所大学本科生群体中招募了 102 名参与者，以金钱报酬作为交换条件。在两个时间段内剔除了 17 个无效和不完整的回应后，来自 85 名参与者的数据被用于分析。被剔除的数据主要是由于在“隐含态度测试”中的回应错误。研究设计如图 1a 所示。

The study was conducted in three stages over two time periods (TP1 and TP2) to allow us to study the change in implicit and explicit attitudes over time. In order to reduce the social desirability bias, both stages of the study were conducted in an experimental laboratory where each participant was seated in private cubicles providing complete privacy. Further, no personal information that could identify individual participants was collected. Participants were asked only for their student numbers, which were solely used to match responses across the two time periods and then discarded.
该研究分为三个阶段，在两个时间段（TP1 和 TP2）内进行，以便我们研究隐性和显性态度随时间的变化。为了减少社会期望偏见，研究的两个阶段均在实验室进行，每位参与者均坐在私人隔间中，提供完全的隐私。此外，没有收集能够识别个体参与者的个人信息。参与者只被要求提供他们的学号，这些学号仅用于匹配两个时间段的回答，然后被丢弃。

Time Period 1 (Stages 1 and 2) In the first time period (TP1), a filler task separated the two stages of data collection, Stages 1 and 2. The second time period (TP2), which we describe later, involved a single instance of data collection, which we label Stage 3. A detailed explanation of the setup is provided in Fig. 1.
时间段 1（阶段 1 和 2）在第一个时间段（TP1）中，一个 填充任务将数据收集的两个阶段，即阶段 1 和 2，分开。第二个时间段（TP2），我们稍后会描述，涉及一次数据收集的单个实例，我们称之为第 3 阶段。设置的详细说明请参见图 1。

Stage 1 of TP1 The baseline levels of explicit and implicit attitudes toward Coca-Cola were measured first. Each participant completed the explicit attitude scale followed by the IAT. This was done to understand participants' initial attitudes toward Coca-Cola, before the presentation of any information. Participants were also asked to indicate their current preference for CocaCola, before being exposed to the article in which positive or negative information was presented. After completing the initial measures of attitudes, participants were randomly assigned to one of the two information conditions where they were asked to read the pretested
TP1 的第一阶段首先测量了对可口可乐的明示和隐含态度的基线水平。每位参与者先完成明示态度量表，然后进行 IAT 测试。这样做是为了了解参与者在接收任何信息之前对可口可乐的初始态度。在被呈现任何信息之前，参与者还被要求指出他们目前对可口可乐的偏好。在完成态度的初始测量后，参与者被随机分配到两种信息条件中的一种，要求他们阅读经过预测试的文章。

Fig. 1 Setup of experiments: timing and collection of measures
图 1 实验设置：时间和测量收集

article outlining either an ethical or an unethical behavior, undertaken by Coca-Cola. After reading the article, participants were asked to read a filler article about the current weather in the city. Both articles had a minimum of reading time allocated to them. The filler task was built-in to provide temporal distance between the measurements of explicit and implicit attitude, post-exposure to brand information.
文章概述了可口可乐公司进行的道德或不道德行为。阅读完文章后，参与者被要求阅读一篇关于城市当前天气的填充文章。这两篇文章都至少分配了 的阅读时间。填充任务是为了在接受品牌信息后，提供明示和隐含态度测量之间的时间距离。

Stage 2 of TP1 Once the two articles had been read, participants once again completed the measures for explicit and implicit attitudes. This was done to enable us to determine whether the baseline attitudes had changed after being exposed to the information (positive or negative). Next, participants answered the questions relating to brand preference again, in order to understand whether preferences changed across the two stages. Finally, they completed questions related to demographics such as gender, income, family size, employment status, age, educational qualification and whether English was their first language. Participants were then thanked for their time and informed that they would be contacted via email for follow-up tests.
TP1 的第 2 阶段一旦两篇文章被阅读完毕，参与者再次完成了明示和隐含态度的测量。这样做是为了让我们能够确定在接受信息（积极或消极）后基线态度是否发生了变化。接下来，参与者再次回答与品牌偏好相关的问题，以了解偏好是否在两个阶段之间发生了变化。最后，他们完成了与人口统计学相关的问题，如性别、收入、家庭规模、就业状况、年龄、教育资格以及英语是否为他们的第一语言。然后感谢参与者的时间，并告知他们将通过电子邮件进行后续测试。
Time Period 2 (Stage 3) Two weeks after completing Stages 1 and 2 at TP1, participants were invited to participate in Stage 3 at time period 2 (TP2). Out of the 102 initial participants, 93 successfully completed Stages 1 and 2 and were invited to Stage 3. After recording participants' identification number, their explicit and implicit attitudes toward Coca-Cola were measured. The measures used for explicit attitudes, implicit attitudes and preferences were identical to those used in TP1. As our intention was to study the longitudinal impact of information presented at Stage 1, no additional information was presented at Stage 3. 

The purpose of this stage was to determine (a) whether the nature of information presented in Stage 1 had a sustained influence on attitudes; (b) the difference in attitudes across the three measurement times; twice in TP1-Stages 1 and 2 (pre- and post-information) and once in TP2, Stage 3 (two weeks later). Participants completed the explicit attitude scale prior to completing the IAT. Finally, participants answered questions relating to brand preference.
本阶段的目的是确定：(a)第 1 阶段呈现的信息性质是否对态度产生持续影响；(b)在三个测量时间点上态度的差异；在 TP1 的两次测量中（第 1 和第 2 阶段，信息呈现前后）以及在 TP2 的第 3 阶段（两周后）各测量一次。参与者在完成 IAT 之前完成了明示态度量表。最后，参与者回答了与品牌偏好相关的问题。

Measurement of Explicit Attitudes As in the pretest, explicit attitudes were measured through a self-report survey using the five statements comprising the Consumer Perceived Ethicality (CPE) scale. Participants were asked to indicate how strongly they agreed or disagreed with the statements on a 7-point Likert scale ( 'Strongly Disagree'; 'Strongly Agree').
明示态度的测量与预测试一样，通过自我报告调查使用包含消费者感知道德性（CPE）量表的五个陈述 来测量明示态度。参与者被要求在 7 点力克特量表上指出他们对这些陈述的赞同或反对程度（ “强烈反对”； “强烈同意”）。

Measurement of Implicit Attitudes Implicit attitudes toward Coca-Cola were measured through the use of an implicit association test (IAT) created by Inquisit (2014), by following the process developed by Greenwald et al. (1998). Based on the research that has utilized the IAT, Greenwald et al. (1998) posit that the test provides a windows into the level of mental operation associated in unthinking fashion based on the unconscious mental operations, the automatic, the implicit and the intuitive, and hence are reflective of 'truer' attitudes than conscious ones. The IAT has also been shown to be resistant to the social desirability bias (Egloff and Schmukle 2002).
对可口可乐的内隐态度进行了测量，通过使用由 Inquisit（2014）创建的内隐联想测试（IAT），遵循了 Greenwald 等人（1998）制定的过程。基于利用 IAT 的研究，Greenwald 等人（1998）认为该测试提供了一个窗口，可以了解与无意识心理操作相关的心理运作水平，自动、内隐和直觉，因此反映了比意识态度更“真实”的态度。IAT 还被证明对社会期望偏见具有抵抗力（Egloff 和 Schmukle，2002）。

Our measure involved a valence-based IAT (Greenwald et al. 1998) consisting of all seven stages of the IAT procedure, and thus allows for truer attitudes than Brief IAT (Sriram and Greenwald 2009), which is more commonly used in the literature. For this test, participants were shown repeated paired groups of ethical (Fairtrade) and unethical (Coca-Cola) brand logos and products combined with positive (e.g., environmentally friendly) and negative (e.g., corruption) words and phrases. When completing the IAT, subjects had to respond to two target concept discriminations: (a) Coca-Cola and (b) Fairtrade. Each of these target concept discriminations was used in combination with the positive and negative words. There were a total of seven images which displayed the Coca-Cola brand logo and seven which displayed the Fairtrade brand logo. These brand logos were used in combination with seven positive words and seven negative words (refer to 'Appendix').
我们的测量涉及基于价值的 IAT（Greenwald 等人，1998 年），包括 IAT 程序的所有七个阶段，因此比 Brief IAT（Sriram 和 Greenwald，2009 年）更常用于文献中，从而允许更真实的态度。对于这个测试，参与者被展示了重复的道德（公平贸易）和不道德（可口可乐）品牌标识和产品的配对组合，结合了积极（例如，环保）和消极（例如，腐败）的词语和短语。在完成 IAT 时，受试者必须回答两个目标概念的区分：（a）可口可乐和（b）公平贸易。这两个目标概念的区分与积极和消极的词语结合使用。总共有七幅显示可口可乐品牌标识的图像，以及七幅显示公平贸易品牌标识的图像。这些品牌标识与七个积极词语和七个消极词语结合使用（参见“附录”）。

As per Greenwald et al.'s (1998) method, the IAT began with an introduction to the target concept discriminationethical or unethical brand logos. Participants then needed to distinguish between logos which they considered to be ethical or unethical by responding to one category with their right hand (e.g., pressing the ' ' key when unethical logos were shown) and one with their left (e.g., pressing the 'e' key when ethical logos were shown). Subsequently, participants were introduced to the attribute dimension, categorizing pleasant meaning words from unpleasant meaning words (using the same method outlined above). 

Measurement of Brand Preference Brand preference was used as the dependent variable since it has been found to usually predict an individual's likelihood and frequency of purchase (Bass et al. 1972). Accordingly, participants were asked to indicate their preference for Coca-Cola on a 7-point Likert scale 'Strongly Disagree'; 'Strongly Agree'). In addition to Coca-Cola, data for other beverage companies (Pepsi, Golden Circle and Lipton) were also collected. 

Explicit Attitudes Explicit attitudes were collected at the baseline level (pre-exposure to information) in Stage 1 (EA1), at Stage 2, post-exposure to information (EA2), and finally, at Stage 3 after the two-week time delay (EA3). In Stage 3, eight participants did not successfully complete the IAT and were hence dropped from the analysis, leaving a final sample of 85 participants. Of these, 43 participants were in the negative information condition and 42 in the positive information condition. The results of the analyses are presented in Table 1 and illustrated in Fig. 2a.
明示态度明示态度在阶段 1（EA1）基线水平（信息曝光前）、在阶段 2 信息曝光后（EA2）以及最终在两周时间延迟后的阶段 3（EA3）收集。在阶段 3，有 8 名参与者未能成功完成 IAT 测试，因此被剔除分析，最终样本为 85 名参与者。其中，43 名参与者处于负面信息条件，42 名处于正面信息条件。分析结果见表 1，并在图 2a 中呈现。

Across condition variations: Pre-exposure (EAI)-As expected, no difference was found across the two conditions for pre-exposure levels . . This shows that participants in both information conditions were introduced to the experiment without any pre-existing differences in explicit attitudes toward the focal brand. Post-exposure (EA2)-After reading the article, the means shifted vs. and were statistically different [Satterthwaite , with participants in the positive condition showing a more positive explicit attitude toward the brand. This provides support for H1. Timedelayed (EA3)-When explicit attitudes were collected after a temporal delay at Stage 3 in TP2, no difference was found across the two means vs. . These results again support H1 by addressing two points. First, they show that the positive and negative conditions were identical in their baseline explicit preference toward the brand, pre-exposure to the information. Second, participants' explicit attitudes changed significantly due to the information but went back to similar levels after the time delay. 

Stimulus induction variations: To test changes in explicit attitudes as a result of the exposure to information, pairedsamples tests were conducted. For the negative condition,
刺激诱导变化：为了测试由于接受信息而导致明示态度的变化，进行了成对样本 t 检验。对于负面条件，

Table 1 Study 1: differential induction based instantaneous and longitudinal changes in dual attitudes and brand preference
表 1 研究 1：基于瞬时和纵向变化的双重态度和品牌偏好的差异诱导

Bold values indicate that the test is statistically significant
粗体数值表示测试在统计上显著

** Implies that the test is significant at
** 表示测试在 上是显著的

results show a significant difference in mean ratings of EA1 and vs. , . The same was found for the positive information condition vs. , . This provides support for H1 since it shows that a single incidence of exposure to positive (negative) information can lead to an increase (decrease) in the explicit attitude toward the brand. 

Longitudinal induction variation: There was no significant difference between the means of EA1 and EA3 for either the negative condition vs. or the positive condition vs. . This provides support to . Explicit attitudes toward the brand are found to return to their pre-manipulation levels; that is, changes in explicit attitude dissipate over the long run. 

Implicit Attitudes Implicit attitudes toward Coca-Cola were analyzed across the two time periods and the three stages; IA1 (pre-exposure, TP1), IA2 (post-exposure, TP1) and IA3 (time delay, TP2). As with explicit attitudes, we tested for across condition, stimulus induction as well as longitudinal induction variations. The results are discussed below and are reported in Table 1 and Fig. 2b. 

Across condition variations: Pre-exposure (IA1)—As expected, no difference was found in the two conditions for pre-exposure levels vs. . That is, participants did not display any pre-existing differences in implicit attitudes toward the brand. Post-exposure (IA2)—After reading the article with positive or negative information about the brand, no significant shift in implicit attitudes was observed vs. ; . Time-delayed (IA3)—When implicit attitudes were collected after a temporal delay in Stage 2, no difference was found across the two means vs. . The results show that across the two information conditions, implicit attitudes are typically stable. This supports hypothesis 1. 

Stimulus induction variations: To test changes in implicit attitudes (pre- and post-information), paired-samples tests were conducted. In the negative information condition, results found no significant difference between means of and vs. , ns]. The same was true for the positive information condition vs. , . The results support , which states that implicit attitudes will not be affected by single exposure to positive or negative information in the short run. 

Longitudinal Induction variation: There was no significant difference between the means of IA1 and IA3 for either the negative vs. ; or the positive vs. information conditions. The results show that as predicted by , no changes occur in implicit attitudes in the relatively long run either, with a single information exposure. 

Brand Preference The self-reported brand preference for Coca-Cola was analyzed across the three stages during the two time periods; PREF1 (pre-exposure, Stage 1), PREF2 (post-exposure, Stage 2) and PREF3 (time delay, Stage 3). The results are condensed below and are reported in Table 1 and Fig. 2c. 

Study 1 - Changes in Dual Attitudes and Brand Preference. 

Fig. 2 Study 1-changes in dual attitudes and brand preferences. (a) Variation in explicit attitudes, (b) variation in implicit attitudes, (c) variation in brand preference 

Across condition variations: Pre-exposure (PREF1)— No difference was found between the means for the two information conditions for the baseline levels vs. . This establishes that participants did not come in with differences in their brand preference for Coca-Cola, across the two conditions. Post-exposure (PREF2)After reading the article, there were substantial changes in self-reported brand preference with participants exposed to the negative condition reporting a decrease 
in preference for the brand versus those exposed to the positive information vs. 4.80; Time-delayed (PREF3)—When preferences were collected after a temporal delay in Stage 3, no difference was found across the means for brand preference across the information conditions vs. . As in the previous cases, we find that significant changes in brand preference occurs across positive and negative information conditions in the short run, but these effects dissipate in the long run. The pattern, therefore, replicates the one seen for explicit attitudes and supports . 

Stimulus induction variations To test changes in preferences, paired-samples tests were conducted across the stages. For the negative information condition, results found a significant difference in means of PREF1 and PREF2 vs. . The same was true for the positive condition vs. . The results, therefore, support such that the brand preferences show an increase in the positive condition and a decrease in the negative condition, in the short run. 

Longitudinal induction variation There was no significant difference in the means of PREF1 and PREF3 for either the negative vs. ; or the positive vs. information conditions. As in the case of explicit attitudes, preference shifts due to positive or negative information decayed over time, supporting . 

The Relationship Between Attitudes and Brand Preference In order to adequately address H3, an OLS regression analysis was run to understand the impact of implicit and explicit attitudes on brand preference. Gender, age, employment status, family size and education were included in the analyses as covariates. The stage of the data collection and the nature of information (positive or negative) were also included for a fixed effects analysis. Due to the lack of variability in age, employment status and education in our student sample, these variables were dropped from further analyses. The (a) brand preference, (b) self-reported explicit attitudes and (c) the implicit attitudes, collected across all three stages, were utilized to estimate multiple OLS models. For the sake of parsimony,  

we only report the results from the pooled data across all stages and both information conditions. The results are displayed in Table 2. Of the focal variables of interest, we found that explicit attitudes predict brand preference . This confirms H3 which states that explicit attitudes will drive brand preference for both positive and negative information conditions in the short run. 

Although we do not hypothesize it, the effect of implicit attitudes was also found to be significant in predicting brand preference . The primary observation from our analyses of the relationships is that explicit attitudes do drive brand preference for participants in both the positive and the negative conditions. However, implicit attitudes are also found to have a significant effect on brand preferences. 

Study 2 had two main objectives. First, Study 2 seeks to replicate the findings from Study 1 and increase their generalizability across different sample populations. Specifically, Study 2 uses a non-student sample. This is in contrast to Study 1, which utilized a student sample. Second, Study 2 extends the findings from Study 1 by adding actual choice to the study since that is a focal variable for business ethics rather than just self-reported preference or intention measures. As in Study 1, a longitudinal betweensubjects design was utilized to achieve the aims of the study. 

The surveys were completed online on a computer in a private cubicle in the laboratory, and participants were instructed to include only pre-assigned, anonymous codes which were to be used to match responses across time periods 1 and 2 (TP1 and TP2), and not for identification purposes. In order to encourage participation, subjects were paid for each round of participation. The experimental design of for Study 2 is presented in Fig. 1b and, other than the choice measures collected, replicates Study 1. A detailed explanation of the setup is provided below. 

Participants were recruited from a panel administered by a university marketing research laboratory. One hundred and two participants completed the study. Responses from seventeen participants were discarded as they failed to follow the instructions correctly and did not answer components of the study, especially the IAT. Data from a further seven subjects had to be removed as they failed to complete Stage 3 of the study. Accordingly, data from a total of 78 participants were used in the analyses of results 

Table 2 Brand preference and actual choice as explained by a dual-attitude model 

** Implies that the test is significant at  

(females years old . As in Study 1, a one-factor, two-level (information: positive, negative) between-subjects design was used in Study 2. 

Time Period 1 (Stages 1 and 2) Participants were required to come to a university computer laboratory to participate in the study. Upon entering, participants were allocated cubicle numbers at random and were instructed to be seated at their cubicle and read the consent form. Once done, participants pressed the next button and completed the study using the same procedure as used for Stages 1 and 2 during time period 1 in Study 1. Before leaving, however, participants were offered a choice of a can of soft-drink to thank them for their participation along with some light snacks. This choice was made in isolation to avoid any possibility of social desirability in choice. These choices were later recorded and matched with the anonymous identification number assigned to each participant. 

Time Period 2 (Stage 3) Two weeks after completing Stages 1 and 2, in time period 1 (TP1), participants were invited to take part in stage 3 of the study. Invitations were sent to 85 participants who had successfully completed the study in TP1. Seven of the 85 responses were removed due to errors and missing information. As in Stage 3 of Study 1, participants were not presented with any information at Stage 3 in TP2. The procedure followed was identical to that from Study 1, Stage 3. However, in addition to the other measures collected in Study 1, we also measured choice. The choice measure is identical to that implemented at the end of Stage 2 in TP1. 

The measures of explicit and implicit attitudes were identical to those used in Study 1 as were the stimuli presented. In addition, we also collected data on brand preference as in the previous study. Participants also answered questions on demographics including gender, income, family size, employment status, age, educational qualification and whether English was their first language. 

While preference has been shown to predict choice behavior, high predictive power is difficult to achieve (Bass et al. 1972). Study 1 measured brand preference, but we wanted to replicate the findings with actual choice behavior in Study 2. As participants were in a laboratory setting, which offered us complete control over the environment, an actual choice element was introduced in the experiment rather than just a self-reported measure of choice. 

To measure actual choice, participants were offered a choice among six carbonated beverages; three from CocaCola and three from Pepsi. To keep the choice task realistic, participants also had the option of not choosing any of the products. We chose Pepsi as the other brand in the choice task as both Pepsi and Coca-Cola are carbonated soft drinks, of similar color and flavor. We also chose flavors for each brand which were comparable to the other. For Coca-Cola, we had Coca-Cola Classic, Coca-Cola Zero and Coca-Cola Diet. Similarly, for Pepsi we had Pepsi, Pepsi Next and Pepsi Max. Participants' choices were recorded by observation. 

To be able to analyze the effect of implicit and explicit attitudes on choice, it is important to ascertain that the results obtained in Study 1 with respect to the shifts in both types of attitudes are also observed in Study 2. We therefore, replicate the analyses conducted for Study 1 on the data collected for Study 2. 

Explicit Attitudes Explicit attitudes were once again measured and analyzed at the three stages-stages 1 (EA1) and 2 (EA2) at TP1 1 and Stage 3 (EA3) at TP2. Independent-samples tests were run to determine whether the mean responses of EA1, EA2 and EA3 were different 
across the two conditions (positive or negative information). There were 36 participants in the negative information condition and 42 in the positive information condition. The results of the analyses are presented in Table 3 and graphed in Fig. 3a. 

Across condition variations No difference was found in the two pre-exposure conditions (EA1) levels 4.65 vs. . After exposure to the information, the means across the conditions were found to be significantly different vs. , such that participants in the positive condition had a higher explicit attitude toward the brand than those in the negative one. Explicit attitudes collected after the two-week delay (EA3) in TP2 reverted to similar levels, and no difference was found across the two means vs. . The results, therefore, replicate our findings in Study 1 and support . 

Stimulus induction variations To assess whether information influenced explicit attitudes, we ran pairedsamples tests across conditions. Results find that for the negative condition, the difference between the means, EA1 and EA2, was significant vs. . The same was found for the positive condition vs. . Thus, the results provide further validation of our findings from Study 1 and for . 

Longitudinal induction variation Again, replicating Study 1, the two-week time delay led to the dissipation of attitude shift due to exposure to information in Stage 2. Specifically, we found no significant difference in the means from Stage 1 (EA1) and Stage 3 (EA3) for either the negative vs. ; , ns] or the positive vs. information conditions. is thus again supported by our findings. 

Implicit Attitudes As in Study 1, the three measures of implicit attitudes toward Coca-Cola collected at different stages and time periods were analyzed; IA1 (pre-exposure, Stage 1, TP1), IA2 (post-exposure, Stage 2, TP1) and IA3 (time delay, Stage 3, TP2). The results are presented in Table 3 and Fig. 3b. 

Across condition variations In keeping with Study 1, there was no significant difference in implicit attitudes across the two information conditions for the baseline levels (IA1) vs. ; . After exposure to the positive or negative brand information in the article, participants did not display a shift in implicit attitudes (IA2), in contrast to the findings for explicit attitudes vs. , ns]. Finally, the time-delayed measure (IA3) was also found to be nonsignificant across the two conditions vs. . These results suggest that implicit (vs. explicit) attitudes are less likely to shift in response to single exposure to new information and thus replicate our findings from Study 1 and . 

Stimulus induction variations Paired-samples tests showed no statistical difference in either the negative vs. or the positive condition vs. ; . As in Study 1, these results support . 

Longitudinal induction variation Finally and as expected, no significant differences were found across the means at baseline (IA1) and post-time delay (IA3) for either condition; negative information vs. and positive information vs. in line with . 

Choice The results of the analyses on both implicit and explicit attitudes mirror the results from Study 1. This gives us confidence in analyzing and extending our findings in the domain of choice to investigate hypotheses 7 and 8 . Participants' actual choice was measured at the two time periods (TP1 and TP2), at the end of Stage 2 (CHOICE1) and at the end of Stage 3 (CHOICE2). Note that in this case, there was no measure of pre-exposure choice measures. The primary reason for this was that it would involve asking participants to make a choice twice within a interval and that was not realistically possible without introducing demand artifacts in the study protocol. However, as our focus is on studying the long-term impact of information exposure on choice, we are able to address this question adequately based on the design implemented. The results of the analyses are reported in Table 3 and Fig. 3c. 

Across condition variations Examining means across the two information conditions revealed no difference at either the post-information measure (Stage 2), CHOICE1 vs. or the time-delayed measure, CHOICE2 vs. . 

Longitudinal induction variation No significant differences were found across the means of CHOICE1 and CHOICE2 for either the negative vs. or the positive vs. information conditions. 

Table 3 Study 2: Differential induction based instantaneous and longitudinal changes in dual attitudes and actual choice 

Bold values indicate that the test is statistically significant 

** Implies that the test is significant at  

The conclusion that we can draw from our results is that just like implicit attitudes, a single exposure to a negative (or a positive) information does not affect the choice of the brand. 

The Relationship Between Attitudes and Choice The primary question that we aimed to answer through Study 2 was to assess the effect of implicit and explicit attitudes on brand choice, that is, hypothesis 4 . Study 1 found a significant effect of explicit as well as implicit attitudes on brand preference, but we need to understand if this preference actually converts to choice. In order to analyze the choice data obtained in Study 2, we performed a logistic regression analysis (as our dependent variable is binary). The use of a non-student sample allowed us to test the generalizability of the results across a more heterogeneous population sample. The analyses used variables such as gender, age, employment status, family size, education as covariates and nature of the information (positive or negative), as independent variables. The (a) recorded choice as well as (b) self-reported explicit attitudes, and (c) the implicit attitudes, collected across TP1 at Stage 2 and TP2 at Stage 3, were used to estimate the model. The results of the analyses can be seen in Table 2. Since none of the covariates was found to be significant, they were dropped from further analyses. The logistic regression provides us a concordance of and a Cox-Snell -square of 0.12 . 

The analyses have two variables that are of focal interest to us, (a) explicit attitudes and their impact on choice, which as per hypothesis 4 should be a nonsignificant relationship and (b) implicit attitudes and their impact on choice, which based on hypothesis 4 should be a significant relationship. Results show that explicit attitudes do not predict choice ( ). This is in line with what we proposed in . 

The effect of implicit attitudes, however, was found to be significant in predicting choice ( ), supporting H4. The odds ratio of 47.42 (the high value here is due to the range of the IAT scores varying between -1 and 1) shows that higher implicit attitudes result in greater choice of the Coca-Cola brand. Thus, we can conclude that actual choice is more influenced by implicit attitudes compared to explicit attitudes which might, in turn, explain the attitude-behavior gap especially, for products where ethics and social desirability might come into play. 

This study set out to better understand why the attitudebehavior gap in ethical consumption exists and how a business manager can go about attenuating this gap by focusing on the two different types of attitudes. Our arguments were underpinned by Wilson et al.'s (2000) model of dual attitudes which suggests implicit and explicit attitudes are impacted by information in a dissimilar way and both influence behavior in a unique way. Our findings are promising and have implications that have both theoretical and managerial relevance. 

To the best of our knowledge, the current research is the first to demonstrate the differential impact of implicit and explicit attitudes on brand preference as well as choice. Our research suggests that the nature of information does 

Study 2 - Changes in Dual Attitudes and Actual Choices. 

Fig. 3 Study 2-changes in dual attitudes and actual choices. (a) Variation in explicit attitudes, (b) variation in implicit attitudes, (c) variation in actual choice 

lead to a readjustment of explicit attitudes, as has been suggested in previous research (Folkes and Kamins 1999). We demonstrate that while explicit attitudes are more malleable in the short term, in the longer term they revert closer to their baseline levels. That is, the nature of information only affects explicit attitudes in the short but not in the longer term. However, this attitude shift does not lead to a shift in choice even in the short run, although it does 
affect self-reported brand preferences. Rather, implicit attitudes guide choice irrespective of the nature of the information presented or the updating of the explicit attitudes. That is, with a single exposure, positive explicit attitudes toward the target brand do not translate into greater brand choice and importantly, negative attitudes toward the target brand do not imply lower brand choice. 

Explicit attitudes seem to be susceptible to social desirability especially, in the context of ethical choices and thus an inaccurate predictor of choice. However, implicit attitudes are much more resilient and hence reliably, affect choice. Thus, the attitude-behavior gap might just be the effect of what one says based on what one is expected to say versus what one feels irrespective of what one is expected to. This is a significant contribution to business ethics literature as it suggests that researchers need to pay closer attention to the influence of implicit attitudes, especially for decisions involving ethics. These results also imply that individuals do indeed hold a dual attitude and that implicit attitudes guide behaviors to a greater extent compared to explicit attitudes. 

Further, the results suggest that previous literature has indeed overestimated the reliability of explicit measures of consumer attitudes and self-reported preferences, perhaps due to the social desirability bias that is prevalent in ethical consumerism (Bray et al. 2011; De Pelsmacker et al. 2005; King and Bruner 2000). Researchers need to be mindful of this bias and ensure they adopt measures that are less susceptible to it. 

Although consumers state that they care about ethics and want to reward companies that do the same (e.g., Cowe and Williams 2000; Creyer and Ross 1997; De Pelsmacker et al. 2005), it does not always translate to behavior, much to the disappointment of business managers focusing on ethical products. The issue is that the change in behavior is dependent on implicit attitudes in both the short term and long term. If business managers want to promote the ethicality of their business practices in the hope of increasing awareness and profits, they need to alter consumers' implicit attitudes. Business managers need to re-evaluate the way in which they construct marketing messages to ensure that implicit (and not just explicit) attitudes are being significantly impacted by their message. 

Our results suggest that implicit attitudes do change, slightly, after direct exposure to stimuli. However, these effects usually wear off in the longer term. Accordingly, business managers trying to promote ethical practices and products need to ensure they give consumers repeated exposure to their marketing message. It is possible that by doing so actual attitude shift will occur. 

Study 1 used a student sample, and thus, the generalizability of its findings was limited. However, we utilized a non-student sample in Study 2 to address this limitation. Further, while both brand preference and choice were measured in the current studies, it can be argued that participants did not actually spend their own money on the product in Study 2 and is hence, a limitation. Although prior research has implemented choice in a similar manner (Shiv and Fedorikhin 1999), we suggest that future research should try to replicate our findings with real monetary stakes. This would add to the robustness of the findings. 

The current setup considers a single exposure to positive/negative information. One would predict that the effects of repeated exposures should influence both implicit and explicit attitudes. Thus, examining the effect of multiple exposures to information regarding the brand could provide insights into two important research questions: (a) the interaction and the dependency between the two attitudes in the short as well as in long run and (b) the amount of social desirability bias that is attenuated (or enhanced) in the explicit attitudes held, over the long run. In this way, marketers will be able to better understand the impact of information on attitudes and behavior in the longer term and can design their communication strategies accordingly. 

It is noteworthy that not only explicit attitudes but also consumption can be susceptible to the social desirability bias. We address this concern in our research through the use of both implicit and explicit attitudes and their relationship to product choice where the former (implicit attitudes) is not subject to this bias and by getting at individual consumption behavior in isolation (participants were in private cubicles), as done extensively in consumption research (e.g., Garg et al. 2007). However, future research could try to replicate our findings in a more externally valid setting such as a field-experiment, to enable observation of consumption in situ and thus allay these concerns. 

Conflict of interest The authors Rahul Govind, Jatinder Jit Singh, Nitika Garg and Shachi D'Silva declare that they have no conflict of interests. 

Ethical Approval All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Declaration of Helsinki and its later amendments or comparable ethical standards.