Consumers as the Demand-Side Buffer of Sustainable Supply Chains: The Formation and Boundary Conditions of Resilient Consumption Through Perceived ESG Legitimacy

Highlights

Please indicate how your work links to systems science via your contributions to systems practice, theory, and/or methodology.

• Extends supply chain resilience theory by treating the consumer as an endogenous node of the system: premium-tolerant consumption acts as a demand-side buffer against sustainability cost shocks.
• Estimates mediation and both moderations within a single covariance-based structural model, testing signal credibility as a continuous latent moderator rather than through a median split.

What are the main findings and/or the implications of the main findings?

• Perceived ESG legitimacy converts ESG signal trust into premium-tolerant consumption; the income trade-off is statistically present but small, so demand-side resilience is not confined to affluent consumers.
• Signal credibility changes the configuration of the pathway, not only its strength, implying that credible third-party verification rather than firms’ own disclosure volume is where leverage lies.

Abstract

Sustainable supply chains carry a cost premium that can make their demand base fragile when economic shocks tighten household budgets, because the more expensive environmental, social, and governance (ESG)-aligned product is often among the first that consumers forgo. We propose extending resilience thinking to the downstream, demand-side node of the supply chain: we frame premium-tolerant ESG consumption as a conceptual indicator of demand-side resilience—an absorptive buffer that may help keep sustainable supply chains viable through disturbance—and draw on signaling theory to model how it is formed. Using a nationally structured, quota-controlled survey of 3000 Korean consumers collected by the Korea Consumer Agency, we estimate a moderated-mediation structural equation model in which ESG signal trust is associated with resilient sustainable consumption through perceived ESG legitimacy, conditioned by income and perceived greenwashing. Because the data are cross-sectional, we report associations rather than causal effects. Trust was positively associated with perceived legitimacy, which in turn was associated with resilient sustainable consumption, so that legitimacy is the proximal correlate linking trust to resilience. Income moderated this conversion only modestly—the sustainability–resilience trade-off is statistically present but small in magnitude—while perceived greenwashing did not attenuate it, and its unexpected positive coefficient proved unstable across specifications and is not interpreted. A multi-group comparison indicated that the pattern of associations differs with the perceived credibility of the ESG signal environment: when third-party signals were perceived as credible, the trust–legitimacy–consumption route was more pronounced, whereas when they were not, trust was associated with consumption more directly. Consumers also expected ESG far more of large firms than of micro-enterprises. No supply-chain-level outcome was measured; the cross-level link is advanced as a proposition for future work. The study reframes supply chain resilience as partly demand-side, potentially sustained by credible, independently verified ESG signals rather than by firms’ own disclosure.

1. Introduction

Contemporary supply chains are asked to be two things at once: sustainable and resilient. They must lower their environmental and social footprint while retaining the capacity to anticipate, absorb, and recover from disruption [1]. The past half-decade has made the second imperative impossible to ignore. The COVID-19 pandemic froze ports, emptied shelves, and exposed how a lean, globally dispersed network can amplify a single shock into a cascade of stock-outs [2]; the 2021 Suez Canal blockage halted an estimated tenth of world trade for days; and the 2022 energy-price spike and broad inflationary surge squeezed household budgets across advanced and emerging economies alike. Such disruptions also propagate downstream, rippling through a network in ways that a firm’s structural choices can either dampen or amplify [3]. Each episode underscored that efficiency without buffering capacity is fragility in disguise.

These same shocks, however, collide with the sustainability agenda in awkward ways. Sustainability investments—cleaner inputs, certified labor practices, transparent governance—typically raise unit costs [4,5], and cost structures that depend on a price premium are exactly the ones most exposed when demand contracts. When a firm that has pledged to cut its carbon footprint is forced to air-freight components to keep a plant running, it buys resilience at the direct expense of its emissions target—a textbook sustainability–resilience trade-off reported by supply chain managers in practice [6]. The relationship between sustainability and resilience is therefore better understood as a portfolio of synergies and trade-offs than as a single virtuous circle [7].

Most research on supply chain resilience concentrates on the supply side: dual sourcing, inventory buffers, supplier integration, and the dynamic capabilities of the focal firm [8,9,10]. Far less attention has been paid to the demand side. Yet a sustainable supply chain is viable only if there exists a stable base of buyers willing to absorb its price premium—especially during the very shocks that resilience is meant to survive. When household budgets tighten, the environmental, social, and governance (ESG)-aligned product, being more expensive, is often the first to be dropped. This is the behavioral counterpart of the well-documented attitude–behavior, or ‘green,’ gap: consumers who profess strong environmental values routinely fail to translate them into purchases when a price or convenience penalty intervenes [11]. We argue that the willingness of consumers to sustain ESG-oriented consumption under price pressure is itself a genuine, measurable form of resilience for the sustainable supply chain—a demand-side absorptive capacity that has been largely overlooked in the operations literature.

Adjacent consumer literatures come close to this idea without occupying it, and it is worth stating precisely where each stops. Willingness to pay concerns the price a consumer will accept under stable conditions; it describes a level rather than a persistence property, and is silent on whether that acceptance survives when budgets tighten [12]. Sustainable purchasing behavior explains why consumers choose sustainable options, but treats the choice context as given rather than as deteriorating [11,13]. Customer loyalty is directed at a brand relationship rather than at a production attribute that carries an unavoidable cost penalty and cannot be verified by the buyer [14]. Consumer resilience concerns the household’s own capacity to absorb and recover from shocks, not the persistence of demand for a particular class of goods [15]. What none of these literatures asks is whether the demand base for sustainably produced goods holds when relative prices bite—which is the question that determines whether a sustainably configured supply chain remains economically viable through a disturbance. Resilience thinking adds to sustainable-consumption research a disturbance-referenced framing in which the quantity of interest is persistence under stress rather than level of preference, which is why the moderators in our model are economic slack and signal credibility rather than attitude strength.

Because the evidence we bring to bear is Korean, it is also worth situating that setting against others at the outset. The institutional environment for ESG claims differs markedly across regions: in the European Union such claims are increasingly anchored in mandatory reporting and claim-substantiation rules, whereas in the United States ESG has become more market-driven and politically contested, and emerging markets typically combine weaker verification infrastructure with higher price sensitivity [16,17,18]. Willingness to accept a sustainability premium differs correspondingly, with survey evidence consistently placing Asia-Pacific and large emerging markets above North America and Western Europe [12,19]. The empirical literature on ESG and consumer behavior is itself concentrated in East Asia, alongside a heavy reliance on cross-sectional survey designs. Korea sits within this concentration as a high-income market with a maturing but, at the time of data collection, not yet mandatory ESG disclosure regime—a configuration that makes it an informative case for the credibility mechanism we examine, and one whose boundaries we return to in Section 6.4.

This study makes the demand-side argument concrete and testable using a large, demographically structured survey of Korean consumers (n = 3000) collected by the Korea Consumer Agency [20]. We operationalize resilient sustainable consumption—which we treat as a consumer-level indicator of the broader conceptual property of demand-side resilience rather than an equivalent of it—with behavioral intention items that explicitly condition consumption on a price penalty (the intention to buy eco-friendly or ethically produced goods ‘even if somewhat more expensive’). We then estimate a single integrated model of moderated mediation grounded in signaling theory. At its core, a mediation chain traces how ESG signal trust is associated with resilient sustainable consumption through perceived ESG legitimacy; onto this chain we layer two moderators that bound the conversion. The first, income, captures the central sustainability–resilience trade-off—whether demand-side resilience is contingent on economic slack. The second, perceived greenwashing, captures whether information asymmetry attenuates the conversion. Finally, a multi-group comparison asks whether the pattern of associations differs with the perceived credibility of the surrounding ESG signal environment. Because the data are cross-sectional, we report associations rather than causal effects throughout, and we treat the link from individual intentions to supply-chain-level resilience as a proposition rather than a demonstrated outcome.

The contributions are of four kinds. Conceptually, we propose a demand-side conceptualization of supply chain resilience that complements the dominant supply-side view, and offer resilient sustainable consumption as its consumer-level indicator rather than as an equivalent of it. Theoretically, we bring signaling theory to bear on the underlying pattern, showing that the perceived credibility of the ESG signal is associated not only with the level of consumer-level resilient consumption but with the configuration of the pathway associated with it. Methodologically, we estimate mediation and moderation within a single covariance-based structural model, test signal credibility as a continuous latent moderator rather than through a median split, and report common latent factor and marker-variable diagnostics alongside effect sizes. Practically, we adjudicate empirically between the trade-off and synergy framings, showing the trade-off to be statistically present but small in magnitude, and we refine the information-asymmetry account of demand: a single perception of greenwashing is not associated with erosion of the conversion of ESG orientation into resilient sustainable consumption, although the two-item greenwashing measure is weak and this coefficient is not interpreted. The remainder of the paper reviews the relevant literature (Section 2), develops the theory and hypotheses (Section 3), describes the data and methods (Section 4), reports the results (Section 5), and discusses implications and limitations (Section 6).

2. Literature Review

2.1. Supply Chain Resilience and the Neglected Demand Side

Supply chain resilience (SCR) became a central construct in operations and logistics research with Christopher and Peck’s foundational treatment, which defined resilience as the capacity of a network to return to a stable state after disruption and identified re-engineering, collaboration, agility, and a risk-management culture as its structural pillars [1]. Sheffi and Rice extended this view by characterizing the resilient enterprise as one that invests in redundancy and flexibility to compress the disruption profile and recover quickly [9]. Two decades of subsequent work have crystallized these ideas into a vocabulary of capabilities—robustness, agility, flexibility, velocity, and visibility—and into a battery of quantitative assessment methods cataloged in systematic reviews [8,10]. The empirical referents of this literature are overwhelmingly upstream events: the COVID-19 pandemic that idled factories and froze logistics networks [2], and disruptions whose effects propagate downstream as a ‘ripple effect’ that simulation studies show can be amplified or dampened by a firm’s structural choices [3]. The supply-side account has itself grown more sophisticated: recent work shows that horizontal and vertical inter-organizational collaboration is an established route to network resilience [21], and that green-finance and circular-economy instruments shape the conditions under which sustainable configurations remain financially viable [22]. These mechanisms sharpen rather than soften the contrast we draw, because they all locate resilience in the firm and its partners, leaving the demand base outside the account.

It is worth stating the boundary between this demand-side notion and supply-side resilience explicitly, since the two are complementary rather than competing. Supply-side resilience concerns the capacity of a firm or a network to source, buffer, and recover, and its locus is the organization; demand-side resilience concerns the persistence of the demand base that makes a sustainably configured supply chain economically survivable when relative prices rise, and its locus is the consumer [23]. A supply chain may be highly resilient in the conventional sense and still fail if demand for its more expensive output collapses under economic strain. What we measure here is an intention-level indicator of that demand-side property rather than the property itself, a distinction we maintain throughout.

The supply-side account has also grown more sophisticated in one respect that sharpens the contrast. Beyond the capabilities of the focal firm, a substantial literature establishes inter-organizational collaboration—horizontal and vertical cooperation, joint capacity pooling, information sharing, and coordinated risk management—as an established route to resilience [21,24,25]. That this collaborative machinery has been elaborated among firms while the demand base has remained outside the account is precisely the asymmetry this study addresses.

What unites this body of work is an upstream, firm-centric gaze. Resilience is treated as something a focal firm or its suppliers build through sourcing strategy, capacity buffers, and inter-organizational coordination [8,10], and disruption is modeled as a supply-side or logistical shock to be absorbed by those buffers [2]. The demand side enters, if at all, only as exogenous volatility. Yet the viability of any supply chain ultimately depends on whether buyers continue to buy. This dependence is especially acute for a sustainable supply chain, whose offerings typically carry a price premium: here the binding disruption may be a demand-side one—an income shock or inflationary squeeze that erodes willingness to pay—rather than a plant fire or a port closure. Reviews of supply chains under multiple, overlapping crises make this explicit, identifying mismatched sustainability expectations and the resulting stakeholder pressure as a first-order tension alongside conventional supply-side disruptions [26]. We therefore propose treating persistent consumer demand for ESG-aligned products under price pressure as a demand-side analog of an inventory or capacity buffer: an absorptive capacity located not in the firm but in its customer base. This reframing extends SCR theory downstream and connects it to the consumer-behavior literature reviewed in Section 2.3.

2.2. The Sustainability–Resilience Nexus: Trade-Offs and Synergies

Sustainable supply chain management (SSCM) developed along a track parallel to SCR, organized around the triple bottom line and the integration of environmental, social, and economic goals into inter-organizational processes [4,5]. Only recently have scholars asked systematically how sustainability and resilience relate to one another. The most comprehensive synthesis, Negri et al.’s systematic review, concludes that the intersection of the two literatures is still nascent and conceptually muddled, and that the field urgently needs to clarify which practices jointly improve—or trade off—sustainability and resilience [7]. Their follow-up framework, tested on firms with at least moderately mature sustainability practices, found that synergies—local sourcing, supplier development, transparency, and reporting—tend to overshadow trade-offs [27]. Simulation evidence sharpens the picture by showing that the sign of the relationship depends on the specific practice: some sustainability measures mitigate the downstream ripple effect, whereas others amplify it [3].

The relationship is, however, contingent and perceptual rather than fixed. Interviewing supply chain managers across Europe and the Middle East, Cotta et al. found that some regarded resilience and sustainability as fundamentally conflictive, others as synergistic, and a notable minority saw no relationship at all [6]. Part of the difficulty is theoretical: as Pagell and Shevchenko argue, SSCM research has tended to focus on the synergistic and the familiar while systematically overlooking trade-offs and the radical innovation that genuinely sustainable chains may require [28]. The empirical evidence is correspondingly mixed—ESG-strong firms proved more resilient during the COVID-19 market crash in some settings [29], yet sustainability commitments can also constrain the response menu during a disruption, as when carbon pledges rule out expedited but high-emission logistics [6]. Crisis-era reviews catalog this ambivalence as a set of recurring tensions—between sustainability and efficiency, between sustainability and resilience, and between global and local operations—that managers must hold in balance rather than resolve [26].

Two gaps motivate the present study. First, this literature is almost entirely supply- and firm-side: it asks whether producing sustainably trades off against being resilient, but not whether consuming sustainably is itself a source of resilience. Second, it rarely treats the trade-off as a matter of degree to be estimated rather than a binary to be asserted. By measuring whether ESG-oriented consumption holds up under an explicit price penalty, and whether that resilience is conditional on consumers’ economic slack, we move the trade-off/synergy debate from the producer to the consumer and from a qualitative to a quantitative footing—directly answering the call to estimate the magnitude of these tensions [7,27].

2.3. ESG, Consumer Trust, and the Attitude–Behavior Gap

On the consumer side, the dominant theoretical scaffold is the theory of planned behavior, in which attitudes shape intentions and intentions shape behavior [30]. Applied to green and ESG consumption, a large body of work documents a robust attitude-to-intention link—meta-analytic evidence confirms that favorable attitudes reliably raise green purchase intention [31]—but a far more fragile intention-to-behavior link. Narrative reviews show that the attitude–intention–behavior gap persists precisely because situational frictions, chief among them price, sever the final step from intention to purchase [11,32]. The gap matters economically because corporate sustainability can pay: firms with stronger sustainability practices exhibit better long-run processes and performance [33], a demand-side return that the attitude–behavior gap threatens to undercut whenever a price premium intervenes.

Why the gap should open at the moment of price is best understood through signaling theory. Because consumers cannot directly audit a firm’s environmental or labor practices, they face an information asymmetry of exactly the kind Spence first formalized: the relevant quality is unobservable, and the buyer must infer it from observable, costly signals [34]. In the canonical account, a signaler with private information about its true quality emits a signal that a receiver observes and interprets within a wider signaling environment; the signal is informative only when it is more costly for low-quality than for high-quality senders to send, so that signal credibility, not signal volume, carries the information [35,36]. ESG claims, certifications, and disclosures are precisely such signals, and consumer trust is the receiver-side response that substitutes for direct verification. This lens also explains how the information environment can fail: when ESG ratings from major agencies diverge, the resulting ‘aggregate confusion’ injects noise that even motivated receivers struggle to resolve, loosening the link between a firm’s true sustainability and the signal consumers perceive [37]. It further implies a distinction we exploit throughout—between the credibility of a signal (the perceived trustworthiness of third-party, independently verifiable ESG information) and the quality of a firm’s own disclosure (the perceived informativeness, transparency, consistency, and factualness of firm- and public-sector ESG reporting)—with signaling theory predicting that the credibility of independent verification, not the quality of self-disclosure, is what moves behavior [36].

A growing body of consumer research operationalizes this logic. Studies grounded explicitly in signaling theory show that perceived corporate ESG practices raise brand trust, which in turn drives sustainable purchase intention: brand trust mediates the path from ESG to green-food purchase intention [38], and perceived ESG dimensions shape brand attitude and trust en route to low-carbon purchase intention [39]. Experimental evidence treats ESG communication as a market signal whose persuasive force runs through perceived ethicality and authenticity, with all three of the environmental, social, and governance dimensions improving behavioral outcomes such as purchase intention and willingness to buy [40]. Consistent with a stakeholder reading of the same phenomenon, ESG practices have been shown to strengthen customer trust, loyalty, and brand-directed behavior [41], and ESG disclosure functions as a credibility-dependent signal whose value to outside observers is conditioned by governance quality [42]. Across these studies, trust and a value- or legitimacy-type judgment recur as the sequential mechanism converting ESG signals into behavior—the mechanism we formalize in Section 3.

The same signaling logic identifies a second friction: perceived greenwashing, the suspicion that ESG claims are public-relations exercises decoupled from substantive action. Greenwashing perception has been shown to erode green trust through heightened consumer confusion and perceived risk [43], to depress green purchase intention both directly and indirectly via diminished green word-of-mouth [44], and to operate through a perceived betrayal of the implicit consumer–firm contract, an effect amplified among consumers with a strong sense of environmental responsibility [45]. Because such skepticism corrodes the very trust on which premium-tolerant demand depends, and because the drivers of greenwashing are structural rather than incidental [46], whether greenwashing perception actually weakens the conversion of ESG orientation into resilient consumption is an empirical question we take up directly.

Finally, this consumer literature has remained largely disconnected from the supply chain resilience literature reviewed above. Recent contributions have begun to bridge consumer- and network-level constructs—linking cultural congruence to market responsiveness as a resilience outcome [47] and partner compatibility and collaboration to risk-response capability [48]—but the consumer as a buffer for sustainable supply chains specifically remains untheorized. The present study addresses that gap by positioning premium-tolerant ESG consumption as the demand-side absorptive capacity of the sustainable supply chain, and by tracing and bounding the signaling pathway that produces it.

3. Theoretical Background and Hypotheses

Building on the demand-side reframing developed in Section 2, we treat premium-tolerant ESG consumption as an absorptive capacity—the resilience dimension most relevant to demand—and now specify how it is produced and bounded. Whether this buffer holds depends, in signaling terms, on whether the surrounding signal environment lets consumers read a firm’s sustainability as credible, a condition we build into the model below [36].

Signaling theory is not the only lens available for sustainable-consumption behavior, and we adopt it deliberately rather than by default. Value–belief–norm theory explains pro-environmental behavior through internalized norms activated by awareness of consequences [49] and accounts well for why consumers hold sustainable preferences, but is largely silent on whether those preferences survive a price penalty attached to an attribute the consumer cannot verify. The theory of planned behavior models intention through attitude, subjective norms, and perceived behavioral control, and predicts sustainable purchase intention in general, but treats the informational status of the sustainability attribute as given rather than as the object of inference [30]. Protection motivation theory is organized around threat appraisal and coping, which fits responses to personal risk better than responses to a firm’s production claims [50]. The defining feature of our problem, set out in Section 2.3, is information asymmetry—precisely the problem signaling theory was built to address—and the reason signal credibility, rather than attitude strength or norm activation, is what conditions the pathway here. We therefore treat these theories as complementary rather than competing, and note in Section 6.4 that integrating signaling theory with the theory of planned behavior or value–belief–norm theory is a promising extension.

3.1. From ESG Attitude to Resilient Consumption: A Mediated Pathway

Signaling theory supplies the mechanism: a firm’s ESG practices function as costly signals, and the receiver’s response is theorized as trust in the sender rather than an appraisal of any single message [34,35,36]; relationship-marketing theory likewise treats such trust as the relational mechanism through which trustworthy conduct converts into favorable response, and consumer research establishes trust in ESG- and corporate social responsibility (CSR)-active firms as a separately measured antecedent of consumer reaction [51,52,53]. Consumers accordingly extend ESG signal trust—a verification-substituting trust in firms that practice ESG—in response to the signals those firms emit. We argue that this trust translates into premium-tolerant behavior not so much directly as through an intervening judgment of perceived ESG legitimacy: the consumer’s assessment that an ESG firm’s offering is socially, ethically, and economically warranted, which is what makes its price premium acceptable. Trust raises legitimacy, and legitimacy, in turn, licenses resilient consumption—a sequence consistent with evidence that brand trust and value-type judgments mediate the path from perceived ESG to sustainable purchase intention [38,40]. Because trust is unlikely to be fully absorbed by a single mediator, we expect a partial mediation in which signal trust also retains a direct effect on resilient consumption [39].

H1. 

ESG signal trust will positively affect perceived ESG legitimacy.

A legitimacy judgment is the evaluative step that a price premium requires: what licenses paying more is not that the consumer likes the firm but that the consumer judges the more expensive offering to be socially, ethically, and economically warranted. Consumer-value research shows that such judgments predict premium-tolerant choice more directly than affect does.

H2. 

Perceived ESG legitimacy will positively affect resilient sustainable consumption.

The same logic allows the receiver to act on trust without completing the evaluative step, particularly where the signal environment supplies little to evaluate. Brand-level trust is a well-validated predictor of purchase and loyalty outcomes in its own right, so we expect a direct association alongside the mediated one [14].

H3. 

ESG signal trust will positively affect resilient sustainable consumption (direct effect).

Taken together, H1 and H2 imply that the association between ESG signal trust and resilient sustainable consumption operates partly through an intervening evaluative judgment rather than wholly as a direct link. We therefore treat perceived ESG legitimacy as a mediator of that association.

H4. 

Perceived ESG legitimacy will mediate the relationship between ESG signal trust and resilient sustainable consumption.

Signaling theory further implies that not all ESG information is equal. A signal shapes behavior only insofar as it is credible—costly to fake and therefore trustworthy—rather than merely abundant [36]. We therefore distinguish the credibility of the ESG signal, captured by the perceived trustworthiness of third-party ratings, from the quality of a firm’s own disclosure, captured by the perceived informativeness, transparency, consistency, and factualness of firm- and public-sector reporting, and expect the former to feed resilient consumption while the latter does not [37]. Beyond a main effect, this credibility precondition suggests that the pathway itself may be reorganized by the trustworthiness of the signal environment: where third-party signals are credible, consumers can afford to reason from trust to a legitimacy judgment and then to behavior, whereas where signals lack credibility, that legitimacy ladder may collapse and trust may drive behavior more directly. We examine this possibility as a research question through a multi-group comparison across high- and low-credibility signal environments (Section 5.4), rather than as a directional hypothesis, because the theory predicts a restructuring of the pathway rather than simply its strengthening or weakening [36].

3.2. The Sustainability–Resilience Trade-Off: Economic Vulnerability as Moderator

If demand-side resilience were costless, every consumer with a positive ESG attitude would absorb the premium; the persistent attitude–behavior gap shows that it is not [11]. The trade-off view of the sustainability–resilience nexus predicts that the conversion of ESG legitimacy into premium-tolerant consumption is constrained by economic slack, so that lower-income consumers—who face a sharper trade-off between sustainable choice and household budget resilience—convert legitimacy into resilient consumption less readily [6,7]. The synergy view predicts instead that ESG legitimacy drives resilient consumption broadly, largely independent of income, mirroring, on the demand side, the producer-side finding that synergies tend to overshadow trade-offs once sustainability practices are mature [27]. We test these as competing expectations by treating income as a moderator of the legitimacy → resilience conversion (throughout, the arrow denotes a hypothesized directional path from the first construct to the second).

H5. 

Income will moderate the conversion of perceived ESG legitimacy into resilient consumption, such that the indirect effect linking ESG signal trust to resilient consumption through perceived legitimacy will be stronger at higher income (a sustainability–resilience trade-off).

A second moderator follows from the signaling account of information asymmetry. If perceived greenwashing erodes the trust on which the legitimacy judgment rests, it should weaken the conversion of ESG orientation into resilient consumption, just as greenwashing perception depresses green trust and purchase intention in prior work [43,44]. We therefore test whether perceived greenwashing attenuates the legitimacy → resilience pathway—an information-asymmetry erosion of demand-side resilience [45].

H6. 

Perceived greenwashing will moderate the conversion of perceived ESG legitimacy into resilient consumption, weakening the indirect effect linking ESG signal trust to resilient consumption through perceived legitimacy (an information-asymmetry erosion of demand-side resilience).

4. Materials and Methods

4.1. Data and Sample

We analyze the consumer survey that underlies the Korea Consumer Agency’s policy study on consumer perceptions of corporate ESG management [20]. The survey was commissioned by the Korea Consumer Agency—a public consumer-protection institution under the Korean government—and fielded as a nationally structured, quota-controlled survey of 3000 adults aged 20–69, with quotas on sex, age, and region set to national population proportions. Consistent with that design, the realized sample is balanced by sex (1500 men, 1500 women) and spans all 17 first-level administrative regions; respondents in the Seoul Capital Area make up 51.7% of the sample, closely tracking that region’s 50.5% share of the national population in 2022 (Statistics Korea). Educational attainment is predominantly tertiary (65.9% hold a four-year university degree), and 61.8% reported prior awareness of the term ‘ESG management.’ Because monthly income is the central moderator in our model, we characterize it by its full distribution rather than a single statistic: the median is 3.0 million Korean won (KRW) and the mean KRW 3.24 million (standard deviation, SD = 2.85 million), and responses span every bracket from under KRW 2 million (23.6% of respondents) to KRW 8 million or more (4.3%). This wide, population-anchored dispersion—rather than a sample concentrated in one income band—is what allows the income-moderation hypothesis to be identified.

Table 1. Sample demographic characteristics.

Variable	Category	n	%
Sex	Male	1500	50.0
	Female	1500	50.0
Age (years)	20–29	522	17.4
	30–39	535	17.8
	40–49	654	21.8
	50–59	696	23.2
	60–69	593	19.8
Education	High school or less	514	17.1
	In college	139	4.6
	College graduate	1978	65.9
	In/Completed graduate	369	12.3
Monthly income	<2 M KRW	708	23.6
	2–3 M	741	24.7
	3–4 M	643	21.4
	4–5 M	327	10.9
	5–6 M	262	8.7
	6–7 M	119	4.0
	7–8 M	70	2.3
	≥8 M	130	4.3
Occupation	Public sector	353	11.8
	Self-employed/owner	350	11.7
	Sales/service	740	24.7
	Homemaker	371	12.4
	Student	158	5.3
	Agriculture/fishery	24	0.8
	Other (incl. unemployed)	1004	33.5
Region	Seoul Capital Area	1552	51.7
	Yeongnam	731	24.4
	Chungcheong	316	10.5
	Honam	277	9.2
	Gangwon	87	2.9
	Jeju	37	1.2

Note. n = 3000. Quota variables: sex, age, and region. KRW = Korean won.

summarizes the sample’s full demographic composition.

The sampling procedure was as follows. Respondents were adults aged 20 and over resident in Korea, selected by modified proportional quota allocation on region, age, and sex, with region defined as the 17 first-level administrative divisions and quotas set to national population proportions. The survey was administered by a professional research firm as a web-based online survey between 10 and 22 November 2022, and the source report states a sampling error of ±2.90 percentage points at the 95% confidence level. The instrument was not pre-tested on a separate quantitative pilot sample; it was developed and validated through expert consultation, in which academic, government and in-house ESG specialists reviewed the structured questionnaire for appropriateness and validity before fielding. We note the absence of a quantitative pilot as a limitation in Section 6.4. The full instrument for every item used in the analysis is provided in Appendix A.

This dataset is well suited to the present study for several reasons, though it also has characteristics we treat as boundary conditions rather than as guarantees of representativeness. First, it derives from a survey administered for a public body, the Korea Consumer Agency, whose quota-controlled sampling aligns the sample with national adult population proportions on sex, age, and region; the correspondence is close on these quota variables—an even sex split and a Seoul-Capital-Area share (51.7%) within roughly one percentage point of the population figure (50.5%) [54]. We therefore describe the sample as quota-structured and demographically aligned with national benchmarks on these dimensions, rather than as nationally representative, because a quota design does not ensure representativeness with respect to non-quota characteristics. Indeed, outside the quota frame, the sample departs from the population: tertiary attainment (65.9%) substantially exceeds the national adult figure, and self-reported ESG awareness is high (61.8%). Because both education and prior ESG awareness plausibly correlate with the constructs we model, and because an online-administered instrument underrepresents low-digital-access consumers, our estimates may not extend to less educated or less ESG-aware segments—a limitation we return to in Section 6.4. Second, the sample is demographically broad—evenly split by sex, spanning all adult cohorts from the twenties to the sixties, and covering all 17 first-level administrative regions—so the patterns we estimate are not confined to a narrow subgroup. Third, the sample spans the full income distribution, from the lowest bracket (under KRW 2 million) to the highest (KRW 8 million or more); because a central question is whether income moderates the conversion of ESG legitimacy into resilient sustainable consumption, this dispersion is essential, as a sample concentrated in a single band could not identify the trade-off. Fourth, the instrument was purpose-built to measure consumer perceptions of corporate ESG management, so its items map onto the trust, legitimacy, and sustainable-consumption constructs the theory requires; a pre-test was conducted prior to full fielding and item wording was refined accordingly. Finally, the sample size (n = 3000) provides ample power for covariance-based structural equation modeling and supports the multi-group analysis, leaving well over a thousand respondents in each credibility subgroup.

4.2. Measures

Latent constructs were formed from the survey’s multi-item blocks and named to reflect their role in the signaling account. All items were originally administered in Korean and were translated into English for presentation in

Table 2. Measurement items.

Construct	Item	Measurement Item (Translated)	Source
ESG signal trust (SIGTRUST)	ST1	ESG-practicing firms are socially respectable.	[12,15]
	ST2	ESG-practicing firms are trustworthy.
	ST3	I feel positive toward ESG-practicing firms.
Perceived ESG legitimacy (ESGLEGIT)	EL1	Using products of ESG-leading firms gives a good impression to others (social value).	[15,37,38]
	EL2	Buying from ESG-leading firms is the right thing to do for the community (ethical value).
	EL3	Products of ESG-leading firms offer good quality for the price paid (economic value).
Resilient sustainable consumption (RESIL)	RS1	I will buy eco-friendly products even if they cost somewhat more.	[15,39]
	RS2	I will buy products from socially responsible firms.
	RS3	I will buy from ethical firms even if they cost somewhat more.
	RS4	I will buy products from firms that comply with laws and principles.

Note. Items adapted from [20]; source also cites each construct’s scale literature.

; every item used a five-point Likert response scale. ESG signal trust (SIGTRUST)—the verification-substituting trust consumers extend to ESG-practicing firms when they cannot directly audit those firms’ practices—combined respect, trustworthiness, and positivity (ST1–ST3), in line with conceptualizations of ESG-based brand trust as a signaling response to information asymmetry [37]. Perceived ESG legitimacy (ESGLEGIT)—the consumer’s judgment that an ESG firm’s offering is socially, ethically, and economically warranted, which is what justifies tolerating its price premium—combined the social, ethical, and economic value items (EL1–EL3); this three-dimensional operationalization is consistent with multidimensional perceived-value scales [55] and with the social–environmental–economic structure of consumer-perceived CSR value [56]. Resilient sustainable consumption (RESIL) was modeled over four price-tolerant consumption items spanning environmental, social, and ethical domains (RS1–RS4); several items explicitly condition the behavior on accepting a higher price, operationalizing resilience under price shock; conditioning sustainable purchase intention on a willingness to bear a price premium follows established green-consumption-value research [57]. Two signal antecedents were used in the conditional-process model: ESG signal credibility (SIGCRED), the perceived trustworthiness of third-party ESG ratings, and ESG signal quality (SIGQUAL), the perceived informativeness, transparency, consistency, and factualness of firm- and public-sector ESG disclosure (eight items). Perceived greenwashing (GW) combined the two publicity-focused information items for firms and the public sector. To keep the conceptual property distinct from what we measured, we treat demand-side resilience as a theoretical property of the demand base and name the measured dependent variable resilient sustainable consumption; resilient sustainable consumption (RESIL) is its label in the model, and the two are not used interchangeably. Consistent with a reviewer’s observation, ESG signal trust as measured is a receiver-side trust in ESG-practicing firms rather than an appraisal of a specific signal instrument, and perceived ESG legitimacy includes a pragmatic (value-for-money) element alongside moral and cognitive ones, following Suchman’s tripartite conception; robustness estimations reported in Section 5 confirm that excluding the value-for-money indicator, or restricting the dependent variable to the two explicitly price-conditioned items, leaves the structural conclusions unchanged. Monthly household income was recorded in eight ordered brackets (from under KRW 2 million to KRW 8 million or more); we assign bracket midpoints and, in Section 5.3, verify the interval assumption against a continuous income item and alternative codings. Perceived greenwashing rests on only two items, one referring to corporate and one to public-sector ESG communication (inter-item correlation 0.543, Spearman–Brown reliability 0.704), so we interpret it cautiously and reproduce both items in Table 2.

4.3. Analytic Strategy

We estimated a single integrated model in which both mediation and moderation are handled entirely within a covariance-based structural equation model (CB-SEM) with maximum-likelihood estimation. Following Anderson and Gerbing’s recommended two-step approach [58], the measurement model was first assessed via confirmatory factor analysis (CFA), evaluating internal consistency (Cronbach’s α and composite reliability, CR), convergent validity (average variance extracted, AVE), and discriminant validity by both the Fornell–Larcker criterion and the heterotrait–monotrait (HTMT) ratio [59,60]. The mediation chain—ESG signal trust → perceived ESG legitimacy → resilient consumption, with a direct signal-trust → resilience path—was then estimated as a latent structural model. Moderation was incorporated within the same model rather than in a separate regression: latent interaction terms (perceived legitimacy × income and perceived legitimacy × perceived greenwashing) were formed with the product-indicator method using double-mean-centered indicators [61] and were entered on the legitimacy → resilience path. The index of moderated mediation and the conditional indirect effects (at −1 SD, the mean, and +1 SD of each moderator) were defined as functions of the model parameters and evaluated with 5000 bias-corrected bootstrap resamples; direct structural paths are reported with z-statistics, whereas indirect effects and the indices of moderated mediation are reported with 95% bootstrap confidence intervals, an interval excluding zero indicating significance. To test whether signal credibility reorganizes the pathway, we additionally fit a two-group model (high vs. low ESG signal credibility, median split) and established configural, metric, and scalar measurement invariance before comparing structural paths with Wald tests. Continuous variables were standardized before forming the product indicators. Common method bias was assessed with three diagnostics: Harman’s single-factor test, a comparison of the single-factor and measurement-model CFA fits, and full-collinearity VIFs [62]. Because all measures are self-reported, we supplemented these procedural diagnostics with two stronger statistical remedies—a common latent factor on which all focal indicators load, in addition to their substantive factors, and a marker-variable analysis using a theoretically unrelated item in the same response format, adjusted following Lindell and Whitney [63]. Signal credibility, which the multi-group model splits at the median, is additionally specified as a continuous latent moderator using the same product-indicator approach, with the multi-group model retained as a supplementary robustness check. Finally, because cross-sectional data cannot establish temporal order, we estimated the fully reversed model and alternative structural specifications to probe reverse causality; results are reported in Section 5.4 and Section 6.4.

Because all constructs were measured within a single self-report instrument, we addressed common method variance both procedurally and statistically. Procedurally, the source instrument separated the construct blocks across different question sections, guaranteed respondent anonymity, used mixed item formats, and was administered by a professional research firm under Korea Consumer Agency quality control. Statistically, beyond Harman’s single-factor test and the single-factor confirmatory comparison, we estimated an unmeasured common latent factor (CLF) on which all focal indicators load, in addition to their substantive factors, and which is constrained orthogonal to those factors, and we re-estimated the structural model with the CLF retained. We further conducted a marker-variable assessment following the measured latent marker variable logic, using a theoretically unrelated item measured on the same five-point self-report format as the marker.

To assess the sensitivity of the specification to the assumed direction of association, we additionally estimated three alternatives within the same measurement model: a fully reversed structure, a structure treating perceived ESG legitimacy and resilient sustainable consumption as correlated outcomes of ESG signal trust, and a structure placing perceived legitimacy prior to signal trust. Because the just-identified structural models among three latent variables are algebraically equivalent, these alternatives cannot be discriminated by fit; we report them for transparency and discuss the grounds for the specified ordering in Section 6.4. ESG signal credibility, previously examined only through a median split, is additionally modeled as a continuous latent moderator using the same product-indicator approach applied to income and perceived greenwashing.

5. Results

5.1. Measurement Model

We first established that the measurement model is reliable and valid across all six latent variables and free of serious common method bias.

Table 3. Confirmatory factor analysis (all six latent constructs).

Construct/Item	Mean	SD	Loading	α	CR	AVE
ESG signal trust (SIGTRUST)				0.893	0.893	0.736
ST1	4.04	0.71	0.844 ***
ST2	3.99	0.74	0.869 ***
ST3	4.11	0.68	0.861 ***
Perceived ESG legitimacy (ESGLEGIT)				0.860	0.860	0.672
EL1	3.61	0.82	0.789 ***
EL2	3.71	0.80	0.843 ***
EL3	3.49	0.82	0.827 ***
Resilient sustainable consumption (RESIL)				0.871	0.873	0.633
RS1	3.49	0.82	0.782 ***
RS2	3.72	0.76	0.792 ***
RS3	3.57	0.81	0.852 ***
RS4	3.84	0.72	0.754 ***
ESG signal credibility (SIGCRED)			0.870–0.900 ***	0.914	0.914	0.781
ESG signal quality (SIGQUAL)			0.726–0.843 ***	0.930	0.931	0.627
Perceived greenwashing (GW)			0.726–0.748 ***	0.703	0.704	0.543

Note. Standardized coefficients; n = 3000. Bootstrap CIs from 5000 resamples. *** p < 0.001.

,

Table 4. Discriminant validity.

Construct	SIGTRUST	ESGLEGIT	RESIL	SIGCRED	SIGQUAL	GW
SIGTRUST	0.858	0.495	0.612	0.360	0.157	0.302
ESGLEGIT	0.499	0.820	0.683	0.810	0.469	0.188
RESIL	0.601	0.677	0.796	0.533	0.308	0.273
SIGCRED	0.361	0.807	0.526	0.884	0.551	0.117
SIGQUAL	0.157	0.466	0.308	0.548	0.792	0.253
GW	0.299	0.191	0.263	0.116	0.258	0.737

Note. Diagonal (bold) = √AVE; lower triangle = correlations; upper triangle = HTMT.

and

Table 5. Common method variance diagnostics.

Diagnostic	Result	Criterion
Harman’s single-factor test	First factor = 52.96% of variance	<50% preferred (weakest diagnostic)
Single-factor CFA fit	CFI = 0.700, RMSEA = 0.228	should fit poorly
Full collinearity VIF (max)	1.80	<3.3
Common latent factor (CLF)	Method variance = 14.7%; all paths remain significant	<25%
Marker variable (unrelated item)	Smallest marker r = 0.133; adjusted correlations remain significant	—

report, in turn, the confirmatory factor analysis, discriminant validity, and common method variance diagnostics, each discussed below.

We first estimated a measurement model containing all six latent variables simultaneously, so that the signal-environment constructs are evaluated on the same terms as the focal constructs. The model fits the data acceptably, χ²(215) = 2512.18, comparative fit index (CFI) = 0.951, Tucker–Lewis index (TLI) = 0.942, root mean square error of approximation (RMSEA) = 0.060. All standardized loadings are significant at p < 0.001 and range from 0.726 to 0.900; Cronbach’s α ranges from 0.703 to 0.930, composite reliability from 0.704 to 0.931, and average variance extracted exceeds 0.50 for every construct, ranging from 0.543 to 0.781 (Table 3). Because perceived ESG legitimacy and ESG signal credibility are the closest pair in the discriminant validity matrix, we tested them additionally by constraining them into a single factor; the merged model fits substantially worse (Δχ²(5) = 1410.44, p < 0.001, ΔCFI = −0.030), supporting their empirical distinctness. Equivalent tests for the other adjacent pairs likewise favor the six-factor solution.

Because two of the four outcome indicators impose an explicit price penalty while two do not, we re-estimated the model using only the two price-conditioned items. The narrower operationalization is internally consistent (α = 0.841, inter-item r = 0.726) and leaves the substantive conclusions unchanged: the association between signal trust and perceived legitimacy is identical at 0.434, legitimacy to resilient consumption is 0.414 against 0.432, the direct association is 0.253 against 0.318, and the indirect association is 0.180 against 0.188, with the income moderation still significant (0.057, p < 0.001). One difference is worth reporting: under the price-conditioned outcome, perceived ESG signal quality becomes significant (0.065, p < 0.001) whereas it was not with the full measure, and signal credibility weakens somewhat (0.069, p = 0.002); we return to this in Section 6.2.

The confirmatory factor analysis in Table 3 confirms a sound measurement model. Estimated across all six latent variables, the model fits the data acceptably by conventional standards [64] (χ²(215) = 2512.18, CFI = 0.951, TLI = 0.942, RMSEA = 0.060); the three-construct model that underlies the structural estimates fits well on the same criteria (χ²(32) = 442.2, CFI = 0.977, TLI = 0.968, RMSEA = 0.065, standardized root mean square residual (SRMR) = 0.031). All standardized loadings were significant and ranged from 0.726 to 0.900, above the 0.70 benchmark for indicator reliability [65]. Internal consistency was adequate for every construct, with Cronbach’s α from 0.703 to 0.930 and composite reliability from 0.704 to 0.931, and average variance extracted exceeded 0.50 throughout, ranging from 0.543 to 0.781. Two constructs warrant comment: perceived greenwashing rests on two items, for which the appropriate reliability coefficient is Spearman–Brown (0.704), and ESG signal quality is empirically two-dimensional, separating corporate from public-sector information provision with a factor correlation of 0.883; we retain the unidimensional specification for comparability and return to both points in Section 6.4.

Table 4 reports discriminant validity, assessed by three complementary criteria. Under the Fornell–Larcker criterion [59], the square root of each construct’s AVE (the bold diagonal) exceeded that construct’s correlations with the others, although the margin is narrow for perceived ESG legitimacy against ESG signal credibility (0.820 against 0.807). We also report the heterotrait–monotrait ratio of correlations (HTMT), a more sensitive criterion [60]. All HTMT ratios fall below the conservative 0.85 cutoff, with the same pair again the highest at 0.810. Because these two constructs are the closest in the matrix, we tested them additionally by constraining them into a single factor; the merged model fits substantially worse (Δχ²(5) = 1410.44, p < 0.001, ΔCFI = −0.030), and equivalent tests for the other adjacent pairs likewise favor the six-factor solution. The three criteria agree that the constructs are empirically distinct, while indicating that the legitimacy–credibility distinction should be interpreted with corresponding care.

Because all measures were self-reported in a single survey, we examined common method bias with five diagnostics, reported in Table 5 [66]; we treat the first three as comparatively weak and the last two as the basis for our conclusion. Harman’s single-factor test returned a first factor accounting for 52.96% of the variance; this exceeds the 50% guideline, and we no longer read it as evidence against method bias, since Harman’s test is widely recognized as the weakest available diagnostic [66]. A confirmatory single-factor model fit poorly (CFI = 0.700, RMSEA = 0.228), far worse than the measurement model (CFI = 0.977); poor single-factor fit is consistent with, but does not by itself demonstrate, the absence of method bias. Full-collinearity variance inflation factors were at most 1.80, below the 3.3 threshold [62]. More diagnostically, a common latent factor loading on all focal indicators in addition to their substantive factors absorbed 14.7% of indicator variance—below the 25% benchmark commonly cited for self-report research—and every structural path remained significant with the CLF retained (trust → legitimacy 0.500 → 0.416, legitimacy → consumption 0.502 → 0.494, direct path 0.350 → 0.302; all p < 0.001), with the proportion mediated barely moving (41.8% → 40.5%). Finally, a marker-variable analysis using a theoretically unrelated item measured in the same response format (smallest focal correlation 0.133) left all focal correlations large and significant after the Lindell–Whitney adjustment [63], and entering the marker as a covariate left the paths essentially unchanged. The two stronger diagnostics—the common latent factor and marker-variable analyses, not Harman’s test—are the basis for our conclusion that common method variance does not account for the observed associations.

5.2. The Integrated Moderated-Mediation Model

Because only the structural paths among the three latent constructs are just-identified, the mediation model’s overall fit coincides with that of the measurement model in Section 5.1 (the full model is over-identified, df = 32, and fits well); the added latent interactions are therefore assessed by the significance of their coefficients and of the index of moderated mediation rather than by a separate global-fit index. The mediation chain was confirmed (

Table 6. Standardized path estimates, indirect effect, indices of moderated mediation, and hypothesis decisions.

Path	β	z	95% CI	Result
Signal trust → ESG legitimacy (H1)	0.434 ***	20.9	[0.393, 0.475]	Supported
ESG legitimacy → Resilient consumption (H2)	0.434 ***	23.4	[0.398, 0.470]	Supported
Signal trust → Resilient consumption (H3)	0.318 ***	17.0	[0.281, 0.355]	Supported
Indirect: Signal trust → ESG legitimacy → Resilient consumption (H4)	0.187 ***	14.8	[0.140, 0.230]	Supported
Index of moderated mediation—Income (H5)	0.027 ***	—	[0.010, 0.042]	Supported
Index of moderated mediation—Greenwashing (H6)	−0.006	—	[−0.022, 0.010]	Not supported

Note. *** p < 0.001.

): ESG signal trust raised perceived ESG legitimacy (β = 0.43, p < 0.001), and both perceived legitimacy (β = 0.43, p < 0.001) and signal trust (β = 0.32, p < 0.001) raised resilient consumption. The trust → legitimacy → resilience indirect effect was significant, indicating partial mediation and supporting H1–H4; perceived legitimacy transmitted a substantial share of signal trust’s influence while a direct path remained. Adding the two latent interactions to the legitimacy → resilience path within the same structural model, the index of moderated mediation for income was positive and significant (0.027, 95% confidence interval (CI) [0.010, 0.042]), supporting H5 in direction, whereas the index for perceived greenwashing was non-significant (−0.006, 95% CI [−0.022, 0.010]); H6 was not supported. As Section 5.3 details, the income effect is present but quantitatively limited, and synergy predominates. The structural model accounted for 46% of the variance in resilient consumption. To gauge practical significance—important given a sample of 3000, at which statistical significance is easy to obtain—we report standardized effect sizes alongside the coefficients. The trust → legitimacy → consumption indirect effect is 0.188 (95% bootstrap CI [0.163, 0.213], 5000 resamples), accounting for 37.1% of the total effect of 0.506. Among predictors, only perceived legitimacy reaches a medium effect (f² = 0.278); trust approaches it (f² = 0.141); every moderator and control falls below Cohen’s small-effect benchmark of 0.02. On the five-point response scale, a one-SD rise in perceived legitimacy corresponds to a 0.285-point shift in resilient sustainable consumption, and a one-SD rise in trust corresponds to a 0.334-point shift in total. Table 6 accordingly reports, for every hypothesis, the path, standardized estimate, test statistic, bootstrap interval, and an explicit supported/not-supported verdict—H1 (0.434), H2 (0.432), H3 (0.318), and H4 (indirect 0.188) supported, H5 (index of moderated mediation 0.027) supported in direction, and H6 not supported.

5.3. Structural Estimates: Signal Credibility, Quality, and the Income Trade-Off

Table 7. Structural model predicting resilient sustainable consumption.

Predictor	Model 1	Model 2	Model 3
Female	+0.312 ***	+0.160 ***	+0.157 ***
Age	+0.148 ***	+0.024	+0.024
Education	+0.029	+0.018	+0.018
Income	+0.056 **	+0.050 ***	+0.049 ***
Greenwashing	—	+0.069 ***	+0.067 ***
Signal credibility	—	+0.087 ***	+0.085 ***
Signal quality	—	+0.022	+0.026
Signal trust	—	+0.317 ***	+0.318 ***
ESG legitimacy	—	+0.361 ***	+0.361 ***
Legitimacy × Income	—	—	+0.059 ***
Legitimacy × Greenwashing	—	—	−0.020
R2	0.049	0.463	0.467
ΔR2	—	+0.415 ***	+0.004 ***

Note. Standardized coefficients. ** p < 0.01, *** p < 0.001.

reports the structural model in which income and perceived greenwashing moderate the legitimacy → resilience path. It was built as a sequence of nested structural models estimated within the same CB-SEM: Model 1 entered the demographic controls; Model 2 added the signal antecedents, the two focal predictors (signal trust, ESG legitimacy), and the moderators (income, greenwashing); and Model 3 added the perceived-legitimacy × income and perceived-legitimacy × greenwashing latent interactions. Beyond demographic controls, perceived ESG legitimacy and signal trust were the dominant predictors of resilient consumption, and ESG signal credibility (SIGCRED, third-party rating trustworthiness) added a significant, if smaller, contribution (β = 0.085, p < 0.001); ESG signal quality (SIGQUAL, the perceived quality of firm- and public-sector disclosure) did not (β = 0.026, n.s.). In this specification, the perceived credibility of independent third-party signals was associated with resilient sustainable consumption, whereas the perceived quality of firms’ own disclosure was not, once credibility was accounted for. This contrast is not robust to how the outcome is operationalized: when the dependent variable is restricted to the two price-conditioned items, perceived signal quality becomes significant (β = 0.065, p < 0.001) while credibility weakens (β = 0.069, p = 0.002). We therefore report the contrast without treating it as evidence that firms’ own disclosure is ineffective (Section 6.2). The legitimacy × income interaction was positive and significant (β = 0.059, p < 0.001), while the legitimacy × greenwashing interaction was not (β = −0.020, n.s.). Perceived greenwashing carried a small positive main effect (β = 0.067, p < 0.001) rather than the hypothesized negative one. Adding the ESG-related predictors raised the variance explained in resilient consumption from 5% (controls only) to 47%. The positive greenwashing coefficient should not be over-read. It is not robust: entered item-by-item, only the corporate-referent item interacts with legitimacy (−0.026, p = 0.014) while the public-sector item does not (−0.004, n.s.); the moderation is non-significant or marginal across covariate sets (−0.016 to −0.020) and changes sign when the outcome is restricted to the price-conditioned items. Because the two-item measure conflates corporate and public-sector referents and the effect does not survive these checks, we treat the positive coefficient as an artifact of measurement rather than a substantive finding and do not interpret it (Section 6.4).

The positive coefficient is also not a suppression artifact. The zero-order correlation between perceived greenwashing and resilient sustainable consumption is already positive at 0.212, and the coefficient attenuates monotonically as predictors are added (0.212 with no covariates, 0.208 with demographics, 0.094 adding signal trust, 0.072 adding perceived legitimacy, and 0.069 in the full model), which is the reverse of a suppression pattern. The association is therefore genuine but largely shared with signal trust and perceived legitimacy.

The conditional indirect effects increased monotonically across the income distribution but only slightly: 0.16 at low income (−1 SD), 0.19 at the mean, and 0.21 at high income (+1 SD)—a gradient of just 0.05 across the full range. Even at the low-income end, the 95% confidence interval excludes zero, so the trust → legitimacy → resilient-consumption pathway holds strongly regardless of income. The trade-off is thus present but quantitatively limited: ESG trust and legitimacy are associated with resilient sustainable consumption across the whole income distribution, and more strongly at higher income, but the gradient is small relative to the level of the association itself [27]. The income moderation is robust to how income is coded (

Table 8. Robustness of the income moderation across alternative codings.

Specification	Legitimacy × Income	Test	p
1. Bracketed variable treated as continuous	0.060	t = 4.59	p < 0.001
2. Continuous income item, log-transformed	0.043	t = 3.50	p < 0.001
3. Continuous income item, untransformed	0.045	t = 3.90	p < 0.001
4. Rank-based ordinal coding	0.059	t = 4.40	p < 0.001
5. Seven dummy-coded brackets	—	F(7, 2978) = 3.46	p = 0.001
Effect size	ΔR2 = 0.004	f2 = 0.007	below small-effect benchmark

Note. Dependent variable: resilient sustainable consumption. All specifications include the same covariates. Standardized coefficients; n = 3000.

)—positive and significant when the ordered brackets are treated as continuous (0.060), when a continuous income item is used in log (0.043) or raw (0.045) form, when income is entered as dummy-coded groups (joint F(7, 2978) = 3.46, p = 0.001), and under a rank-based specification (0.059)—so it is not an artifact of the interval assumption. In magnitude, however, it is small: the interaction adds 0.004 to R² (f² = 0.007, below Cohen’s 0.02), and the legitimacy slope rises from 0.368 in the lowest income bracket to 0.594 in the highest. We therefore describe the trade-off as statistically robust but modest, and report these figures rather than attaching a label.

5.4. Multi-Group Analysis: Does Signal Credibility Change the Pathway?

Signaling theory implies that a signal shapes behavior only insofar as it is credible. We therefore tested whether the credibility of the broader ESG signal environment changes how demand-side resilience is produced, splitting the sample at the median of ESG signal credibility (perceived trustworthiness of third-party ratings) into a high-credibility group (n = 1384) and a low-credibility group (n = 1616). Because the structural comparison between the two groups is only meaningful if the constructs are measured equivalently in each, we formally tested measurement invariance in a sequence of nested multi-group confirmatory factor models estimated with a robust maximum-likelihood (MLR) estimator; the results appear in

Table 9. Measurement invariance across signal-credibility groups (high n = 1384; low n = 1616).

Model	CFI	TLI	RMSEA	SRMR	ΔCFI	ΔRMSEA	Decision
Configural	0.972	0.960	0.065	0.030	—	—	—
Metric	0.971	0.963	0.063	0.032	−0.001	−0.003	Supported
Scalar	0.966	0.961	0.065	0.037	−0.005	+0.002	Supported
Strict	0.964	0.963	0.063	0.040	−0.002	−0.002	Supported

Note. Δ relative to the preceding model. Procedure and criteria in Section 5.4.

. Following the standard four-step procedure [67,68], we evaluated, in increasing order of stringency: (i) configural invariance, in which the two groups share the same factor structure but all parameters are freely estimated; (ii) metric (weak) invariance, which constrains the factor loadings to be equal across groups and thereby establishes that the constructs have the same meaning in both; (iii) scalar (strong) invariance, which additionally constrains the item intercepts and is required before observed scores can be compared across groups; and (iv) strict (residual) invariance, the most demanding level, which further constrains the item residual variances so that the constructs are measured with equal precision in both groups. Because the chi-square difference test is oversensitive in large samples such as ours (n = 3000), invariance was judged by the change in approximate fit indices rather than by chi-square: a model retains invariance when the decrement in fit relative to the preceding model satisfies ΔCFI ≤ 0.01 [69] and ΔRMSEA ≤ 0.015 [70]. The configural model fit the data well (CFI = 0.972, RMSEA = 0.065, SRMR = 0.030). Each successive constraint left fit essentially unchanged: metric (ΔCFI = −0.001, ΔRMSEA = −0.003), scalar (ΔCFI = −0.005, ΔRMSEA = +0.002), and strict (ΔCFI = −0.002, ΔRMSEA = −0.002) all fell comfortably within the thresholds. Strict measurement invariance—the most demanding level—therefore holds, so the loadings, intercepts, and residual variances can be treated as equivalent across the high- and low-credibility groups and the structural coefficients compared directly.

All three structural paths differed across the groups, with between-group differences tested by Wald z-statistics (

Table 10. Multi-group comparison of structural paths by ESG signal credibility (standardized coefficients).

Structural Path	High-Credibility (n = 1384)	Low-Credibility (n = 1616)	Difference
Signal trust → ESG legitimacy	0.51 ***	0.37 ***	z = 3.54, p < 0.001
ESG legitimacy → resilient consumption	0.54 ***	0.40 ***	z = 2.16, p = 0.031
Signal trust → resilient consumption (direct)	0.21 ***	0.42 ***	z = −4.64, p < 0.001

Note. z = Wald test of the between-group difference in the path coefficient. *** p < 0.001. Bold = dominant route within each group.

). The two groups showed visibly different patterns of association with resilient sustainable consumption. Where signal credibility was high, the mediated route was more pronounced—signal trust was more strongly associated with perceived ESG legitimacy (β = 0.51 vs. 0.37; z = 3.54, p < 0.001) and legitimacy with resilient sustainable consumption (β = 0.54 vs. 0.40; z = 2.16, p = 0.031)—while the direct path was comparatively weak (β = 0.21). Where signal credibility was low, this ordering reversed: the mediated association weakened, and the direct association dominated, with signal trust related to consumption directly and more strongly (β = 0.42 vs. 0.21; z = −4.64, p < 0.001). In associational terms, a credible signal environment is one in which trust is accompanied by a legitimacy judgment that, in turn, is linked to premium-tolerant behavior, whereas in a less credible environment, trust is associated with behavior more directly. The perceived credibility of the signal environment is thus associated not only with the level of demand-side resilience but with the configuration of the pathway through which it appears—consistent with the credibility condition of signaling theory, though, given the cross-sectional design, we advance this as a difference in the pattern of associations rather than a demonstrated restructuring.

Because dichotomizing a continuous variable discards information and can manufacture group differences, we re-estimated signal credibility as a continuous latent moderator (

Table 11. ESG signal credibility as a continuous moderator.

Estimate	β	Test	p/CI
Legitimacy × credibility → resilient consumption	0.045	t = 4.88	p < 0.001
Trust × credibility → resilient consumption (direct)	−0.030	t = −2.68	p = 0.007
Both interactions entered simultaneously	0.063/−0.059	—	both p < 0.001
Trust × credibility → perceived legitimacy	0.007	t = 0.72	p = 0.473 (n.s.)
Index of moderated mediation	0.010	95% CI [0.004, 0.016]	excludes zero
Conditional indirect effect (−1 SD/mean/+1 SD)	0.076/0.088/0.101	—	monotonic
Effect size of the interaction	f2 = 0.008	—	below small-effect benchmark

Note. Standardized coefficients; n = 3000; 5000 bootstrap resamples.

) and treat the median-split multi-group model above as a supplementary robustness check that illustrates the same pattern. The continuous specification reproduces and sharpens it: credibility positively moderates the legitimacy → consumption path (β = 0.045, t = 4.88, p < 0.001) and negatively moderates the direct trust → consumption path (β = −0.030, t = −2.68, p = 0.007), but does not moderate the trust → legitimacy path (β = 0.007, p = 0.473), so the reorganization concerns the downstream conversion and the direct route rather than signal uptake. The index of moderated mediation through credibility is 0.010 (95% bootstrap CI [0.004, 0.016]), and the conditional indirect effect rises monotonically from 0.076 at −1 SD to 0.101 at +1 SD of credibility. The interaction is small in magnitude (f² = 0.008, below Cohen’s small benchmark), which we also state rather than reporting significance alone. The continuous and multi-group results agree, so the pattern is not an artifact of the median split.

5.5. Practical Significance

With a sample of 3000, statistical significance is readily attained and is therefore a weak guide to importance. We accordingly report effect sizes and express the focal associations in the original response metric. Effect sizes differ markedly across predictors. Perceived ESG legitimacy is the only predictor reaching a medium effect by Cohen’s benchmarks (f² = 0.278), and ESG signal trust approaches it (f² = 0.141). Every moderator and control falls below the benchmark for a small effect: perceived greenwashing: 0.009, the legitimacy × income interaction: 0.007, the legitimacy × greenwashing interaction: 0.001, income itself: 0.004, and sex: 0.012. The two focal constructs, in other words, carry the model, while the moderators qualify rather than drive the association.

In the original five-point metric, a one-standard-deviation increase in perceived ESG legitimacy is associated with a shift of 0.285 points in resilient sustainable consumption, and a one-standard-deviation increase in signal trust with 0.334 points in total, of which 0.210 is direct and 0.124 runs through legitimacy. The incremental variance explained follows the same ordering: demographics alone account for R² = 0.049; adding signal trust raises this to 0.300; adding perceived legitimacy raises this to 0.454; and the full model with the signal-environment variables and moderators reaches 0.467. Read together, these figures indicate that the managerially consequential quantities in this model are the level of ESG signal trust and the accompanying legitimacy judgment, rather than the conditions under which their association is stronger or weaker.

6. Discussion

Before turning to the implications, we restate the level at which our evidence sits. What we measure and test is consumer-level resilient sustainable consumption: an individual, intention-level disposition to keep buying sustainably produced goods when they carry a price penalty. Supply chain resilience proper—the capacity of a network to absorb and recover from disturbance—is not measured here, and no supply-chain-level outcome enters our models. The implications we draw below about supply chain viability are therefore theoretical and managerial extensions of a consumer-level result, not empirical conclusions about supply chains, and we mark them as such throughout.

These findings reframe a problem usually posed on the supply side. The viability of a sustainable supply chain depends not only on how robustly upstream firms can source, buffer, and recover [1,8], but on whether downstream demand holds when the premium bites. Our results indicate that this demand-side pattern is associated primarily with a cognitive route—trust accompanied by perceived legitimacy, and trust and legitimacy together associated with premium-tolerant behavior—rather than with demographics. The proximal correlate is perceived legitimacy, which is itself associated with trust, pointing managers and policymakers toward credible, legitimacy-conferring ESG communication rather than mere awareness-raising [11,33].

On the central question of whether sustainability and resilience trade off or reinforce each other, our evidence is nuanced. A trade-off is present: the ESG–resilience conversion is stronger among higher-income consumers, consistent with the idea that demand-side resilience is partly a function of economic slack [6,7]. But it is small in magnitude (f² = 0.007). The ESG slope remained large and positive even at low income, so ESG orientation is associated with stable demand across the income distribution. This mirrors, on the demand side, the producer-side finding that synergies tend to overshadow trade-offs once sustainability practices are in place [27]. For supply chain design, this is encouraging: investments that deepen consumer ESG trust and perceived legitimacy are associated with broad-based demand resilience rather than a narrow, affluence-dependent niche.

The greenwashing results warrant caution rather than interpretation. Although perceived greenwashing depresses green purchase intention in prior work [44,46], here it did not moderate the conversion of ESG orientation into resilient sustainable consumption, and its small positive main effect did not survive item-wise and alternative specifications. Because the measure combines corporate and public-sector referents in only two items, we treat this coefficient as a measurement artifact and refrain from the post hoc reading offered in earlier drafts. Whether information asymmetry [37] erodes demand-side resilience is best revisited with a properly specified, multi-item greenwashing scale.

The multi-group results sharpen the signaling interpretation while remaining associational. Greenwashing measured as a single perception was not associated with erosion of the conversion, but the perceived credibility of the broader signal environment was associated with a different configuration of the pathway. Where third-party ESG signals were credible, resilient sustainable consumption was reached through a legitimacy judgment—trust accompanied by a verdict that the premium was warranted. Where signals were not credible, that step was attenuated, and trust was associated with behavior more directly. This is consistent with the credibility precondition of signaling theory in the demand for sustainability, though, given the cross-sectional design, we read it as a difference in the pattern of associations rather than a demonstrated restructuring. For managers, the implication is that credible, independently verified ESG disclosure is associated with the value-based reasoning that appears to make demand-side resilience durable.

Finally, the size-graduated expectation structure speaks to sustainability that is inclusive of small and medium-sized enterprises (SMEs). Consumers expected ESG far more strongly of large and mid-sized firms (M = 4.30) than of small firms (M = 3.88) or micro-enterprises (M = 3.50; large-firm versus micro-enterprise contrast paired t = 46.7, p < 0.001). This suggests that verification effort may be most productively concentrated where expectations are highest; we are careful, however, not to read lower expectations toward micro-enterprises as grounds for lighter regulation, a point we develop in Section 6.3.

6.1. Theoretical Implications

Consumers are an under-theorized node of supply chain resilience. The dominant view locates resilience upstream, in the capacity of firms to source, buffer, and recover, but the willingness of consumers to keep absorbing sustainably produced goods at a premium, even under economic strain, is a plausible and complementary foundation of a resilient sustainable supply chain. The contributions below follow from taking that willingness seriously as an object of theory.

We separate the study’s conceptual contributions from its empirical ones. Conceptually, the study extends resilience thinking to that downstream node. We articulate a demand-side counterpart—a base of consumers who continue to absorb sustainably produced goods at a premium under economic strain—and propose resilient sustainable consumption as its consumer-level indicator. We present this as a conceptual extension that builds on adjacent constructs (willingness to pay, sustainable purchase behavior, customer loyalty, consumer resilience), not as an entirely new construct; establishing it as such would require discriminant validation against those constructs and evidence from behavior under an actual disturbance, which we identify as necessary next steps.

Empirically, the study brings signaling theory to bear on sustainable consumption beyond a main-effect logic. The multi-group result is the central empirical finding: the configuration of associations differs with perceived signal credibility. Where third-party ESG signals are credible, trust is accompanied by a legitimacy judgment that is linked to premium-tolerant behavior; where signals lack credibility, that legitimacy step is attenuated and trust is associated with behavior directly. This moves the question from ‘do signals matter?’ toward ‘how does signal credibility relate to the structure of consumer reasoning?’—while remaining, given the design, an associational rather than a mechanistic claim. We note that the integrated moderated-mediation specification and the latent-interaction approach are methodological rather than theoretical contributions, and label them as such.

The study also adjudicates empirically between the trade-off and synergy framings of the sustainability–resilience nexus [6,7,27]. Rather than asserting either, we quantify both: income moderates the legitimacy → resilience conversion (the trade-off is statistically present), but the conditional indirect effect remains positive even at low income and the interaction is small in magnitude (f² = 0.007), so synergy predominates. This quantified answer—a real but small trade-off—is more useful than a binary verdict and mirrors, on the demand side, producer-side findings upstream. Finally, the perceived-greenwashing results refine the information-asymmetry account only cautiously: a single greenwashing perception was not associated with attenuation of the conversion, but because the two-item measure is weak and its coefficient unstable, we draw no positive inference from it.

6.2. Managerial Implications

For supply chain and marketing managers, the results suggest priorities stated at the level the evidence supports. Because the evidence is consumer-level, the following should be read as managerial implications derived from a consumer-level pattern rather than as demonstrated effects on supply chain outcomes. Because perceived legitimacy is the proximal correlate of resilient sustainable consumption and is itself associated with signal trust, firms may find more leverage in credible, legitimacy-conferring ESG communication than in expanding their own disclosure. In this sample, perceived signal credibility was associated with resilient sustainable consumption, whereas perceived disclosure quality was not, once credibility was accounted for. We are careful not to over-read this contrast: it is not evidence that disclosure is ineffective—the two may be empirically entangled, disclosure may operate through credibility rather than independently of it, our measure captures perceived rather than audited quality, and in a robustness specification restricted to price-conditioned outcome items disclosure quality does become significant (+0.065, p < 0.001). The practical implication is therefore to pair disclosure with independent assurance rather than to abandon it: third-party certification and external verification act as demand-resilience investments that support the legitimacy reasoning associated with price-tolerant demand.

A second managerial implication concerns market segmentation. Because the ESG → resilience conversion stays strongly positive across the income distribution, managers should not treat sustainable products as a narrow luxury niche reserved for affluent buyers; demand-side resilience is broad-based. At the same time, the income moderation indicates that higher-income segments deliver a modest additional buffer, so a tiered strategy—broad sustainable positioning supported by premium variants for higher-income segments—captures both the base and the margin. Finally, the multi-group finding tells managers where to concentrate signaling effort: in markets or segments where rating credibility is low, raising the trustworthiness of signals can shift consumers onto the more durable, legitimacy-mediated route rather than leaving demand dependent on diffuse trust alone.

6.3. Policy Implications

For policymakers, the findings point toward strengthening the credibility infrastructure of ESG information rather than simply mandating more disclosure. Because consumers in this sample responded to the perceived trustworthiness of signals more than to the perceived quality of firms’ own disclosure, public measures that standardize ESG ratings, accredit third-party verifiers, and penalize greenwashing may do more to support demand-side resilience than rules that only expand firms’ reporting. A credible signaling environment functions, in effect, as public infrastructure for sustainable demand—though, as a consumer-perception result, this is one input to policy design rather than a determinant of it.

The size-graduated expectation structure invites a differentiated, rather than a deregulatory, reading. Consumers expected ESG far more of large firms than of small and micro-enterprises, which suggests that verification effort is most productively placed where expectations—and reputational stakes—are highest. We do not conclude that micro-enterprises should face lighter regulatory obligations: consumer expectation is only one input to regulatory design, which must also weigh environmental impact, administrative capacity, and equity between firms, and our data cannot, by themselves, set obligation thresholds. As differentiated guidance rather than a finding, one might expect independently assured, comparable disclosure from large firms; standardized but proportionate reporting, supported by a shared verification infrastructure, of mid-sized firms; and participation in collective or sector-level verification schemes, rather than firm-level reporting burdens, for micro-enterprises. We mark this discussion as extending beyond what our data establish.

6.4. Limitations and Future Research

The data are cross-sectional and measure behavioral intention rather than realized purchasing, so the resilience interpretation rests on stated premium tolerance. The intention–behavior gap is well documented in sustainable consumption, where stated intention overstates realized purchasing [11]; our estimates should therefore be read as evidence about stated premium tolerance rather than about behavior under an actual shock. This gap is not incidental to the present case. Sustainable and ethical consumption is the domain in which the attitude–behavior gap has been most consistently documented: consumers who report strong pro-environmental or ethical intentions frequently fail to act on them at the point of purchase, and the discrepancy widens precisely where the sustainable option carries a price penalty—the condition our outcome items impose. Stated premium tolerance should therefore be treated as an upper bound on realized premium-tolerant purchasing rather than as an estimate of it, and the magnitudes we report are best read comparatively, across consumers and conditions, rather than as absolute levels of demand-side resilience.

A cross-sectional design cannot establish temporal precedence, and we describe our estimates as associations throughout. We estimated the fully reversed model, a model with legitimacy and consumption as correlated outcomes of trust, and a model with legitimacy preceding trust; all are statistically equivalent to the hypothesized model, each returning χ²(32) = 442.18, CFI = 0.977 and RMSEA = 0.065 with identical information criteria. This equivalence is an algebraic property of just-identified structural models among three latent variables rather than a diagnostic result about these data, so fit cannot adjudicate directionality in any study of this form. Theory and measurement nonetheless favor the ordering we specify: signaling theory posits that inference about an unobservable attribute precedes evaluation and that evaluation precedes intention, and the trust and legitimacy items refer to ESG-practicing firms as a class whereas the consumption items refer to the respondent’s own prospective behavior. Longitudinal or experimental designs are required to settle the question.

No supply-chain-level outcome was measured. Demand stability, recovery time, supplier continuity, sales volatility, and network performance lie outside our data, so the link from individual premium-tolerant intentions to supply-chain-level resilience is advanced as a proposition rather than a finding. It requires that intentions translate into purchases at a non-trivial rate, that the intending segment is large enough to affect order volume, and that the effect persists under actual economic strain, none of which is tested here.

Four measurement caveats qualify the constructs. Discriminant validity between perceived ESG legitimacy and signal credibility is satisfied but not comfortably (HTMT = 0.810), and perceived legitimacy includes a value-for-money element, so distinguishing it from perceived value remains an open question for a purpose-built instrument. Signal quality is empirically two-dimensional, and perceived greenwashing rests on two items whose referents differ, which is why we do not interpret its coefficient. ESG signal trust, as measured, is trust in ESG-practicing firms rather than an appraisal of a specific signal instrument. The greenwashing limitation warrants a further word, since it bears directly on a null result. With only two items, one referring to corporate and the other to public-sector communication, the measure cannot separate skepticism about firms’ claims from skepticism about institutional communication, and it has no items capturing the specific forms greenwashing takes—vague claims, selective disclosure, unsubstantiated imagery. A two-item scale also attenuates reliability, which biases an interaction term toward zero. The non-significant moderation we report is therefore consistent with two quite different states of the world: greenwashing perception genuinely does not condition the conversion of legitimacy into premium-tolerant consumption, or it does so, but the measure is too coarse to detect it. We cannot distinguish these with the present data, and a purpose-built multi-item greenwashing scale that separates corporate from institutional referents is required before the question can be settled.

The study is set in a single national market with high institutional trust, a maturing but not yet mandatory ESG disclosure regime, and high digital penetration. Consumer ESG perception differs markedly across the more regulation-anchored EU, the more market-driven United States, and price-sensitive emerging markets, and the credibility-conditioning result is the most likely to be context-dependent, since it turns on the maturity of third-party verification infrastructure. The sample is quota-structured rather than representative and skews toward higher education and ESG awareness, so it may underrepresent less educated or less ESG-aware consumers. Accordingly, the estimates reported here should be generalized to other national and cultural contexts only with caution, and we frame every implication that follows as conditional on the Korean setting in which the data were collected.

Future research should, in order of priority, use longitudinal or panel designs capable of observing persistence through an actual demand shock; replicate across countries with differing ESG regulatory regimes; validate the construct against realized purchasing using transaction data; and integrate the demand-side buffer into formal models of supply chain resilience alongside supply-side capabilities [8,10].