Unverifiable Green Signals and Consumer Response in E-Commerce: Evidence from Platform-Level Data

Abstract

This study investigates the effects of unverifiable green signals—vague environmental claims, trust-substitute cues, and function-stacking—on consumer purchasing behaviors in e-commerce settings. Using detailed product-level data collected from two major Chinese online platforms, Taobao and Pinduoduo, during the peak shopping period in November 2023, we analyze the impact of these signals on product sales using ordinary least squares (OLS), instrumental variable (IV), and propensity score matching (PSM) methods. Results indicate that vague environmental language and function-stacking significantly boost sales across platforms, highlighting consumers’ preference for easily interpretable and seemingly comprehensive products. However, trust-substitute signals exhibit mixed effects, with them being beneficial on platforms with stronger credibility frameworks (Taobao) and less effective or even detrimental on platforms characterized by price competition and weaker governance (Pinduoduo). This study contributes to the literature on consumer trust and digital greenwashing by identifying platform-specific responses to unverifiable eco-claims and underscoring the importance of heuristic processing theories and trust formation mechanisms in digital marketing contexts. These findings underscore the complex dynamics of greenwashing strategies and stress the necessity for enhanced regulation and clearer communication standards to protect consumers and genuinely support sustainable consumption.

1. Introduction

From the UN’s Sustainable Development Goals to the European Union’s Green Deal, promoting sustainable consumption among consumers is world-widely acknowledged. Eco-labels and green claims have become key reference points in product evaluation, especially in online marketplaces [1]. However, many claims lack third-party certification or technical validation, leading to widespread unverifiable or misleading environmental messaging [2]. These claims mislead consumers, damage trust in legitimate eco-products, and weaken broader efforts toward sustainable consumption [3]. The problem is particularly severe in online energy-saving product categories. Recent research shows that electricity-saving plugs and battery rejuvenators are among the most common products associated with unverifiable green claims, which often advertise benefits like “saves 30% electricity” or “military-grade performance,” without credible evidence [4].

While certified eco-labels have been shown to enhance consumer trust and willingness to pay [5], uncertified products trigger different consumer responses. In the absence of verification, consumers increasingly rely on heuristic signals such as vague environmental language or brand-related cues [6,7]. Experimental evidence indicates that over 70% of consumers shift to heuristic processing when certification is missing [7].

Among these heuristic signals, two types are especially common in digital commerce: (a) Vague environmental signals, such as “green technology” or “energy-efficient,” offer unspecific claims without measurable benchmarks [8,9]. (b). Trust-substitute signals, such as “endorsed by experts” or “lifetime warranty,” substitute for actual technical proof [9,10]. And sellers often engage in function-stacking—highlighting multiple technical features or exaggerated functionalities to convey product reliability or superiority [11]. Although these strategies are widely used, little is known about their actual effects on purchasing behavior [12]. Existing studies tend to focus on greenwashing as a general phenomenon, without clearly differentiating between specific types of unverifiable signals or assessing how these signals interact and operate in practice. The role of “function-stacking” in shaping consumer trust and the conditions under which it may be effective remain under-explored.

Platform design further shapes how green signals are interpreted. Platforms emphasizing reputation and quality may amplify the effectiveness of trust-based claims, while price-focused platforms may foster consumer skepticism [13,14,15]. However, empirical studies comparing platform differences remain limited. Specifically, how platform governance and user base characteristics moderate consumer responses to unverifiable green signals has not been sufficiently studied. Given these limitations, our study raises the following research questions: (1) How do vague signals, trust-substitute signals, and function-stacking influence the sales of unverifiable green products in online marketplaces? (2) To what extent do platform characteristics—such as brand trust, price sensitivity, and governance strength—condition the effectiveness of these signals?

This study addresses these gaps by examining how vague signals, trust-substitute signals, and function-stacking affect online sales of unverifiable green products. Using product-level data from Taobao and Pinduoduo collected during China’s major shopping season in November 2023, we apply ordinary least squares (OLS), instrumental variables (IVs), and propensity score matching (PSM) methods. We also investigate how platform features shape consumer responses.

This study makes three key contributions. First, it refines the conceptual classification of unverifiable green signals by introducing “function-stacking” as a distinct and testable signal type, extending existing greenwashing frameworks that primarily focus on vague or trust-based claims. Second, it reveals how the effectiveness of different eco-signals varies across platforms with distinct institutional and consumer profiles, bridging the gap between green marketing theory and the platform governance literature. Third, it provides empirical evidence to support targeted regulatory interventions, especially for digital marketplaces that lack robust eco-claim verification mechanisms. Our study not only contributes to the theoretical understanding of signal heterogeneity and consumer heuristics in online green marketing but also informs platform-level governance and consumer protection strategies.

2. Literature Review

2.1. Green Signals in E-Commerce

Green signals are claims or indicators designed to communicate a product’s environmental benefits, such as energy savings, carbon footprint reduction, or eco-friendly manufacturing [16]. In traditional contexts, credible green signals help consumers identify sustainable options and reduce decision-making uncertainty [17]. Verified labels like Energy Star or government-backed eco-labels have been shown to increase consumers’ willingness to pay for green products by signaling authentic environmental commitment [3].

However, the growing emphasis on sustainability has also led to the proliferation of unverifiable green signals in digital marketplaces. Sellers exploit unverifiable green signals to create the illusion of environmental friendliness without ensuring actual product performance [18]. The existing literature identifies three common types of unverifiable green signals: (a) vague signals (generalized claims such as “high-efficiency”, “green technology,” or “saves 30% energy” without any standard reference [7]); (b) trust-substitute signals (heuristic trust triggers, like “lifetime warranty,” “military-grade,” or endorsements from prestigious organizations without verifiable links to environmental benefits [11]); and (c) function-stacking (exaggerating product capabilities through extensive feature listings (e.g., “overvoltage protection,” “real-time monitoring”) to imply high technological sophistication and eco-performance [13]).

The influence of green signals can be interpreted through the signaling theory, which addresses the issue of information asymmetry between buyers and sellers in the market [19]. Eco-labels help draw consumers’ attention to otherwise unobservable environmental attributes, allowing them to distinguish more environmentally friendly products from their less sustainable counterparts [20]. Compared with certified eco-labels, unverifiable sustainability claims—such as vague green statements—have also shown a positive effect on consumer choice and willingness to pay. Notably, this effect is mediated by perceived familiarity, which enhances consumers’ acceptance of such claims [21].

While green signals are intended to foster sustainable consumption, unverifiable green signals undermine this goal by distorting consumer perceptions and decision-making. Thus, we propose the following hypothesis:

H1: 

Products employing more unverifiable green signals—vague claims, trust-substitute cues, or function-stacking—achieve higher sales in e-commerce platforms, regardless of actual environmental performance.

2.2. Consumer Responses to Unverifiable Green Signals

Unverifiable green signals (UGSs) influence consumer decision-making primarily by activating heuristic trust mechanisms. In the absence of third-party certification, consumers tend to rely on alternative cues—such as vague slogans or implied authority—to assess product credibility [6].

Empirical research confirms that vague environmental language (e.g., “green technology,” “eco-friendly”) can enhance perceived product effectiveness, even when lacking substantiated evidence [15]. Similarly, trust-substitute signals (e.g., “military grade,” “endorsed by experts”) function as peripheral cues that evoke reliability by association rather than verified performance [22]. These cues often appeal to consumers’ desire to support pro-environmental goals [9].

However, the impact of UGSs is not uniform. Consumers with strong environmental values are more susceptible to heuristic green claims, as these align with their identity and attitudes [23]. In contrast, low-involvement consumers rely more heavily on external validations, such as user reviews and sales rankings, before accepting green claims as credible [24]. Still, the presence of multiple vague or exaggerated claims can trigger greenwashing fatigue and consumer skepticism [25].

Function-stacking, defined as listing numerous claimed product benefits, adds another persuasive dimension. By inflating the number of features—e.g., “voltage stabilization,” “battery optimization,” “overheat protection”—sellers can signal technological sophistication [25]. While moderate stacking enhances appeal, excessive claims without validation may lower authenticity and increase consumer hesitation.

The risks of unverifiable green signals are considerable. Studies indicate that such signals can mislead consumers into purchasing inferior or ineffective products, undermine trust in authentic green innovations, and ultimately erode the credibility of broader sustainability efforts [26]. Perceptions of greenwashing have also been shown to negatively affect consumers’ environmental beliefs and product evaluations, while reducing their sense of well-being during interactions with company websites [27].

Based on this, we propose the following hypothesis:

H2: 

Unverifiable green signals—including vague claims, trust-substitute cues, and function-stacking—positively influence product sales by enhancing perceived trust and product appeal, but their effects vary depending on consumer orientation and the coherence of signal content.

2.3. Platform Context and Greenwashing in Digital Markets

In online marketplaces, platform design fundamentally shapes how consumers interpret and respond to green signals. Without physical inspection or interpersonal interaction, users rely on digital intermediaries—platform structure, seller reputation, and governance mechanisms—to assess credibility [28].

Reputation systems, such as seller ratings, reviews, and transaction history, act as amplifiers for unverifiable signals. On platforms like Taobao, where the visibility of seller performance is high, trust-substitute cues become more persuasive due to their alignment with the platform-level trust architecture [29]. In contrast, on platforms like Pinduoduo, which emphasize aggressive pricing and limited merchant screening, consumers tend to be more skeptical and less responsive to unverifiable green claims [30].

Platform governance also moderates signal effects. Strong oversight—such as proactive content audits or green-claim vetting—can mitigate the spread of greenwashing, while loosely regulated platforms allow unverifiable claims to proliferate unchecked [31]. This institutional gap increases consumer vulnerability and erodes confidence in green marketing.

Demographic composition further differentiates platform response. Taobao tends to attract relatively affluent and environmentally aware consumers, who are more inclined to interpret green signals positively when presented in a trusted environment [32]. By contrast, Pinduoduo users—typically more price-sensitive—exhibit heightened caution and require stronger substantiation for green claims [33].

Finally, platform brand trust operates as a form of indirect certification. Green signals on a highly trusted platform are more likely to be accepted, while the same message on a platform with low perceived credibility may be disregarded or questioned [34].

In summary, platform-level features—including governance strength, reputational infrastructure, and user composition—jointly moderate how unverifiable green signals are interpreted and acted upon. This leads to the following hypothesis:

H3: 

The positive effects of unverifiable green signals on product sales are stronger on platforms with higher governance strength, reputational trust, and sustainability-oriented user demographics.

2.4. Theoretical Framework

To explain how unverifiable green signals shape consumer behavior, this study integrates heuristic processing theory, cue utilization, and trust formation mechanisms into a coherent conceptual chain:

Firstly, vague signals (e.g., “eco-friendly”, “green tech”) operate through affect heuristics, which substitute affective impressions for complex evaluations in low-information settings [35]. Such emotionally resonant but technically imprecise terms trigger pro-environmental instincts without requiring factual scrutiny. Studies show that green affect cues increase perceived sincerity and brand warmth, especially among moderately engaged consumers [36,37]. Moreover, trust-substitute signals (e.g., “lifetime warranty”, “expert endorsed”) reflect cue utilization theory, where peripheral cues replace absent central information [38]. These cues simulate authority and reduce cognitive effort, acting as trust triggers in environments lacking verification [39]. Their effectiveness hinges on contextual factors such as platform reputation and user expectations [40]. Finally, function-stacking—listing multiple technical claims—taps into feature-based heuristics, where more attributes signal higher value [41]. This aligns with the complexity–utility association: users equate numerous features with advanced performance, even without validation [42]. In digital commerce, such bundling boosts perceived credibility and reduces uncertainty [43], but excessive stacking may trigger skepticism [44].

3. Methodology and Data

3.1. Research Idea

This study investigates how unverifiable green signals influence consumer purchase behavior in e-commerce markets, focusing on three types of promotional strategies: vague signals, trust-substitute signals, and function-stacking.

The analysis begins with the construction of a product-level dataset covering electricity-saving devices and battery rejuvenators from two major Chinese e-commerce platforms, Taobao and Pinduoduo. Textual analysis techniques were applied to classify products based on the presence of specific green signals, using predefined keyword dictionaries. Ordinary least squares (OLS) regression models were first estimated to examine the associations between the different types of unverifiable signals and log-transformed sales. To address potential endogeneity concerns—particularly the possibility that sellers self-select signaling strategies based on unobserved product quality—an instrumental variable (IV) approach was implemented, using the category-level average warranty length as an instrument for trust-substitute signals. Robustness checks were then conducted. A placebo test was performed by randomly reassigning the trust-substitute signal variable, and extreme price trimming was applied to exclude products priced below CNY 10 or above CNY 500. Additionally, a propensity score matching (PSM) strategy was employed to estimate the average treatment effect of trust-substitute signals while improving covariate balance. Finally, platform-specific heterogeneity was examined by conducting subsample analyses for Taobao and Pinduoduo to explore how consumer responses vary depending on platform context. While the initial collection included JD.com and Douyin, the final dataset used for regression and heterogeneity analysis focused on Taobao and Pinduoduo. JD.com and Douyin presented several limitations: fewer listings of green plug-type products, inconsistent availability of key metadata (e.g., warranty information, green claims), and technical restrictions that impeded reliable and systematic crawling. To maintain data integrity and ensure comparability across variables, only platforms with high-quality, structured data were retained.

As the dataset is based on product-level listings rather than user-level behavioral data, demographic characteristics of platform users (e.g., age, gender, education) are not directly observable. To address this limitation, we incorporate several control variables that proxy for user segmentation and purchasing preferences—such as product price, discount rate, brand reputation, warranty availability, and green signal richness. Additionally, platform fixed effects are introduced to account for systematic differences in user bases between Taobao and Pinduoduo. These measures help mitigate potential bias stemming from unobserved demographic heterogeneity.

The overall research design is summarized in Figure 1.

3.2. Research Methods

To estimate the effects of unverifiable green signals on product sales, we employ three empirical strategies: ordinary least squares (OLS), instrumental variables (IVs) regression, and propensity score matching (PSM). Each method complements the others by addressing potential biases from different angles.

3.2.1. Ordinary Least Squares (OLS)

OLS serves as the baseline estimation strategy. The model regresses log-transformed product sales on three types of green signals, controlling for price and platform effects:

$$\begin{gathered} {\mathit{log}\_\mathit{sales}}_i=\alpha +{\beta }_1\cdot {\mathit{log}\_\mathit{price}}_i+{\beta }_2\cdot {\mathit{vague}\_\mathit{signal}}_i+{\beta }_3\cdot {\mathit{trust}\_\mathit{substitute}}_i \\ +{\beta }_4\cdot {\mathit{function}\_\mathit{stack}}_i+\gamma \cdot {\mathit{Platform}}_i+{ϵ}_i \end{gathered}$$

(1)

This model allows for the direct estimation of how each signal type correlates with consumer purchasing behavior. Standard errors are clustered at the seller level to account for potential intra-merchant correlation.

3.2.2. Instrumental Variable (IV) Regression

To mitigate endogeneity—particularly the concern that sellers with better products may be more likely to use strong trust signals—we use a two-stage least squares (2SLS) approach. The average warranty length within the same product category and platform, excluding the focal observation, is used as an instrument for the trust-substitute signal. The category-level average warranty length is expected to influence a seller’s signaling decision because sellers often benchmark their signal strategies—such as offering trust-substitute claims—against peer norms within the same product category and platform. When the prevailing warranty duration among competitors increases, individual sellers are more likely to adopt similar trust-enhancing strategies to remain competitive.

The below estimation follows:

First stage:

$${\mathrm{trust}\_\mathrm{substitute}}_{\mathrm{i}}={\mathsf{\delta }}_0+{\mathsf{\delta }}_1\cdot {\mathrm{avg}\_\mathrm{warranty}}_{-\mathrm{i}}+{\mathsf{\delta }}_2\cdot {\log \_\mathrm{price}}_{\mathrm{i}}+{\mathsf{\delta }}_3\cdot {\mathrm{vague}\_\mathrm{signal}}_{\mathrm{i}}+{\mathsf{\delta }}_4\cdot {\mathrm{function}\_\mathrm{stack}}_{\mathrm{i}}+{\mathsf{\mu }}_{\mathrm{i}}$$

(2)

Second stage:

$${\log \_\mathrm{sales}}_{\mathrm{i}}=\mathsf{\alpha }+{\mathsf{\beta }}_1\cdot {\log \_\mathrm{price}}_{\mathrm{i}}+{\mathsf{\beta }}_2\cdot {\mathrm{vague}\_\mathrm{signal}}_{\mathrm{i}}+{\mathsf{\beta }}_3\cdot {\widehat{\mathrm{trust}\_\mathrm{substitute}}}_{\mathrm{i}}+{\mathsf{\beta }}_4\cdot {\mathrm{function}\_\mathrm{stack}}_{\mathrm{i}}+{\mathsf{ϵ}}_{\mathrm{i}}$$

(3)

This approach isolates exogenous variation in the trust-substitute variable. The instrument is chosen for its relevance (warranty norms within a category influence signaling) and its plausibility of satisfying the exclusion restriction. Specifically, the category-level average warranty length—our instrumental variable—is unlikely to directly influence product-level sales. This variable reflects broader industry norms across all products within a subcategory, rather than product-specific quality or marketing efforts. Since consumers on digital platforms typically do not observe or consider the average warranty of similar products in the category, its impact on purchase decisions should operate only indirectly—through the seller’s choice to include a trust-substitute signal.

3.2.3. Propensity Score Matching (PSM)

To further reduce bias from observable confounders, we implement PSM. Products with and without trust-substitute signals are matched based on their log price, function-stacking, and vague signal presence, using nearest neighbor matching with a 1:1 ratio.

Let ${\mathrm{T}}_{\mathrm{i}}$ be the binary treatment (trust-substitute signal), and ${\mathrm{X}}_{\mathrm{i}}$ is the set of covariates. We estimate the average treatment effect on those treated (ATT):

$$\mathrm{ATT}=\mathbb{E}\left[{\mathrm{Y}}_{\mathrm{i}}\left(1\right)-{\mathrm{Y}}_{\mathrm{i}}\left(0\right)\mid {\mathrm{T}}_{\mathrm{i}}=1\right]$$

(4)

This non-parametric method ensures that treated and control groups are comparable on observed characteristics, strengthening causal inference in the presence of selection on observables.

3.3. Data Source and Description

This study utilizes product-level data collected from four major Chinese e-commerce platforms: Taobao, Pinduoduo, JD.com, and Douyin. Data collection was conducted between 1 November and 30 November 2023, a period that includes China’s “Double 11” shopping festival, during which promotional activities and consumer purchasing intensify. This timeframe ensures that the dataset captures a representative snapshot of active marketing and consumer behavior during a critical retail season.

Product data was obtained by assessing open-access web pages from the respective platforms. Two categories prone to unverifiable green claims—electricity-saving devices and battery rejuvenators—were targeted. Products were initially retrieved using relevant functional keywords (e.g., “power saver,” “battery repair”) to maximize coverage. Subsequently, a manual screening process was conducted to eliminate irrelevant listings and ensure that only products genuinely claiming green or energy-saving attributes were included. This two-stage process of keyword-based retrieval and human verification aligns with best practices for data accuracy and relevance in online marketplace studies.

The original dataset contained over 500 products. After removing duplicates, irrelevant entries, and products with missing key information, a final analytical sample of 337 products was retained.

Among these, the distribution across platforms was as follows: Taobao (115 products), Pinduoduo (117 products), JD.com (72 products), and Douyin (34 products). In subsequent platform heterogeneity analyses, we focused exclusively on two platforms, Taobao and Pinduoduo, to ensure adequate statistical power. This decision was based on the consistency and completeness of the data from these platforms. Although JD.com and Douyin were included in the initial search, they lacked key signaling variables and exhibited poor metadata quality.

Each product’s detailed textual description—including features, promotional slogans, and warranty terms—was extracted. From these texts, three core types of unverifiable green signals were systematically identified:

Vague signals: Claims implying environmental benefits without verifiable standards (e.g., “energy-saving,” “green technology,” “intelligent efficiency”), detected via keyword dictionary matching;

Trust-substitute signals: Heuristic cues such as “lifetime warranty,” “military-grade quality,” or references to reputable institutions;

Function-stacking count: The number of distinct technical features claimed, based on a standardized functional keywords list.

When products exhibited both types of signals, each was coded independently without aggregation. All signal extraction followed predefined, non-subjective text mining protocols to ensure consistency.

Descriptive statistics for the key variables are presented in

Table 1. Descriptive statistics and platform comparison (N = 337).

Variable	Description	Unit	Mean	SD	Min	Max	Taobao	Pinduoduo	p-Value
Log sales	Log of product sales volume	log(units)	3.90	2.60	0.00	9.21	3.355	4.917	<0.001 ***
Log price	Log of product price	log(CNY)	3.52	0.84	0.70	5.39	3.607	3.350	0.0177 *
Function-stack	Number of technical feature claims	count	2.39	1.51	1	8	2.250	2.658	0.0260 *
Trust-substitute	Count of trust-related words	count	1.83	1.64	0	6	1.882	1.726	0.3882
Vague signal	Count of vague green promo terms	count	0.88	0.94	0	3	0.886	0.855	0.7678
Warranty duration	Warranty duration (text-extracted)	days	55.6	58.5	1	365	54.413	59.074	0.7551

Note: p-values are based on two-sample t-tests between Taobao and Pinduoduo. *** p < 0.001, ** p < 0.01, * p < 0.05; platform differences were tested via two-sample t-tests.

.

The mean of log-transformed monthly sales (log sales) was 3.90 (SD = 2.60), while the mean log-transformed price (log price) was 3.52 (SD = 0.84). Notably, Pinduoduo products demonstrated a higher average count of trust-substitute signals compared to Taobao products (p < 0.01), whereas vague signal prevalence was relatively higher on Taobao (p < 0.05). These platform-level differences in signal deployment highlight important contextual factors to be explored in subsequent analyses.

4. Results

4.1. Main Regression Results

Table 2. Regression results of log (Sales).

Variables	OLS (Full)	OLS (Taobao)	OLS (Pinduoduo)	IV Model	Placebo Test	Robustness Trim
Intercept	3.798 (0.619) ***	3.672 (0.922) ***	3.835 (0.761) ***	−12.242 (56.042)	3.923 (0.634) ***	3.357 (0.836) ***
log(price)	−0.672 (0.158) ***	−0.744 (0.227) **	−0.265 (0.212)	−9.604 (30.843)	−0.634 (0.155) ***	−0.562 (0.215) **
Vague signal	0.841 (0.147) ***	0.684 (0.184) ***	1.097 (0.230) ***	−12.367 (45.519)	0.894 (0.140) ***	0.784 (0.154) ***
Trust-substitute	0.103 (0.088)	0.231 (0.102) *	−0.305 (0.155)∙	24.957 (85.555)	−0.031 (0.077)	0.103 (0.090)
Function-stacking	0.643 (0.088) ***	0.589 (0.124) ***	0.586 (0.118) ***	6.326 (19.609)	0.618 (0.086) ***	0.672 (0.091) ***
Placebo signal	—	—	—	—	−0.031 (0.077)	—
R2/Adj.R2	0.229/0.219	0.169/0.153	0.349/0.326	−186.3/−188.5	0.226/0.217	0.212/0.202
F/Wald test	F = 24.62 ***	F = 10.89 ***	F = 15.00 ***	Wald = 0.12 (ns)	F = 24.22 ***	F = 20.21 ***
N	337	220	117	337	337	305

Notes: Standard errors are in parentheses. *** p < 0.001, ** p < 0.01, * p < 0.05, ∙ p < 0.1, — indicates not applicable.

presents the results from the ordinary least squares (OLS) and instrumental variable (IV) regressions, with log sales as the dependent variable.

In the OLS full-sample model, both vague signal and function-stacking show significantly positive effects: a one-unit increase in vague signal frequency is associated with a 0.84 unit increase in log sales (p < 0.001); and each additional function claim leads to a 0.64 unit increase in log sales (p < 0.001).

The effect of trust-substitute signals is relatively weak overall, but platform-level analysis reveals significant variation: On Taobao, trust-substitute signals are positively associated with sales (coefficient = 0.23, p =0.025). On Pinduoduo, the effect is negative and marginally significant (coefficient = −0.31, p = 0.052), suggesting that consumers on lower-trust platforms may discount institutional or warranty-based claims.

The instrumental variable (IV) model was also tested, using average warranty length at the platform category level as an instrument for trust-substitute signals. However, due to the weak instrument, IV results are presented for completeness but are not interpreted further.

4.2. Robustness Checks

Two robustness checks were conducted to validate the consistency of the OLS results: a placebo test and a price trimming procedure, shown in

Table 3. Robustness check results (placebo test and trimmed sample).

Variable	Placebo Test	Trimmed Sample
Log price	−0.634 ***	−0.562 **
Vague signal	0.894 ***	0.784 ***
Placebo signal	−0.031	—
Trust-substitute	—	0.103
Function-stacking	0.618 ***	0.672 ***
Constant	3.923 ***	3.357 ***
Observations	337	303
Adj. R2	0.2166	0.2017

Note: In the placebo test, a randomized placebo signal is used to replace a trust-substitute; in the trimmed sample, products priced below CNY 10 or above CNY 500 are excluded; standard errors are robust. *** p < 0.001, ** p < 0.01, * p < 0.05.

.

In the placebo test, the trust-substitute signal variable was randomly reassigned to mimic a false treatment. As expected, the placebo signal shows no significant association with log sales (coefficient = −0.03, p = 0.686), confirming that the original effect was not due to random data structure. Meanwhile, the effects of vague signal and function-stacking remain statistically significant and in magnitude near the baseline OLS estimates, indicating that their influence on sales is robust.

The price trimming check excludes extreme observations with prices below CNY 10 or above CNY 500. After trimming, both vague signal use and function-stacking still exhibit positive and significant effects on log sales, while log price continues to show a negative impact. Trust-substitute signals remain statistically insignificant in this trimmed sample.

These robustness tests support the validity of the main findings: vague environmental language and functional over-claiming consistently promote product sales, whereas trust-substitute signals exhibit weaker or unstable effects across samples.

4.3. Treatment Effects from Matching

Propensity score matching (PSM) is used to test the effect of trust-substitute signals under more comparable conditions. Listings that contain at least one such signal are matched one-to-one with listings that do not, using log price, the number of vague signals, and the count of stacked functions as matching variables. After matching, these variables are well balanced; for instance, the mean gap in vague signals falls to zero.

Table 4. Average treatment effect (ATT) from propensity score matching.

Model	ATT Estimate	Std. Error	p-Value	Implied Sales Increase
PSM (Optimal 1:1 Matching, ATT)	0.816	0.318	0.011	2.26

Note: The dependent variable is log sales. Matching covariates include log price, vague signal use, and function-stacking count.

reports an average treatment effect of 0.816 log units (p = 0.011), which equals roughly 2.3 extra monthly sales for products that display trust-substitute cues. The positive effect appears only in the matched sample, explaining why OLS gave weaker results: PSM compares listings that are alike in price and other green signals, while OLS mixes very different products. Overall, trust-substitute signals can raise sales when competing items are similar, but the size of the gain is likely to differ across platforms and settings.

4.4. Summary of Key Patterns Across Platforms

Figure 2 provides a concise comparison of the effects of three types of unverifiable green signals across Taobao and Pinduoduo. The results reveal two notable patterns. First, vague signals and function-stacking consistently exert positive and statistically significant influences on product sales across both platforms, reinforcing their persuasive power in online green marketing. In contrast, trust-substitute signals demonstrate divergent effects: while they are positively associated with sales on Taobao (p < 0.05), they exert a negative, marginally significant effect on Pinduoduo (p ≈ 0.1). This divergence may reflect differences in how platform-specific trust architectures shape consumer interpretations of trust-related language. Second, the overall signal effectiveness appears to be stronger on Pinduoduo, with larger coefficient magnitudes observed for both vague signals and function-stacking, suggesting that consumers on price-sensitive platforms may rely more heavily on heuristic cues when evaluating product claims. These findings underscore the importance of the platform context in moderating green marketing effectiveness.

5. Discussion

5.1. Summary of Key Findings

This study provided empirical evidence on how unverifiable green signals influence consumer decisions in digital marketplaces. Based on product-level data from two major e-commerce platforms in China, we found that vague signals—claims that lack technical clarity but suggest environmental friendliness—significantly increase product sales. This effect is consistent across platforms and robust across multiple estimation strategies.

In contrast, trust-substitute signals such as “lifetime warranty” or institutional affiliations show mixed effects. On Taobao, these signals are positively associated with higher sales, whereas on Pinduoduo they appear to reduce consumer trust, yielding marginally negative effects. This platform-level heterogeneity underscores the contextual dependence of signal interpretation.

Furthermore, we found a strong and stable effect of function-stacking—the number of distinct functional claims in product descriptions—on sales. This promotional tactic enhances perceived product value regardless of platform.

While our instrumental variable (IV) model failed to produce precise causal estimates due to weak instrument strength, results from propensity score matching (PSM) validate the main finding: trust-substitute signals are associated with significantly higher sales, after adjusting for observed confounders.

5.2. Revisiting the Effectiveness of Unverifiable Green Signals

Our results indicated that vague signals and function-stacking significantly enhance product sales, while trust-substitute signals exhibit more limited effects. This pattern highlights the nuanced roles different forms of unverifiable green signals play in shaping consumer decisions.

Vague green claims, although lacking formal certification, appeal to consumers’ environmental aspirations and are often perceived as benign or well-intentioned. Prior studies suggest that when third-party verification is absent, vague signals can still serve as emotional shortcuts, especially for consumers with moderate environmental awareness [22].

Function-stacking exerts a particularly strong influence. By listing multiple technical features—such as “energy saving,” “voltage protection,” and a “smart chip”—sellers increase the perceived complexity and utility of their products. This tactic aligns with prior findings showing that multi-claim bundling boosts product attractiveness by triggering consumers’ mental models of value accumulation [25]. Even without proof, the sheer number of functions listed seems to legitimize green narratives in the eyes of users.

However, the limited effect of trust-substitute signals suggests growing consumer selectivity. While past studies have shown that terms like “military grade” or “endorsed by experts” boost perceived credibility [11], our results imply that such signals may be less effective in over-saturated environments. The diminishing impact may stem from increased consumer familiarity with these tactics, especially when platforms lack mechanisms to validate such claims.

5.3. Contextualizing Platform-Specific Effects

Our results reveal that the effects of unverifiable green signals differ markedly between Taobao and Pinduoduo, suggesting that platform context conditions signal interpretation.

On Taobao, both vague and trust-substitute signals show stronger effects. This aligns with the literature showing that Taobao’s reputation system and seller scoring mechanisms enhance the perceived credibility of green claims, even those lacking verification [45]. In contrast, on Pinduoduo, only vague signals and function-stacking significantly influence sales. Trust-substitute signals have a negligible or even negative effect. Prior studies suggest that Pinduoduo’s price-sensitive users, who are less likely to trust high-credibility cues without supporting evidence, are more skeptical of unverifiable claims [30]. Additionally, the platform’s low-regulation environment allows for widespread signal inflation, which may erode the effectiveness of formalistic cues over time [31].

Interestingly, the positive impact of function-stacking persists across both platforms, underscoring its generalizability. This may reflect a broader consumer tendency to conflate feature quantity with functional depth, regardless of platform characteristics [25].

Taken together, our findings support the view that platform governance, user demographics, and trust infrastructure jointly moderate how consumers process unverifiable green signals. In digital markets with contrasting user profiles and credibility structures, the same green message may yield dramatically different behavioral outcomes.

5.4. Theoretical Contributions and Policy Relevance

This study was motivated by the growing prevalence of unverifiable green claims in digital commerce and the lack of fine-grained empirical evidence on how specific heuristic signals affect consumer behavior. By examining three distinct forms—vague signals, trust-substitute signals, and function-stacking—we addressed the overarching question of how unverifiable eco-signals shape product performance in heterogeneous platform environments.

Our findings support the proposed hypotheses: vague signals and function-stacking consistently boost sales, while the effect of trust-substitute signals varies by platform. These results confirm that different signals trigger distinct cognitive responses, depending on signal form and platform context. This helps validate a signal-specific framework of greenwashing, moving beyond general treatments of unverifiable claims.

Theoretically, we contribute by refining the classification of unverifiable signals in green marketing and revealing their heterogeneous effects. We also advance the understanding of platform conditioning by showing how governance structures and user trust profiles moderate the persuasiveness of eco-claims.

Practically, our results suggest that platform-based regulation should adopt differentiated approaches. Stronger signal validation mechanisms—especially for trust-substitute claims—are needed on platforms with low buyer trust. At the same time, educational nudges that promote critical evaluation of vague claims and multi-function marketing may help consumers navigate the information asymmetry of digital markets.

6. Conclusions

This study investigates how unverifiable green signals influence consumer behavior in digital marketplaces, with a focus on three signal types—vague environmental claims, trust-substitute cues, and function-stacking strategies. Using product-level data from Taobao and Pinduoduo, we applied OLS, IV, and PSM models to assess how these signals affect short-term sales. Our results indicate that trust-substitute signals are particularly effective in boosting sales, especially when supported by longer warranties or platform structures that amplify consumer trust. However, the effectiveness of vague and stacked signals is more context-dependent, varying across platforms and pricing levels.

These findings provide evidence that unverifiable green marketing practices are not uniformly deceptive or ineffective but rather contextually shaped by platform environments and consumer expectations. The study contributes to the literature on greenwashing and digital commerce by offering a refined signal typology and identifying conditions under which different signals influence consumer behavior.

Despite its contributions, this study has several limitations. First, the analysis is limited to two product categories—electricity-saving plugs and battery rejuvenators—on two Chinese e-commerce platforms. While this focus ensures internal consistency, it restricts generalizability. Second, although both IV and PSM methods were employed to address endogeneity, the instrumental variable used has limited strength, reflecting the challenge of causal inference in high-endogeneity contexts. Third, the study only captures short-term sales responses and does not assess long-term consequences such as trust erosion or brand credibility.

Future research could enhance these findings by incorporating user-level behavioral data to examine repeat purchases and trust formation, expanding the scope to other product categories, and evaluating the impact of recent regulatory efforts targeting unverifiable green claims.