An important aspect of this research is its contribution through non-significant findings. While much of the extant literature assumes and reports positive engagement-loyalty relationships, the absence of such effects in our rigorous panel data analysis carries substantial theoretical value. Our study is positioned not to validate the mainstream assumption that social media engagement drives brand loyalty, but rather to investigate why engagement may fail to convert into loyalty in emerging market contexts within short-term observation windows. The pattern of null results across all hypotheses challenges the prevailing narrative in digital marketing research and invites a more nuanced understanding of the boundary conditions under which engagement-loyalty relationships operate. By documenting what does not work in a specific empirical context, this research contributes to theory building by highlighting the gap between engagement metrics and relationship outcomes, and by suggesting that the translation of digital interactions into consumer loyalty may require longer time horizons or different mechanisms than commonly assumed.

The rest of this paper is structured in the following way. Section 2 is a theoretical framework and elaborates the research hypotheses. Section 3 is the description of the research methodology, i.e., data collection, the variable operation and the analytical approach. Section 4 includes the empirical findings. The findings and their implications are discussed in section 5. Section 6 ends with findings restriction and future studies.

This study uses the Stimulus-Organism-Response (S-O-R) model to identify the working mechanism under which the engagement with social media may affect brand loyalty. The S-O-R framework was originally suggested by Mehrabian and Russell (1974), but since that time, there are numerous studies that used the S-O-R model to explore the connection between consumer behavior and environmental stimuli (Kim et al., 2020). It assumes that the stimuli in the external environment cause cognitive and affective processing in the organism, and it then results in approach or avoidance behaviors.

The brand-consumer interactions on social media platforms such as Instagram form a part of environmental stimuli that may shape the consumers in terms of their internal appraisals of the brand in the situation of the social media marketing. Such stimuli will be content exposure, likes, comments, shares, and other interactive elements involved in social media communication (Ming et al., 2021). The organism component relates to the cognitive and affective reaction of the consumers to the stimuli such as perceptions about brand trust, satisfaction and emotional attachment. The response component gives behavioral consequences like brand loyalty, re purchase intentions and recommendation behavior.

The S-O-R framework offers a valuable theoretical perspective on social media marketing impacts since it clearly takes into consideration the mediation of the effects of internal psychological states. The framework does not presuppose direct influence of marketing stimuli on consumer behaviors but recognizes that they occur via cognitive and emotional processing (Liu et al., 2023). This view is in line with the current view of consumer decision-making as complex and entailing rational and affective involvement in decision-making. The framework has seen effective implementation in different online marketing settings such as online shopping behavior, the adoption of mobile commerce and the use of virtual reality tourism online.

However, it is important to note that the S-O-R framework was originally developed in the context of physical environmental stimuli and immediate behavioral responses. Its application to high-frequency social media data, where stimuli occur continuously and responses may develop over extended time periods, represents an extension that warrants empirical scrutiny. The framework’s assumption of relatively immediate stimulus–response linkages may not fully capture the slow-moving nature of trust and loyalty formation in digital brand relationships.

Complementary theoretical background to the relationship-building process between the brands and the consumers is provided by the Social Exchange Theory (SET). SET was developed by Homans (1958) and Blau (1964) as an explanation of social behavior as a sequence of interdependent transactions with parties aiming to increase the benefits and reduce cost. The theory brings the concept of reciprocity as a central theme when it comes to the formation and continuation of relationships (Cropanzano and Mitchell, 2005).

Applied to brand-consumer relationships, SET suggests that when social media engagement creates value on the brands, consumers may develop a sense of obligation and reciprocate it by displaying trust and loyalty behaviors (Teichmann, 2021). This exchange process could potentially be facilitated through social media whereby brands and consumers can communicate directly, deliver content personally, and provide feedback through the use of social media. As a result of the consumer being made to believe that brands are always willing to provide value by way of engaging content, timely responses, and meaningful interactions, the consumers may end up trusting the reliability and benevolence of the brand (Aggarwal, 2004).

The social media interaction mechanism behind the SET provides a theoretical basis for understanding how brand loyalty might be increased through the use of social media. Brands that provide valuable content and interactions to consumers may create a sense of psychological indebtedness that could drive reciprocal actions like further engagement, a positive-word of-mouth, and brand preference (Kumar and Kumar, 2020). Nevertheless, SET also implies that this reciprocity has its limits, too much or too invasive interaction can be seen as a transgression of exchange rules, which might reduce, instead of enriching, the quality of relationships. This theoretical consideration provides a basis for investigating potential nonlinear patterns in the engagement-loyalty relationship, though the specific functional form remains an empirical question.

Based on the theoretical discussions presented in the section above, we formulate exploratory hypotheses on the potential relationship between social media engagement and brand trust and brand loyalty. These hypotheses are based on synthesis of the principles of S-O-R framework, the insights of the Social Exchange Theory, and the findings of the previous studies in the area of digital marketing and consumer behavior. Given the limited prior research in emerging market contexts and the potential for boundary conditions to operate differently across settings, we approach these hypotheses with appropriate caution regarding their expected support.

An important theoretical consideration that informs our empirical approach is the distinction between short-term and long-term effects of social media engagement. The theoretical frameworks reviewed above—S-O-R and Social Exchange Theory—do not explicitly specify the time horizon over which engagement effects should manifest. This ambiguity has significant implications for empirical testing.

Trust and loyalty are widely recognized as slow-moving psychological constructs that develop through accumulated positive experiences over extended time periods (Delgado-Ballester and Munuera-Alemán, 2005). The formation of genuine brand relationships typically requires consistent, repeated interactions that gradually build confidence and emotional attachment. In contrast, social media engagement metrics fluctuate on a weekly or even daily basis, reflecting short-term variations in content posting, platform algorithms, and consumer attention.

This temporal mismatch raises the possibility that short-term fluctuations in engagement may not produce immediate corresponding changes in trust and loyalty. Weekly variations in engagement may constitute “noise” relative to the underlying relationship-building processes that unfold over months or years. Consumers may not update their trust and loyalty dispositions in response to each week’s engagement levels; rather, these psychological states may respond to cumulative engagement patterns over longer horizons.

Furthermore, the nature of social media engagement in emerging markets may differ qualitatively from that in developed markets. Engagement in these contexts may be more entertainment-oriented, driven by content virality or novelty rather than genuine brand relationship building (Zeqiri et al., 2024). Such “symbolic” engagement may generate high metrics without translating into the reciprocal relationship dynamics posited by Social Exchange Theory.

These considerations suggest that our weekly panel data analysis may be better positioned to detect short-term engagement effects (if they exist) than to capture the long-term relationship-building processes emphasized in the theoretical literature. The absence of short-term effects would not necessarily imply that engagement is ineffective, but rather that its effects may operate over longer time horizons than our observation window permits.

The data used in this study were collected from publicly available Instagram brand accounts and user-generated comment content across multiple countries. This study adopts a longitudinal panel research design to examine the short-term relationships among social media engagement, brand trust, and brand loyalty. Panel data combine cross-sectional and time-series information, allowing researchers to control for unobserved heterogeneity across units and to model dynamic relationships using lagged specifications (Hsiao, 2014). Compared with purely cross-sectional designs, panel methods provide greater statistical efficiency and stronger causal inference under appropriate assumptions.

The panel structure consists of five global brands observed across five emerging market countries over a 24-week period, yielding a theoretical maximum of 600 brand–country–week observations. The observation window spans from January 1, 2024 to June 16, 2024, which enables the analysis of high-frequency weekly variation in engagement and consumer response indicators while maintaining relevance to current platform dynamics. After incorporating one-period lags in the explanatory variables to mitigate simultaneity concerns, the final analytical sample includes 575 observations. This sample size provides adequate statistical power to detect moderate effect sizes and supports the use of fixed-effects panel estimation.

The brand sample includes five internationally recognized companies from diverse industries: Nike (sportswear), Samsung (electronics), Coca-Cola (beverages), L’Oréal (cosmetics), and Toyota (automotive). These brands were selected based on several criteria: inclusion in the Interbrand Best Global Brands ranking, active and consistent presence on Instagram across sampled countries, representation of heterogeneous product categories, and variation in brand age and competitive positioning. This diversity enhances the external validity of the findings and reduces the likelihood that results are driven by industry-specific characteristics.

The country sample consists of five major emerging economies: India, Brazil, Indonesia, Mexico, and South Africa. These countries represent different geographic regions and exhibit substantial variation in economic development, internet penetration, and social media adoption. This cross-national heterogeneity allows the analysis to capture contextual variation while maintaining a focused emphasis on emerging market environments.

This study operationalizes three core constructs: social media engagement (independent variable), brand trust (mediator), and brand loyalty (dependent variable). All measures were constructed using objective platform data and text-based indicators derived from user-generated content.

Although NLP-based measures offer the advantage of using naturally occurring behavioral data, they may not fully capture the multidimensional psychological nature of trust and loyalty. Several sources of measurement error warrant consideration. First, consumers may hold trust and loyalty attitudes without explicitly expressing them in comment text. Second, keyword-based classification and sentiment algorithms may generate false positives or false negatives. Third, the relatively low frequency of explicit trust and loyalty expressions may reduce the signal-to-noise ratio.

These limitations may introduce attenuation bias, whereby classical measurement error biases coefficient estimates toward zero. Consequently, non-significant findings should be interpreted cautiously, as they may reflect either genuine absence of effects or reduced statistical power due to imperfect measurement. Future research may benefit from triangulating NLP-based indicators with survey-based measures or behavioral purchase data.

To test the proposed hypotheses, fixed-effects panel regression models were employed. Fixed-effects estimation controls for time-invariant unobserved heterogeneity across brand and country units (Wooldridge, 2010). Given the nested structure of the data, the models incorporate brand fixed effects, country fixed effects, and week fixed effects to account for brand-specific characteristics, country-specific institutional environments, and common temporal shocks (Table 1).

Prior to hypothesis testing, correlation analysis was conducted to examine bivariate relationships among the main variables. Table 2 presents the correlation matrix with corresponding significance levels. Brand loyalty is strongly correlated with brand trust, providing preliminary support for the theoretical association between these constructs. Engagement shows weak but statistically significant correlations with both trust and loyalty, indicating potential relationships that warrant further examination under multivariate controls.

High correlations among certain country-level macro indicators are absorbed by fixed effects and do not raise multicollinearity concerns for the regression models. Overall, the correlation patterns suggest no severe multicollinearity issues among the primary time-varying predictors.

The core finding of this study is that none of the hypothesized relationships achieved statistical significance. All coefficients for the engagement and trust variables in the main regression models are non-significant at conventional levels (p > 0.10). This section reports these null results in detail.

Table 3 presents the fixed-effects panel regression results examining the relationship between social media engagement, brand trust, and brand loyalty. Model 1 tests the direct short-term effect of lagged engagement on brand loyalty. As shown in column (1), the estimated coefficient of lagged engagement is small in magnitude (β = 0.0209) and statistically insignificant (SE = 0.1398, p = 0.883). This finding indicates that week-to-week fluctuations in engagement intensity do not produce immediate changes in brand loyalty outcomes. Accordingly, Hypothesis 1 is not supported.

To examine potential nonlinear dynamics, Model 2 incorporates both linear and quadratic engagement terms. Neither the linear term (β = −0.3364, SE = 0.3624, p = 0.363) nor the quadratic term (β = 3.0031, SE = 3.0679, p = 0.337) reaches conventional significance levels. These results suggest that the engagement–loyalty relationship does not exhibit detectable diminishing returns, saturation effects, or other curvilinear patterns within the observed engagement range. Instead, the estimated relationship remains statistically indistinguishable from zero across engagement levels. Thus, Hypothesis 2 is not supported.

Model 3 evaluates whether engagement predicts brand trust. The coefficient of lagged engagement is negative but extremely small (β = −0.0106) and statistically insignificant (SE = 0.1686, p = 0.950). This result suggests that short-term variation in engagement does not systematically influence trust-related sentiment expressed in consumer comments. Therefore, Hypothesis 3a is not supported.

Model 4 examines the association between brand trust and brand loyalty. The estimated effect of lagged trust on loyalty is also close to zero (β = −0.0037, SE = 0.0394, p = 0.926). This finding indicates that short-term changes in trust-related expressions do not translate into contemporaneous loyalty-related responses at the weekly level. Hypothesis 3b is therefore not supported.

Model 5 jointly includes engagement and trust predictors. The coefficients remain statistically insignificant and similar in magnitude to the previous specifications, suggesting that the absence of effects is not driven by omitted variable bias within the proposed mediation framework.

To further examine the potential nonlinear pattern implied by Model 2, additional diagnostic statistics are reported in Table 4. The estimated turning point occurs at an engagement value of 0.056 (log scale), which is close to the sample mean. The marginal slope is negative at low engagement levels and positive at higher levels. However, because both the linear and quadratic coefficients are statistically insignificant, this pattern cannot be distinguished from a flat relationship. These diagnostics reinforce the conclusion that engagement does not exhibit a meaningful curvilinear association with brand loyalty in the short term.

Although the estimated coefficients are statistically insignificant, graphical representations are presented to illustrate the predicted patterns implied by the regression models. Figure 2 displays predicted brand loyalty values across the observed engagement range based on Model 2, with control variables held constant. The fitted curve exhibits a shallow U-shaped pattern, but the 95% confidence intervals are wide and overlap substantially across engagement levels, indicating high estimation uncertainty and limited practical variation in predicted loyalty.

Figure 3 presents a scatter plot of the raw data with a quadratic fitted line. The dispersion of observations around the fitted curve is substantial, and no systematic trend is visually apparent.

Figure 4 reports marginal effects of engagement on loyalty across the engagement distribution. The marginal effects remain close to zero and statistically insignificant throughout the entire range, with confidence intervals consistently encompassing zero. Taken together, these visual diagnostics further corroborate the regression results that short-term engagement does not exert a meaningful effect on brand loyalty.

Formal mediation tests were conducted following the Baron and Kenny (1986) procedure and supplemented with Sobel tests and bootstrap confidence intervals. Table 5 reports the mediation results. The estimated indirect effect of engagement on loyalty via brand trust is effectively zero (0.00008). The Sobel test statistic is close to zero (z = 0.01, p = 0.992), and the bootstrap 95% confidence interval includes zero (−0.0158, 0.0160).

These findings provide no statistical evidence that brand trust mediates the relationship between social media engagement and brand loyalty within the short-term observation window. Accordingly, Hypothesis 4 is not supported.

The indirect effect is essentially zero (0.00008). The Sobel test is non-significant (z = 0.01, p = 0.992). The bootstrap confidence interval includes zero (CI: −0.0158, 0.0160). H4 is not supported.

To assess the robustness of the main findings, alternative engagement specifications and estimation approaches were examined. Table 6 reports results using cumulative engagement measures based on 4-week and 8-week rolling averages. Both coefficients remain statistically insignificant (β = 0.0312, p = 0.809 for the 4-week measure; β = 0.0489, p = 0.752 for the 8-week measure), suggesting that the null results are not driven by the one-week lag specification.

Table 7 presents results from alternative panel estimators, including random-effects and between-effects models. Across all specifications, engagement coefficients remain statistically insignificant. The Hausman test indicates no systematic difference between fixed-effects and random-effects estimates (χ² = 12.34, p = 0.263), although the fixed-effects specification is retained based on theoretical considerations regarding unobserved heterogeneity. Overall, the robustness analyses confirm that the absence of statistically significant engagement effects is stable across alternative model specifications and measurement approaches.

Table 8 summarizes the results of all hypothesis tests. None of the proposed hypotheses reach conventional significance levels. This consistent pattern of null findings persists across baseline models, nonlinear specifications, mediation tests, and robustness checks, indicating a systematic absence of short-term engagement effects on brand loyalty within the studied context.

All five hypotheses failed to achieve statistical significance. This consistent pattern of null results persists across the main models and robustness checks.

This study examined the short-term relationship between social media engagement and brand loyalty in emerging markets, with particular attention to the mediating role of brand trust. Across all model specifications, the empirical results consistently show that social media engagement does not exert a statistically significant effect on brand loyalty. Neither linear nor nonlinear engagement terms are significant, and brand trust does not mediate the engagement–loyalty relationship within the weekly observation window. These findings remain robust across alternative lag structures, cumulative engagement measures, and different panel estimation approaches. Together, the results suggest a systematic absence of short-term engagement effects on loyalty-related outcomes in the studied context, rather than isolated model-specific anomalies.

The findings carry several important theoretical implications for social media marketing and consumer–brand relationship research.

First, the results challenge the widely held assumption that higher engagement automatically translates into stronger brand loyalty. Much of the existing literature implicitly treats engagement metrics—such as likes, comments, and shares—as proxies for relationship capital. However, the absence of detectable short-term effects suggests that observable engagement behavior may not reflect substantive psychological attachment or commitment. Engagement volume alone may capture surface-level interaction rather than deeper relational bonds.

Second, the findings highlight a temporal mismatch between engagement dynamics and relationship formation processes. Engagement fluctuates at high frequency, often responding to platform algorithms and content characteristics, whereas trust and loyalty are slow-moving psychological constructs that typically develop through accumulated experiences over longer horizons. Weekly engagement variation may therefore represent short-term noise rather than meaningful relationship-building signals. This temporal misalignment helps explain why short-term engagement shocks fail to produce immediate loyalty responses.

Third, the results point to potential boundary conditions of the Stimulus–Organism–Response (S-O-R) framework in high-frequency digital contexts. While S-O-R theory assumes relatively direct stimulus-to-response pathways mediated by internal states, the present findings suggest that such mechanisms may operate over extended time scales rather than immediate weekly cycles. The framework may require temporal extensions to account for cumulative exposure effects and delayed psychological adaptation in digital brand environments.

Together, these theoretical implications suggest the need to move beyond static or instantaneous interpretations of engagement effectiveness and toward dynamic models that explicitly incorporate time horizons and relationship accumulation processes.

From a methodological perspective, the findings underscore the importance of using panel data designs with appropriate controls for unobserved heterogeneity. Many prior studies rely on cross-sectional survey data or short-term experiments, which may overestimate engagement effects by conflating brand-specific characteristics with engagement outcomes. The use of brand, country, and time fixed effects in this study provides a more stringent test of engagement effectiveness by isolating within-unit variation over time.

In addition, the results highlight potential limitations of NLP-based sentiment measures for capturing complex psychological constructs such as trust and loyalty. While text-based indicators offer scalability and behavioral realism, they may suffer from attenuation bias due to imperfect classification and low base rates of explicit attitudinal expressions. Future research may benefit from triangulating NLP-derived indicators with survey-based measures or transactional purchase data to improve construct validity.

Finally, the high-frequency nature of social media data raises important methodological challenges. Weekly variation may be too granular to capture slow-moving consumer attitudes. Researchers should carefully align temporal measurement scales with the theoretical time horizons of the constructs under investigation. Longer observation windows and cumulative exposure measures may be better suited to studying relationship-building processes.

The findings also carry important implications for marketing practitioners operating in emerging markets. First, the results suggest that engagement volume alone should not be interpreted as a direct indicator of relationship success. Metrics such as likes and comments may reflect short-term attention rather than durable brand commitment. Overreliance on engagement-based key performance indicators (KPIs) may therefore lead to overly optimistic assessments of campaign effectiveness.

Second, managers should adopt a longer-term perspective when evaluating social media performance. Relationship-building outcomes such as trust and loyalty are unlikely to respond immediately to short-term engagement fluctuations. Strategic consistency, cumulative exposure, and sustained brand communication efforts may be more important than short-term spikes in interaction metrics.

Finally, the findings imply that emerging market consumers may engage with brand content in ways that are more entertainment-oriented or symbolic rather than relational. Firms may need to complement engagement-oriented tactics with deeper relationship-building strategies, such as customer service integration, community management, and offline–online coordination, to translate digital interactions into durable brand equity.

This study examined the short-term relationship between social media engagement and brand loyalty in emerging markets, with a particular focus on the potential mediating role of brand trust. Using a multi-country panel dataset and fixed-effects regression models, the empirical analysis consistently shows that social media engagement does not exert a statistically significant short-term effect on brand loyalty.

In addition, the results indicate that brand trust neither responds significantly to short-term engagement fluctuations nor mediates the engagement–loyalty relationship within the weekly observation window. These findings remain robust across nonlinear specifications, cumulative engagement measures, and alternative estimation approaches. Together, the evidence points to a systematic absence of short-term engagement effects rather than isolated model-specific anomalies.

The findings carry important implications for understanding the role of social media engagement in brand relationship building. First, the results suggest that observable engagement metrics do not necessarily represent genuine relationship capital. Likes, comments, and shares may reflect momentary attention or entertainment-driven interaction rather than durable psychological attachment or behavioral commitment.

Second, the findings highlight the importance of time horizons in evaluating engagement effectiveness. Trust and loyalty are slow-moving psychological constructs that typically develop through accumulated experiences. Short-term engagement variation may therefore be insufficient to trigger measurable changes in these outcomes. This temporal mismatch helps explain why engagement shocks observed at high frequency fail to translate into immediate loyalty responses.

Together, these insights call for greater theoretical attention to dynamic relationship formation processes and caution against interpreting short-term engagement indicators as direct measures of brand relationship strength.

Importantly, this study does not claim that social media engagement is ineffective in general. Instead, the findings demonstrate that short-term fluctuations in engagement do not generate immediate loyalty responses in emerging market contexts. Engagement may still contribute to long-term brand equity through cumulative exposure, repeated interaction, and sustained communication strategies. However, such long-run effects fall outside the temporal scope of the present analysis.

By clearly distinguishing between short-term null effects and potential long-term relationship-building mechanisms, this study avoids overgeneralization and provides a more nuanced interpretation of engagement effectiveness.

Several directions for future research emerge from the present findings. First, future studies should employ longer observation windows, such as multi-year panels, to capture cumulative engagement effects and delayed relationship formation processes.

Second, methodological triangulation is needed to improve measurement validity. Combining NLP-based sentiment indicators with survey-based trust and loyalty measures or transactional purchase data would allow researchers to assess whether null short-term effects persist across alternative measurement approaches.

Third, platform heterogeneity warrants further investigation. Engagement dynamics and algorithmic exposure mechanisms differ substantially across platforms such as Instagram, TikTok, Facebook, and regional social media services. Comparative platform-level analyses would help clarify whether the observed patterns are platform-specific or reflect broader digital marketing dynamics.

Finally, future research should examine cultural and institutional moderators in emerging markets, including consumer media usage norms and market maturity, to better understand boundary conditions of engagement effectiveness.