The Effect of Environmental, Social, and Governance (ESG) on the Persistence of Firm Value: Evidence from Survival Analysis

Abstract

This study examines the effect of environmental, social, and governance (ESG) performance on the persistence of firm value among publicly listed companies in Taiwan from 2016 to 2023, using survival analysis. This approach addresses a gap in the literature, which has largely overlooked the temporal dimension of firm value. The findings indicate that only higher social scores are significantly associated with a longer duration of firm value persistence, whereas environmental and governance scores do not exhibit this effect. Furthermore, the analysis reveals that within the social pillar, only product quality and safety contribute meaningfully to sustaining firm value. Although previous studies have often linked sustainability practices to higher firm value, the present findings suggest that such effects may not endure over time. These results underscore the importance of aligning ESG initiatives with core business strategies and enhancing disclosure credibility to ensure authentic commitment.

1. Introduction

Responsible investment has emerged as a focal topic within the financial sector in recent years [1,2]. As the global awareness of sustainable development continues to grow, investors are increasingly inclined to consider environmental, social, and governance (ESG) performance in their decision-making. This investment approach aims not only to achieve financial returns but also to deliver positive societal and environmental outcomes [3,4]. Consequently, an expanding body of the literature has examined the relationship between ESG performance and firm value [5], including analyses of each ESG dimension individually—environmental (E) [6], social (S) [7], and governance (G) [8]. ESG-related effects have been shown to operate through several channels, including the cost of debt [9,10], systematic risk [11], cash flows [12], and corporate growth [13], all of which may ultimately influence firm value.

Despite these contributions, much of the existing literature relies on panel data and regression techniques such as ordinary least squares (OLS), weighted least squares, generalized least squares, lasso, and ridge regression. While these approaches are effective in estimating correlations and average effects, they are limited in evaluating how long firms can sustain elevated firm value levels. For instance, OLS primarily captures the magnitude of firm value at a given point in time, rather than its continuity. In addition, it is unsuitable for time-to-event data, which are essential for assessing the persistence of firm value. Whether an outcome improves and how long it persists represent two fundamentally distinct analytical questions [14,15]. Furthermore, OLS does not account for censored observations, introducing potential bias when firms do not experience a value decline within the observation period.

To address these limitations, this study applies survival analysis—a method commonly used in biomedical and economic research but seldom utilized in ESG-related studies—to examine the temporal aspect of firm value persistence. Survival analysis is particularly well suited for assessing how long a firm maintains a certain condition, in this case, firm value persistence, before it experiences a failure event, defined as a decline in value.

This study departs from traditional approaches to firm value by focusing on the persistence of firm value. In this context, firm value is defined as the overall market-based value of a company, rather than an isolated outcome attributable solely to ESG performance. The persistence of firm value is measured as the number of consecutive years in which a firm’s value exceeds that of the previous year. Using survival analysis, this study investigates whether higher environmental, social, and governance performance is associated with a longer duration of firm value persistence among publicly listed firms in Taiwan between 2016 and 2023. Taiwan provides a compelling research setting due to its recognized leadership in ESG evaluation within the Asia–Pacific region, as acknowledged by ISS ESG [16]. Moreover, the ESG scores used in this study are sourced from the Taiwan Economic Journal (TEJ), which compiles ESG ratings authorized and certified by the Sustainability Accounting Standards Board (SASB) [17], thereby ensuring data credibility and consistency.

The findings indicate that while overall ESG scores are not significantly related to firm value persistence, firms with higher S Scores demonstrate significantly longer durations—especially in aspects related to product quality and safety. In addition, control variables show that stronger financial performance and heightened investor sentiment are associated with extended firm value persistence. These findings contribute to the literature by addressing a critical gap in understanding the temporal dimension of firm value and highlight the need to consider duration-based perspectives. The academic contribution of this study lies in introducing a survival analysis perspective—commonly seen in medical and economic research but seldom in ESG-related studies—to underscore time-oriented inquiry. Practically, it underlines that the causal link between ESG actions and extended firm value remains unclear. Companies should align ESG investments with their core operations to maximize benefits, and regulators may need to guard against greenwashing to ensure stakeholders can identify genuine ESG commitments.

The remainder of this paper is organized as follows: Section 2 reviews the relevant literature and presents the research hypotheses; Section 3 describes the data and methodology; Section 4 reports the empirical results; Section 5 conducts additional tests and sensitivity analyses, including assessments of endogeneity; Section 6 discusses the findings and their implications; and Section 7 concludes the study.

2. Literature Review and Hypothesis Development

2.1. Expanding Perspectives on Firm Value

The process of determining a firm’s value has traditionally centered on financial metrics derived from accounting data and market information [18,19]. Measures such as earnings, cash flows, and asset-based valuations have long been regarded as reliable indicators of a company’s underlying worth and growth potential [5,20]. However, as business environments grow more complex, scholars and practitioners increasingly recognize that firm value extends beyond these conventional metrics [21]. Recent studies emphasize the growing importance of intangible elements—such as corporate reputation, innovation capacity, stakeholder relationships, and strategic positioning—in shaping investor perceptions of firm performance [2,22].

In parallel with this broader perspective, growing attention has been directed toward ESG performance [23]. These dimensions assess how responsibly firms manage environmental impacts [24], how ethically they engage with employees, customers, and communities [25], and how effectively they implement governance structures that ensure accountability and transparency [26]. As companies operate in increasingly globalized and sustainability-oriented markets, ESG considerations may influence current value assessments and also the capacity of firms to preserve or enhance that value over time [5,27]. These studies therefore encourage a more holistic view of firm value, incorporating both financial fundamentals and the broader societal and environmental contexts in which firms operate.

2.2. Positive Perspectives Linking ESG and Firm Value

Multiple theoretical frameworks and cross-national empirical findings support the view that ESG performance contributes positively to firm value. A foundational perspective derives from stakeholder theory [28], which emphasizes that firms must address the interests of a broad spectrum of stakeholders—including customers, employees, suppliers, and communities—to improve overall firm value. By adopting responsible social initiatives, implementing ethical sourcing, and minimizing environmental harm, firms can reduce reputational risks, strengthen stakeholder trust, and enhance legitimacy [1,27]. These benefits, in turn, foster investor confidence and mitigate value volatility.

From the resource-based view, ESG capabilities are seen as strategic assets that are valuable, rare, and difficult to imitate [29]. Firms that integrate ESG into their core operations can improve resource efficiency, enhance risk management, and differentiate themselves from competitors. Effective governance mechanisms also help mitigate agency conflicts and reduce information asymmetries [8], supporting firm value enhancement.

Empirical evidence across multiple regions corroborates these theoretical claims. In developed markets such as the United States and the United Kingdom, ESG performance has been consistently associated with higher firm value, particularly in relation to governance indicators and firm value metrics such as Tobin’s Q [30]. Similarly, a multi-country analysis covering 34 countries found that sustainability disclosures were positively associated with firm value, though their impact on operating performance was more nuanced [31].

In China, while ESG scores correlate with firm value, their impact varies across industries, highlighting the importance of sector-specific ESG integration [32]. Other studies also suggest that firm size and growth option value may moderate the ESG–value relationship, as seen in India, the US, and China [33,34].

Regional governance conditions further influence the strength of ESG’s impact. In countries with higher regulatory quality and political stability, ESG practices are more likely to translate into financial benefits [35]. Additionally, cross-regional comparisons show variation in causal mechanisms: for instance, environmental performance tends to yield stronger value effects in European contexts, while social and governance factors are more influential in Anglo-American systems [36].

Finally, industry heterogeneity plays a decisive role. In environmentally sensitive sectors, ESG performance—especially environmental initiatives—tends to enhance profitability and firm value. In contrast, in sectors such as agriculture and utilities, governance and social factors have demonstrated stronger effects [32].

In summary, both theoretical reasoning and empirical evidence converge on the view that ESG performance can increase firm value. However, the strength and nature of this relationship are shaped by firm-specific attributes, industry characteristics, ownership structures, and national governance contexts. These insights underscore the importance of tailoring ESG strategies to organizational and institutional environments to maximize firm value benefits.

2.3. Negative or Neutral Perspectives Linking ESG and Firm Value

Not all scholars concur that ESG investments invariably lead to higher firm value. Critics apply agency theory to argue that executives might employ ESG initiatives as symbolic or superficial gestures—referred to as “greenwashing”—to bolster personal reputations or divert shareholder scrutiny without delivering meaningful performance improvements [4,37,38]. Under such conditions, ESG commitments may amount to non-essential spending or managerial entrenchment rather than authentic value enhancement.

Furthermore, ESG endeavors can entail significant initial costs and extended time horizons before yielding financial returns. Investors with short-term horizons may be skeptical, thus assigning little immediate value premium [6]. In highly competitive industries or under permissive regulatory environments, ESG efforts may fail to confer a clear advantage if customers or investors focus predominantly on price and earnings. In such cases, ESG investments may dilute earnings and make the firm less appealing to valuation frameworks that prioritize near-term results.

Recent empirical studies further indicate that the firm value impact of ESG is highly contextual. For example, while ESG performance generally correlates positively with firm value in developed markets such as the European Union [39], the relationship is more nuanced or even negative in emerging economies. In China and Southeast Asia, institutional weaknesses, short-term investment priorities, and less stringent regulatory environments have been shown to attenuate or reverse ESG’s value effects [40,41]. Similarly, firms in the Asia–Pacific region exhibit a mixed relationship, with country-specific institutional conditions and investor sophistication playing moderating roles [42].

Industry also plays a pivotal role. Companies in environmentally sensitive sectors often face heightened ESG-related risks, with controversies leading to pronounced value erosion, particularly in developed economies with stronger enforcement [43]. Conversely, in sectors where ESG risks are less visible or quantifiable—such as agriculture or retail—the value impact may be neutral or negligible.

Governance structures further condition the ESG–value nexus. High ownership concentration, especially prevalent in emerging markets, can intensify agency conflicts and undermine the intended benefits of ESG disclosures [41]. By contrast, firms with more transparent reporting and independent boards tend to mitigate the adverse effects of ESG controversies and improve stakeholder trust [44,45].

Taken together, these findings suggest that ESG’s effect on firm value is neither universally positive nor consistent across contexts. It depends on a complex interplay of regional, industry, and firm-level governance factors. What may enhance firm value in one setting could be ineffective—or even detrimental—in another. Therefore, nuanced interpretation is essential when assessing the implications of ESG strategies on financial performance.

2.4. Mixed Empirical Evidence and Methodological Constraints

Numerous studies exploring the ESG–firm value relationship have produced mixed outcomes, reflecting the theoretical ambivalence noted above. Some find that firms with stronger ESG performance enjoy higher firm value, often attributed to enhanced risk management, reputational gains, and stronger stakeholder relations [7,46]. Others fail to detect a significant association or even uncover a negative link, indicating that ESG initiatives may not always be positively received by the market [4].

These divergent findings suggest that the ESG–firm value link is highly context-specific and may evolve over time. Sector-specific factors, regulatory dynamics, market maturity, and investor preferences mediate how ESG credentials affect firm value. In addition, most empirical approaches face methodological challenges. Much of the literature uses cross-sectional or panel regressions, estimating average ESG effects on firm value at particular points in time. Although insightful, these methods generally do not incorporate the temporal dimension of value, such as how long firms maintain certain levels of firm value or whether ESG performance influences its persistence.

Therefore, to address the literature’s inability to determine how ESG might affect the period over which firm value is sustained, the present study employs survival analysis—an approach rarely utilized in ESG research. Building on the predominantly positive findings in the literature regarding the ESG–firm value relationship, this study extends the analysis to a temporal dimension and proposes the following hypothesis:

Hypothesis: ESG performance is positively associated with the persistence of firm value.

3. Data and Methods

3.1. Data Source and Study Period

This study focuses on Taiwanese publicly listed companies from 2016 to 2023, with all data obtained from the TEJ database. The main reason for beginning in 2016 is that TEJ started to provide comprehensive ESG scores from that year onward, thereby ensuring a consistent observation period.

3.2. Survival Analysis Framework

Survival analysis was employed to assess the persistence of firm value among these companies. Firm value was proxied by the Tobin’s Q ratio (TQ ratio), reflecting each firm’s value over the 2016–2023 period. The observation for each firm began when its value first exceeded that of the previous year. “Survival time” was defined as the length of time a firm continues to exceed its yearly value from the prior period. “Failure” was noted when a firm’s value fell below that of the previous year, an occurrence referred to as “the event”.

To address right-censoring, the study’s final sample included instances in which a failure event had not occurred by the end of 2023, thereby reducing potential bias in survival analysis. This approach, which accommodates censored observations, is a hallmark of survival analysis methods. Figure 1 visually outlines the methodology, illustrating the observation period for six representative cases. Among these, cases 1 and 2 are censored, indicating that no value decline occurred by 2023. The right side of the figure demonstrates how these cases were processed for survival analysis, clarifying the computation of survival time.

3.3. Cox Regression with a Time-Dependent Covariate

This study employed a Cox regression with a time-dependent covariate, a semi-parametric model that accommodates time-varying covariates and benefits from a non-parametric baseline hazard function. The primary rationale for using Cox regression with a time-dependent covariate lies in the limitations of traditional parametric models. Parametric distribution models are typically estimated via OLS, but OLS is unsuitable for censored data, which is inherent to survival analysis.

Instead, this study utilized Akaike’s Information Criterion (AIC) and maximum likelihood estimation (MLE) to determine the optimal distribution type. AIC was applied to compare model fit by evaluating log-likelihood estimates; while a higher log-likelihood indicates a better fit, the model with the smallest AIC value is considered the most appropriate [47]. Although the log-normal distribution initially appeared to provide a plausible fit, further testing using MLE and likelihood ratio statistics resulted in its rejection.

Based on these outcomes, the analysis adopted Cox regression with a time-dependent covariate because it does not require assumptions regarding the specific distribution of the baseline hazard function and can accommodate the time-varying nature of control variables. This flexibility allows the baseline hazard to take any form—fixed, increasing, or decreasing—according to the data. Although the Cox model does not directly estimate the baseline hazard function, necessitating additional methods if such estimates are required, its adaptability makes it well suited for examining the temporal dynamics of firm value persistence.

3.4. Covariates and Statistical Model

A Cox regression with a time-dependent covariate model was used to evaluate the influence of ESG scores on the persistence of firm value, controlling for financial performance, investor sentiment, the COVID-19 pandemic, and firm size. Initially, a broad set of potential covariates was considered based on the existing literature. A stepwise Cox regression narrowed the model to the most impactful control variables: financial performance, investor sentiment, COVID-19, and firm size. Specifically, earnings and firm size capture internal financial and operational characteristics, while investor sentiment and COVID-19 reflect external market and environmental conditions. The model is represented as Equation (1):

h(t;X(t)) = h₀(t)⋅exp(α₁X_ESG(t) + α₂X_EPS(t) + α₃X_Sentiment(t) + α₄X_Covid-19(t) + α₅X_Size(t))

(1)

where h(t;X(t)) represents the hazard function that describes the instantaneous risk of an event occurring at time t, given the covariates X(t). h₀(t) is the baseline hazard function representing the hazard when all covariates are zero, and exp(.) is the exponential function, which ensures that the hazard function h(t;X(t)) is always positive.

The ESG scores (X_ESG), which range from 0 to 100, are provided by TEJ. TEJ’s methodology integrates data from more than 20 public sources—including annual shareholder meetings, the Market Observation Post System, and official company websites—aligning with SASB guidelines. This variable serves as the primary explanatory variable in this study and is used to test the proposed hypothesis.

Financial performance (X_EPS) is incorporated as a control variable, given its significant influence on firm value [48]. It is measured using a dummy coded 1 if the earnings per share (EPS) in a given year exceed those of the previous year and 0 otherwise.

Investor sentiment (X_Sentiment), another key determinant of firm value, is proxied by the annual average stock turnover rate for each firm, computed by averaging daily turnover rates. This measure captures investor confidence and interest, which are essential contributors to market value [22].

COVID-19 (X_Covid-19) is also incorporated as a control, represented by a dummy coded 1 for the years 2019 and 2020 (the peak pandemic years) and 0 for other years. This variable isolates pandemic-related disruptions and their impact on firm value [49].

Lastly, firm size (X_Size) is controlled by taking the natural logarithm of total assets, addressing disparities in size across firms.

The final sample for this analysis comprised 1637 firms across multiple industries, as detailed in

Table 1. Distribution of final sample firms by SASB’s main industry classification.

SASB Main Industry	Original Sample Firms	Firms with Missing Data	Final Sample Firms	%
Technology and Communications	661	11	650	40%
Resource Transformation	292	7	285	17%
Consumer Goods	188	4	184	11%
Public Infrastructure	115	2	113	7%
Healthcare	103	8	95	6%
Transportation	84	1	83	5%
Extraction and Mineral Processing	72	0	72	4%
Services	48	2	46	3%
Financials	45	0	45	3%
Food and Beverages	45	0	45	3%
Renewable Resources and Alternative Energy	21	2	19	1%
Total	1674	37	1637	100%

.

Table 2. Distribution of final sample firm-years by SASB main industry classification.

SASB Main Industry	2016	2017	2018	2019	2020	2021	2022	2023	Total	%
Technology and Communications	293	472	400	146	135	98	67	56	1667	39%
Resource Transformation	145	201	166	54	48	38	30	33	715	17%
Consumer Goods	59	107	91	65	63	41	34	24	484	11%
Public Infrastructure	53	85	69	37	33	20	12	11	320	7%
Healthcare	11	34	34	32	41	30	23	20	225	5%
Transportation	34	58	44	18	20	16	12	6	208	5%
Extraction and Mineral Processing	39	57	49	18	9	10	10	7	199	5%
Financials	28	41	42	16	15	4	3	0	149	3%
Food and Beverages	23	27	24	18	16	12	11	9	140	3%
Services	12	24	20	13	11	11	10	11	112	3%
Renewable Resources and Alternative Energy	6	12	10	6	5	5	2	2	48	1%
Total	703	1118	949	423	396	285	214	179	4267	100%

Note: The analysis includes data from 1637 firms over 4267 observation periods.

presents the distribution of firm-years across the final sample, categorized by the SASB’s main industry classification. For instance, in the technology and communications industry, 293 firms entered the observation in 2016, defined as the point at which a firm’s value first exceeded that of the preceding year. By 2017, some firms continued meeting the survival conditions, whereas others experienced the event—indicated by a firm value decline relative to the prior year—and new firms entered, bringing the total number of firms to 472 in 2017. This pattern persisted annually, and by 2023, 56 firms remained under observation without experiencing an event.

4. Results

4.1. Descriptive Statistics

Table 3. Descriptive statistics.

Variable	Mean	Std	Min	25%	50%	75%	Max
ESG Scores	54.95	7.39	30.46	49.83	54.56	59.74	79.53
E Scores	54.89	10.94	26.70	46.11	53.19	62.43	90.41
S Scores	55.52	10.35	27.71	47.82	54.87	62.64	88.08
G Scores	54.59	10.38	22.14	47.48	55.08	61.79	84.41
EPS	0.55	0.50	0.00	0.00	1.00	1.00	1.00
Investor Sentiment	0.86	1.50	0.00	0.14	0.36	0.96	23.84
COVID-19	0.16	0.37	0.00	0.00	0.00	0.00	1.00
Size	15.69	1.60	11.43	14.62	15.38	16.38	22.95

Note: The analysis includes data from 1637 firms over 4267 observation periods.

presents descriptive statistics for this study, calculated at the firm-year level rather than the individual-firm level. The range of ESG and their components, spanning from 22.14 to 90.41, demonstrates considerable variation throughout the sample. Additionally, the mean EPS is 0.55, indicating that in 55% of firm-year observations, the current year’s EPS exceeded that of the previous year. The mean COVID-19 value of 0.16 signifies that 16% of the firm-year observations occurred during the primary pandemic years.

4.2. Survivorship Curve

Figure 2 presents survival curves stratified by industry, with firms’ ESG scores categorized into tertiles within their respective SASB classifications. These curves illustrate survival probabilities for firms in the highest (top third) and lowest (bottom third) ESG tiers, revealing differences in the persistence of firm value across industries. The 95% confidence intervals are shown as shaded areas, providing a visual depiction of the precision surrounding each estimated survival probability.

A log-rank test was applied to determine whether survival probabilities differed significantly between high and low ESG groups. The results indicate that only the “food and beverages” industry exhibited a statistically significant difference (p-value < 0.05). In contrast, no other industries demonstrated a significant difference, implying that the influence of ESG quality on the persistence of firm value is not universal across all sectors but rather appears in specific industries.

4.3. Results from Time-Dependent Cox Regression Analysis

Table 4. Time-dependent Cox regression of ESG scores on the persistence of firm value.

Covariate	Coef	Exp(coef)	95% CI (Lower, Upper)	z-Value	p-Value
ESG Scores	−0.003	0.997	(0.990, 1.005)	−0.765	0.444
EPS	−0.408	0.665	(0.599, 0.739)	−7.616	0.000
Investor Sentiment	−0.155	0.856	(0.802, 0.914)	−4.655	0.000
COVID-19	−0.022	0.978	(0.929, 1.030)	−0.846	0.397
Size	−0.092	0.912	(0.863, 0.965)	−3.198	0.001
Model Fit	Partial log-likelihood −9263.52 Partial AIC 18,537.03

Note: 1. The analysis includes data from 1637 firms, with 1467 of these firms experiencing the event. 2. The event is defined as the point at which a firm’s value becomes lower than that of the preceding year.

presents the empirical results for Model (1). In Table 4, “Coef” refers to the coefficient: a positive value suggests an increased risk of the event occurring, whereas a negative value indicates a decreased risk. In this analysis, the event is defined as a decrease in a firm’s value compared to the previous year. The hazard ratio, shown as “exp(coef)”, interprets values greater than 1 as indicating a higher risk of the event, while values less than 1 suggest a lower risk. The 95% confidence interval (CI) (lower and upper) offers a range for the hazard ratio; if the interval includes 1, the covariate’s effect on the risk is not statistically significant. Statistical significance is determined by the z-value and p-value, with p < 0.05 denoting statistical significance. The Cox time-dependent survival analysis was performed on data from 1637 firms across 4267 firm-year observations, identifying 1467 events in which firm value declined relative to the prior year.

In Table 4, the coefficient for ESG is −0.003, and the hazard ratio is 0.997. With a p-value of 0.444, this effect is not statistically significant, indicating that no substantial evidence supports a correlation between ESG scores and the persistence of firm value.

Regarding other control variables, the results for EPS (hazard ratio = 0.665, p < 0.001), Sentiment (hazard ratio = 0.856, p < 0.001), and Size (hazard ratio = 0.912, p = 0.001) imply that firms with improving financial performance, higher investor sentiment, and a larger size tend to sustain their value for a longer period. This observation indicates that better operating performance and greater investor attention are conducive to prolonged firm value.

In contrast, the hazard ratio for COVID-19 is 0.978 (p = 0.397) and is thus not statistically significant. One possible explanation for this finding is that over 39% of the sample belongs to the “technology and communications” industry, which—along with remote interactions and economic easing policies—may have mitigated negative pandemic impacts.

Given that this study primarily investigates the effect of ESG scores, the results do not demonstrate a significant influence on the persistence of firm value. In light of these findings, Section 4 disaggregates ESG into its individual components (E, S, and G) and introduces additional tests to examine their respective impacts on the persistence of firm value more precisely.

5. Additional Test and Sensitivity Analysis

5.1. Detailed Analysis of ESG Component Scores

In this part of the study, the E, S, and G component scores replace the original ESG variable in Model (1) to further investigate which elements potentially influence the persistence of firm value. Model (2) is specified as follows:

h(t;X(t)) = h₀(t)⋅exp(ß₁X_E(t) + ß₂X_S(t) + ß₃X_G(t) + ß₄X_EPS(t) + ß₅X_Sentiment(t)
+ ß₆X_Covid-19(t) + ß₇X_Size(t))

(2)

Table 5. Time-dependent Cox regression of ESG component scores on the persistence of firm value.

Covariate	Coef	Exp(coef)	95% CI (Lower, Upper)	z-Value	p-Value
E Scores	0.026	1.027	(0.968, 1.088)	0.882	0.378
S Scores	−0.074	0.928	(0.874, 0.987)	−2.391	0.017
G Scores	0.015	1.015	(0.964, 1.069)	0.573	0.566
EPS	−0.408	0.665	(0.599, 0.739)	−7.619	0.000
Investor Sentiment	−0.156	0.855	(0.801, 0.913)	−4.687	0.000
COVID-19	−0.023	0.977	(0.928, 1.028)	−0.895	0.371
Size	−0.077	0.926	(0.873, 0.983)	−2.537	0.011
Model Fit	Partial log-likelihood −9260.88 Partial AIC 18,535.76

Note: 1. The analysis includes data from 1637 firms, with 1467 of these firms experiencing the event. 2. The event is defined as the point at which a firm’s value becomes lower than that of the preceding year.

shows the empirical results for Model (2). Among these, only the S Scores exhibit a statistically significant impact on the persistence of firm value (hazard ratio = 0.928; p = 0.017). This suggests that firms with higher S Scores are more likely to sustain their value over a longer period. Conversely, the E Scores and G Scores do not show a significant effect, and the other control variables remain consistent with the primary findings.

5.2. Converting ESG Components from Scores into Ranking Percentiles

To bolster the robustness of Section 5.1, which examines the effect of E, S, and G Scores on the persistence of firm value, E, S, and G component scores are converted into industry-specific percentile rankings in accordance with SASB classifications.

For example, consider a firm (Firm A) in the technology and communications industry. Its S Scores in 2019 are ranked 22nd among 146 firms in that industry (as indicated by the distribution in Table 2). Firm A’s S Scores ranking percentile is thus 15% (22/146). A lower percentile indicates stronger S performance relative to industry peers, whereas a higher percentile signals weaker performance.

After converting these ESG components into ranking percentiles, the data are tested again using Model (2), with percentile ranks replacing the original ESG component scores.

Table 6. Time-dependent Cox regression of ESG component percentile rankings on the persistence of firm value.

Covariate	Coef	Exp(coef)	95% CI (Lower, Upper)	z-Value	p-Value
E Scores	−0.027	0.974	(0.794, 1.194)	−0.255	0.799
S Scores	0.204	1.226	(0.997, 1.508)	1.932	0.053
G Scores	−0.047	0.954	(0.796, 1.144)	−0.510	0.610
EPS	−0.407	0.665	(0.599, 0.739)	−7.611	0.000
Investor Sentiment	−0.154	0.857	(0.803, 0.915)	−4.641	0.000
COVID-19	−0.023	0.978	(0.929, 1.029)	−0.866	0.387
Size	−0.080	0.923	(0.871, 0.977)	−2.745	0.006
Model Fit	Partial log-likelihood −9261.86 Partial AIC 18,537.72

Note: 1. The analysis includes data from 1637 firms, with 1467 of these firms experiencing the event. 2. The event is defined as the point at which a firm’s value becomes lower than that of the preceding year.

presents the time-dependent survival analysis results associated with these ranking percentiles. The key outcome is that only the S Score ranking percentile shows marginal statistical significance (hazard ratio = 1.226, p = 0.053). This finding implies that as the S Scores ranking percentile increases—indicating comparatively weaker social performance within the industry—the likelihood of the event, defined as a decline in firm value, increases. Other results remain consistent with earlier analyses. Hence, the main conclusion persists: firms with more favorable S Scores rankings demonstrate a stronger persistence of firm value, while E and G rankings do not yield significant effects.

5.3. Further Analysis of S Component Factors

Section 5.1 reveals that firms with superior S Scores maintain their firm value for a longer period. Consequently, additional analysis explores which specific sub-dimensions of S most influence the persistence of firm value. The S component comprises five factors: human rights and community relations (S_HRCR), data security (S_DS), product quality and safety (S_PQS), employee information disclosure (S_EID), and employee health and safety (S_EHS). These five sub-scores replace the aggregated S Score in Model (2), yielding Model (3):

h(t;X(t)) = h₀(t)⋅exp(ɣ₁X_E(t) + ɣ₂X_{S_HRCR}(t) + ɣ₃X_{S_DS}(t) + ɣ₄X_{S_PQS}(t) + ɣ₅X_{S_EID}(t) + ɣ₆X_{S_EHS}(t) + ɣ₇X_G(t) + ɣ₈X_EPS(t) + ɣ₉X_Sentiment(t) + ɣ₁₀X_Covid-19(t) + ɣ₁₁X_Size(t))

(3)

Table 7. Time-dependent Cox regression of ESG component sub-scores on the persistence of firm value.

Covariate	Coef	Exp(coef)	95% CI (Lower, Upper)	z-Value	p-Value
E Scores	0.038	1.038	(0.978, 1.103)	1.222	0.222
S HRCR Scores	−0.031	0.970	(0.918, 1.025)	−1.087	0.277
S DS Scores	0.001	1.001	(0.950, 1.056)	0.050	0.960
S PQS Scores	−0.054	0.948	(0.898, 1.000)	−1.961	0.050
S EID Scores	−0.022	0.978	(0.926, 1.034)	−0.783	0.434
S EHS Scores	−0.052	0.950	(0.890, 1.013)	−1.573	0.116
G Scores	0.015	1.015	(0.964, 1.070)	0.577	0.564
EPS	−0.407	0.666	(0.599, 0.740)	−7.574	0.000
Investor Sentiment	−0.152	0.859	(0.805, 0.917)	−4.581	0.000
COVID-19	−0.023	0.977	(0.928, 1.029)	−0.882	0.378
Size	−0.073	0.929	(0.870, 0.992)	−2.204	0.028
Model Fit	Partial log-likelihood −9259.26 Partial AIC 18,540.53

Note: 1. The analysis includes data from 1637 firms, with 1467 of these firms experiencing the event. 2. The event is defined as the point at which a firm’s value becomes lower than that of the preceding year.

provides the time-dependent survival analysis results for Model (3). Among the five S sub-dimensions, only product quality and safety (S_PQS) is statistically significant (hazard ratio = 0.948; p = 0.050), indicating that higher product quality and safety scores extend the persistence of firm value. By contrast, the remaining four dimensions—human rights and community relations, data security, employee information disclosure, and employee health and safety—do not show evidence of a relationship with value persistence.

5.4. The Moderating Effect of Firm Size on ESG and the Persistence of Firm Value

Section 5.1 reveals that firm size (Size) exhibits a significant positive correlation with the persistence of firm value. To further explore the moderating role of firm size in the relationship between ESG performance and the persistence of firm value, the full sample was ranked in descending order by total assets, and the top and bottom one-third of firms were extracted for separate analysis. Model (2) was then applied to these two subsamples, and the results are presented in

Table 8. Time-dependent Cox regression of ESG components on the persistence of firm value by firm size.

Covariate	Top One-Third Size Exp(coef)	p-Value	Bottom One-Third Size Exp(coef)	p-Value
E Scores	1.054	0.211	0.931	0.314
S Scores	0.891	0.014	0.905	0.035
G Scores	1.036	0.424	0.979	0.655
EPS	0.579	0.000	0.756	0.004
Investor Sentiment	0.927	0.271	0.795	0.000
COVID-19	0.938	0.204	0.980	0.649
Size	0.960	0.421	1.007	0.966
Model Fit	Partial log-likelihood −2718.14 Partial AIC 5450.29		Partial log-likelihood −2340.90 Partial AIC 4695.80

Note: 1. The top one-third Size subsample consists of 546 firms, with 517 experiencing the event. 2. The bottom one-third Size subsample consists of 546 firms, with 455 experiencing the event. 3. The event is defined as the point at which a firm’s value becomes lower than that of the preceding year.

.

The results in Table 8 indicate that, in contrast to the findings from the full sample, firm size (Size) is no longer statistically significant in either the top or bottom one-third subsamples. This discrepancy arises because the size effect observed in the full sample is primarily driven by cross-group variation—where firms with substantially different sizes exhibit distinct patterns in the persistence of firm value. However, once firms are grouped by size, the within-group variation decreases, thereby reducing the statistical power of Size as an explanatory variable. This finding suggests that within more homogeneous subsamples, other firm-specific characteristics—such as ESG practices—may exert a stronger influence on the persistence of firm value than firm size itself. This is consistent with the prior literature [50], which notes that size effects tend to be more pronounced in cross-sectional analyses but become less relevant when stratified by firm size (e.g., segmented or subsample-based studies).

Despite the lack of significance for Size, the primary variable of interest—S Scores—remains statistically significant in both the top and bottom one-third asset subsamples. Specifically, for the largest firms, S Scores exhibit a hazard ratio of 0.891 (p = 0.014), indicating that firms with stronger social performance exhibit a greater persistence of firm value. Similarly, for the smallest firms, S Scores remain significant with a hazard ratio of 0.905 (p = 0.035), reinforcing the robustness of the social dimension’s influence across different firm sizes.

Notably, neither E Scores nor G Scores achieve statistical significance in either group, mirroring the findings in Section 5.1. This suggests that the influence of social performance on the persistence of firm value holds consistently across different firm sizes, whereas the effects of the environmental and governance dimensions remain inconclusive.

5.5. Lagged Variables to Address Endogeneity

To mitigate potential endogeneity and enhance robustness, a sensitivity analysis was performed by lagging the main explanatory variables by one period. This procedure reduces the risk of reverse causality, as the lagged variables reflect information from the prior period, making them less influenced by the current period’s outcome variable. Moreover, lagging helps capture delayed effects of explanatory variables on the persistence of firm value, thereby improving causal interpretation and result stability.

In the adjusted models, ESG, EPS, Sentiment, and Size are replaced with their lagged values from t − 1, while COVID-19 remains unlagged since it represents a contemporaneous event. Accordingly, Model (1) and Model (2) are reformulated as Model (4) and Model (5):

h(t;X(t)) = h₀(t)⋅exp(α₁X_ESG(t − 1) + α₂X_EPS(t − 1) + α₃X_Sentiment(t − 1) + α₄X_Covid-19(t)
+ α₅X_Size(t − 1))

(4)

h(t;X(t)) = h₀(t)⋅exp(ß₁X_E(t − 1) + ß₂X_S(t − 1) + ß₃X_G(t − 1) + ß₄X_EPS(t − 1) + ß₅X_Sentiment(t − 1) + ß₆X_Covid-19(t) + ß₇X_Size(t − 1))

(5)

The results from Model (4) and Model (5) remain consistent with the primary findings: only S Scores are significantly associated with greater persistence of firm value, whereas ESG, E, and G Scores do not show significant effects.

6. Discussion

The empirical results presented in Section 4 and Section 5 indicate that overall ESG scores are not significantly associated with the persistence of firm value. However, among the three ESG components, only the S dimension demonstrates a significant and robust effect in extending the period over which firm value is maintained. This finding stands in contrast to prior studies that generally report a positive static association between ESG scores and firm value [27,51]. Those studies often conclude that ESG engagement enhances transparency and mitigates risk, thereby attracting investor confidence and increasing firm value. For instance, ESG disclosures have been shown to serve as credible signals of corporate strategy, indirectly influencing firm value through heightened stakeholder trust [27]. Similarly, effective ESG performance has been found to moderate the relationship between enterprise risk management and firm value [51].

However, the current study, through the application of survival analysis, emphasizes the temporal dimension—specifically, how long a firm can maintain value above the prior year—rather than simply whether ESG performance improves value. This distinction reflects the view that whether performance improves and how long it persists are analytically and conceptually distinct issues [14,15]. Static methods may overestimate the implications of ESG if they ignore how firm value evolves over time. By adopting a persistence-based framework, this study highlights a critical risk overlooked in the prior literature: short-term ESG effects may not guarantee sustained value creation.

A key contribution of this research lies in the differentiated effect of ESG components. While E and G Scores do not exhibit significant relationships with the persistence of firm value, the S Score is positively associated with it. This finding supports the angel halo effect [46], whereby firms recognized for social responsibility gain intangible goodwill that contributes to prolonged investor confidence. For investors, this underscores the importance of monitoring firms’ social practices when evaluating the continuity of value generation.

To explore the mechanism underlying this social effect, the S component was further disaggregated into five sub-dimensions. Among these, only product quality and safety (PQS) demonstrates a statistically significant association with the persistence of firm value. This suggests that social initiatives that directly affect end consumers—such as ensuring product safety—are more readily perceived and rewarded by the market, possibly due to their direct impact on revenue streams. In contrast, other sub-dimensions such as human rights and community relations, data security, employee disclosure, and workplace safety, while socially desirable, do not exert a measurable influence on value persistence. These findings highlight that ESG investments lacking immediate and visible impact may be less likely to be capitalized into sustained firm value.

In contrast, the environmental component typically involves long-term investments and high upfront costs, with benefits that may only materialize over extended horizons. As such, firms focused on near-term outcomes may underinvest in environmental initiatives, even if these are strategically sound. This underscores the need for firms to adopt multi-dimensional performance metrics and extended evaluation windows to align short-term actions with broader sustainability goals.

Governance practices, while essential for corporate accountability, often arise from regulatory compliance rather than a voluntary strategic initiative. As such, their effects may be perceived as baseline expectations rather than signals of value creation, thereby limiting their impact on the persistence of firm value.

While the data are drawn from Taiwanese firms, the findings offer implications that extend beyond the local context. As reviewed in Section 2.2 and Section 2.3, the ESG–firm value relationship varies significantly across countries, shaped by institutional environments, regulatory rigor, and cultural expectations. For instance, governance and environmental components are more influential in developed economies such as the US and UK [30,36], whereas social performance may resonate more in contexts with heightened consumer scrutiny. Conversely, in emerging markets such as China and Southeast Asia, weaker regulatory frameworks and short-term investment horizons may limit or even reverse ESG’s value-enhancing effects [40,41,42]. The persistence-based findings of this study contribute to this debate by showing that even in a regionally prominent ESG market like Taiwan [16], only specific ESG attributes—particularly social initiatives tied to product integrity—translate into sustained firm value.

In sum, this study highlights that the relationship between ESG performance and the persistence of firm value is not uniformly positive across dimensions but is instead driven by the distinct effect of social performance. Firms are advised to integrate ESG strategies into their core operations to enhance their relevance and visibility to stakeholders. For example, a food packaging company could prioritize the upstream adoption of sustainable materials—such as plant-based biodegradable plastics—instead of relying solely on downstream efforts like coastal clean-ups. Initiatives that are closely aligned with operational outcomes and deliver observable value are more likely to generate lasting firm value benefits.

Nevertheless, the challenge of distinguishing authentic ESG practices from greenwashing persists. Investors may find it difficult to assess whether ESG disclosures reflect substantive commitments or superficial compliance. Policymakers are therefore encouraged to implement stronger verification mechanisms to promote transparency and assist stakeholders in identifying firms that genuinely commit to sustainability.

Finally, while survival analysis offers a valuable approach for examining the temporal dimension of firm value, its application entails both strengths and limitations. It excels in handling time-to-event data and accommodating censored observations, making it particularly suitable for evaluating how long a firm maintains a certain level of firm value. However, since survival models focus on timing rather than magnitude, they may fail to capture short-term fluctuations in firm value. Given that traditional regression methods are well established in the literature for analyzing the magnitude of value changes, combining survival analysis with such approaches can provide a more comprehensive understanding—capturing both the persistence and the scale of environmental, social, and governance influences on firm performance.

7. Conclusions

This study contributes to the literature by employing a survival analysis framework to investigate the relationship between ESG performance and the persistence of firm value. Unlike conventional methods that evaluate ESG’s impact at a single point in time, survival analysis captures the temporal aspect of firm value, offering deeper insights into whether ESG-driven benefits endure over time.

The findings reveal that while overall ESG scores do not significantly affect the persistence of firm value, the social dimension—especially product quality and safety—plays a key role in extending the period over which firm value is maintained. In contrast, the environmental and governance components do not exhibit consistent effects. These results highlight the complexity of ESG’s role in firm value and suggest that not all ESG investments contribute equally to sustaining firm value over time.

Though based on a Taiwanese context, this study has broader implications. The methodology and insights can be applied in international settings to explore whether similar patterns hold in different institutional and regulatory environments. Future research is encouraged to extend the current model by incorporating alternative firm value metrics, broader ESG indicators, and cross-country comparisons.

Ultimately, this research highlights the utility of time-oriented methods for capturing the persistence of firm value and the differentiated impact of ESG strategies. By shifting the focus from whether ESG improves firm value to how long such value is sustained, the study provides a meaningful contribution to both academic discourse and strategic decision-making in sustainability-oriented investment.