Digital Economy Policies and Fisher Household Vulnerability: Evidence from China’s Fisheries Sector

Abstract

The formulation of effective digital economy policies is crucial for advancing high-quality industrial development in the digital era. This study employs panel data from 29 Chinese provinces over the period of 2011–2023 and adopts a Difference-in-Differences (DID) approach to empirically assess the impact of digital economy policies on fisher household vulnerability. The results show that these policies significantly alleviate household vulnerability and enhance risk resilience. Mechanism analysis reveals that digital economy policies reduce fisher vulnerability mainly through enhancing fishery digital enabling capacity and fisher digital literacy. Fishery digital enabling capacity significantly promotes information access, industrial digitalization, and operational efficiency within the fishery sector, while fisher digital literacy further strengthens individual adaptive capacity. Further heterogeneity analysis reveals that while digital economy policies significantly reduce fisher household vulnerability in both coastal and inland provinces, the effects are stronger in coastal regions. In addition, the policies exhibit a stronger vulnerability mitigation effect in areas with higher policy intensity relative to those with lower intensity. These results underscore the importance of reinforcing digital economy policy frameworks, enhancing policy coordination and intensity, and strengthening digital capacity-building to mitigate fisher household vulnerability, improve adaptive resilience, and foster the sustainable and high-quality development of the fisheries sector.

1. Introduction

Oceans cover approximately 71 percent of the Earth’s surface, far exceeding the 29 percent occupied by land. As terrestrial resources become increasingly depleted and the marginal returns on land-based development decline, the ocean is emerging as a critical frontier for resource extraction and economic development. In this context, marine fisheries play a central role in the marine economy, making significant contributions to food security, regional development, and sustainable growth. China’s aquatic industry has experienced rapid expansion, with total aquatic product output reached 73.66 million tons in 2024, 15.9 times that of 1978, maintaining its position as the world’s largest producer for 35 consecutive years. In the current era of digitalization, developing the digital economy and promoting industrial digital transformation have become national strategic priorities. As early as February 2019, the policy documents issued by China’s Ministry of Agriculture and Rural Affairs explicitly emphasized the need to integrate modern information technologies, such as big data and artificial intelligence, with aquaculture and to promote the development of digital fishery demonstration projects. The development of digital fishery models grounded in advanced technologies such as cloud computing and big data plays a critical role in modernizing the fisheries industry. This approach not only promotes industrial upgrading and enhances the efficiency of resource utilization but also injects strong momentum into the implementation of China’s maritime power strategy and the revitalization of rural fishing ports.

Fisheries constitute a vital pillar of China’s agricultural sector and national economy, and the development of fishing households directly influences the quality and sustainability of marine economic growth. However, fishing households remain highly vulnerable to external shocks such as market fluctuations, natural disasters, and technological limitations. Enhancing their resilience and improving their livelihood conditions has thus become a key priority in promoting the modernization of marine fisheries and advancing the goal of common prosperity. This study employs a DID approach to examine the impact of digital economy policies on the vulnerability of fishing households. It aims to assess policy effectiveness, identify digital empowerment mechanisms within the fishery sector, and enrich the theoretical framework linking the digital economy and fisheries. The findings offer empirical support for promoting fishery modernization and enhancing the welfare of fishing communities. In addition, the findings support the broader application of new-generation information technologies in the fisheries sector. This helps strengthen the digital capacity of the industrial chain and promotes the localized implementation of the Digital China strategy. Furthermore, the results provide empirical evidence for the formulation of effective digitalization policies in fisheries, facilitate the accurate identification and mitigation of fishing household vulnerability, and offer practical insights into promoting high-quality development of the fishery economy and advancing China’s national marine strategy.

2. Literature Review

The concept of the digital economy was first introduced by Bowman in 1996. Since then, its evolution has generally followed three main stages: the information economy, the internet economy, and the new economy (Moulton, 1999 [1]; Turcan and Juho, 2014 [2]). Definitions of the digital economy have varied across historical periods, with early studies primarily emphasizing the role of digital technologies in enhancing productivity. As research deepened, scholarly attention gradually shifted toward the broader economic functions of digital technologies and their transformative impact on production relations. Despite growing interest, a universally accepted definition of the digital economy has yet to emerge. From an international perspective, definitions of the digital economy differ across countries. China, South Korea, and Russia generally conceptualize it as a type of economic activity, although each emphasizes different aspects (Chen et al., 2022 [3]). In contrast, countries such as the United States and France tend to define the digital economy through the lens of measurement and statistical frameworks. As the digital economy continues to expand rapidly, establishing effective methods for assessing its development has become a focal point for both academic researchers and policymakers. At present, two primary approaches are widely used to measure the digital economy: the index-based method and the value-added method. The index-based method constructs multidimensional evaluation frameworks to generate composite indices that reflect the overall level of digital economy development (Atkinson and Nager, 2014 [4]; OECD, 2014 [5]; Graham et al., 2017 [6]; Wang et al., 2021 [7]). This approach has become one of the most widely adopted and academically recognized tools. In contrast, the value-added method emphasizes calculating the share of value added by the digital economy and related industries in the overall economy in order to assess their contribution to economic growth (Xiang and Wu, 2019 [8]; Xu and Zhang, 2020 [9]).

The concept of vulnerability has been interpreted from multiple disciplinary perspectives. The World Bank defines vulnerability as the probability of exposure to risks and the likelihood that such shocks will lead to a decline in livelihood conditions below a critical minimum threshold. The World Food Programme (WFP) introduced an analytical framework in 1995 for assessing rural vulnerability. This framework includes three core dimensions: risk exposure, coping capacity, and the effectiveness of social support systems. Moser (1998) conceptualizes vulnerability as the result of heightened livelihood risks stemming from inadequate asset endowments at the individual, household, or community level [10]. Focusing specifically on farming households, Li et al. (2007) define vulnerability as the probability of encountering shocks and the corresponding actual and potential capacity to respond and adapt effectively [11].

Existing theoretical frameworks and conceptual definitions provide a solid foundation for assessing the vulnerability of fishing households. International scholarship in this area is relatively advanced, with sustained attention given to the combined effects of external shocks and internal structural dynamics on household livelihoods. Vulnerability is commonly understood as the result of the interaction between exposure to external stressors and the capacity of households to respond and adapt (Linnekamp et al., 2011 [12]). The Livelihood Vulnerability Index (LVI) has been widely employed as an analytical tool in empirical studies. For example, Ahsan et al. (2014) and Bennett et al. (2015) applied the LVI to examine how fishing households responded to various risk scenarios [13,14]. In contrast, research on fishing household vulnerability in China remains limited. Existing studies have predominantly focused on inland farming and pastoral households, with relatively few empirical investigations targeting fishing communities. Against the backdrop of the national strategies of building a Digital China and promoting maritime power, exploring the vulnerability of fishing households and pioneering the analysis of how the digital economy policies affect such vulnerability carries both significant theoretical value and urgent practical relevance.

3. Theoretical Analytical Framework

3.1. Theoretical Foundations

This study adopts the Sustainable Livelihoods Framework (SLF) and Digital Transformation Theory (DTT) as the core analytical foundations to construct a theoretical framework for examining how the digital economy affects the livelihood vulnerability of fishing households.

The SLF, initially proposed by Chambers and Conway (1992) and systematically developed by Scoones (1998) [15,16], conceptualizes livelihood systems as comprising three interrelated elements: livelihood assets, livelihood strategies, and livelihood outcomes. The framework emphasizes that the sustainability of individual or household livelihoods depends on access to and the effective utilization of five types of capital, namely human capital, natural capital, physical capital, financial capital, and social capital, within the constraints imposed by external institutional and environmental factors (Scoones, 1998 [16]). For fisher households, vulnerability primarily manifests as the erosion of livelihood capital, restrictions on livelihood strategies, and reduced adaptive capacity in response to multifaceted shocks from resources, markets, policies, and environmental changes (Allison and Ellis, 2001 [17]). Meanwhile, Digital Transformation Theory highlights the role of information technology in reshaping economic structures and social relations. The proliferation of digital infrastructure, data flows, and intelligent systems can fundamentally alter patterns of resource allocation and value creation (Brennen and Kreiss, 2016 [18]).

Recent empirical studies examining the coupling coordination between the digital economy and ecological resilience in China have demonstrated that digital development can improve overall system coordination; however, notable regional disparities remain (Huang et al., 2025 [19]). These findings imply that the effects of digitalization on household livelihood vulnerability are likely to vary across regions. Building on these insights, this study integrates the Sustainable Livelihoods Framework and Digital Transformation Theory to investigate how the digital economy may mitigate the livelihood vulnerability of fishing households by reshaping livelihood capital structures and influencing their livelihood strategy choices.

3.2. Theoretical Pathways of Digital Economy Affecting Fisher Household Vulnerability

By integrating the Sustainable Livelihoods Framework with Digital Transformation Theory, digital economy policies can be conceptualized as external enabling factors that influence the vulnerability of fisher households by reshaping both their livelihood assets and strategies. Specifically, digital interventions enhance livelihood assets by improving human capital through access to digital skills and information, optimizing natural and physical capital via intelligent monitoring and advanced equipment, expanding financial capital through digital finance and insurance services, and strengthening social capital through online networks and community platforms. These enhancements collectively increase households’ capacity to withstand environmental, market, and policy-related shocks. Concurrently, digital economy initiatives shape livelihood strategies by facilitating income diversification, promoting more effective risk management practices, and enabling industrial upgrading across the fisheries value chain. The synergistic effect of strengthened assets and optimized strategies reduces households’ exposure and sensitivity to external shocks, thereby mitigating vulnerability. Moreover, the adaptive capacity of households is enhanced, enabling proactive responses to emerging risks and supporting the maintenance of stable and resilient livelihoods over time.

In the process through which digital economy policies influence fisher household vulnerability, digital capacity plays a critical mediating role and can be conceptualized at two levels. First, Fishery Digital Enabling Capacity (FDEC) represents an industry-level mediator that captures the extent of digital support, industrial environment, technology extension, and infrastructure within the fisheries sector. Improvements in Fishery Digital Enabling Capacity enhance households’ livelihood assets, including natural, physical, financial, and social capital, providing the necessary conditions for accessing information, utilizing technology, and implementing livelihood strategies, thereby indirectly reducing vulnerability.

Second, fisher digital literacy (FDL) serves as an individual-level mediator, bridging the implementation of policies and the resulting behavioral changes. FDL reflects fishers’ capacity to acquire, comprehend, and apply digital knowledge in both production and daily life. Higher levels of digital literacy enable fishers to use digital resources and technologies more effectively, ensuring that the impact of digital economy policies is fully transmitted through the optimization of livelihood assets and strategies and directly contributing to the reduction in vulnerability.

Overall, digital economy policies reduce fisher household vulnerability by enhancing livelihood assets and optimizing livelihood strategies(see Figure 1). Fishery Digital Enabling Capacity and fisher digital literacy operate at the industry and individual levels, respectively, jointly promoting household resilience. This theoretical framework provides a rigorous foundation for empirical analysis and clarifies the mechanisms through which digital interventions support sustainable and high-quality development in fisheries livelihoods.

4. Research Design

This study selects 2016 as the policy intervention point for the digital economy, aiming to evaluate the impact of digital economy policies on the vulnerability of fishing households. The primary rationale lies in the fact that 2016 marked the first time China systematically introduced a series of national-level policy documents to promote the development of the digital economy. This year is widely recognized as the beginning of China’s digital economy era, signifying the elevation of digital economy development to a national strategic priority. The sample includes 31 provincial-level regions in mainland China, excluding Hong Kong, Macao, and Taiwan. Due to significant data gaps, Qinghai and Tibet are excluded, resulting in a final sample of 29 provinces. The study period spans from 2011 to 2023, providing a sufficient time frame for empirical analysis and enabling a comprehensive evaluation of policy impacts before and after the implementation of digital economy policies.

4.1. Variable Description

4.1.1. Dependent Variable

To systematically and rigorously assess the vulnerability of fishing households, this study adopts a three-dimensional framework encompassing exposure, sensitivity, and adaptive capacity, consistent with widely recognized approaches to vulnerability assessment (IPCC, 2014 [20]). These dimensions capture the key processes through which households experience, respond to, and recover from external shocks. Specifically, exposure reflects the extent to which fishing households are subject to environmental and economic disturbances; sensitivity measures the degree to which their livelihoods depend on vulnerable resources or structures; and adaptive capacity captures the ability of households to absorb, cope with, and recover from adverse impacts through access to resources, technology, and institutional support.

Based on this conceptual framework and data availability, a set of representative indicators was selected to reflect the multidimensional nature of vulnerability while ensuring objectivity and comparability across provinces. The rationale for each dimension and its indicators is detailed below. Descriptive statistics (mean and standard deviation) are reported in

Table A1. Annual descriptive statistics of fisher household vulnerability sub-indicators under the exposure dimension.

Year	E1 Disaster-Affected Population Density		E2 Direct Economic Loss Rate from Natural Disasters		E3 Direct Economic Loss Rate from Fishery Disasters		E4 Disaster-Affected Rate of Aquaculture Area		E5 Loss Rate of Aquatic Products
	Mean	Std. Dev.	Mean	Std. Dev.	Mean	Std. Dev.	Mean	Std. Dev.	Mean	Std. Dev.
2011	0.31	0.24	0.01	0.01	0.05	0.05	0.19	0.19	0.05	0.07
2012	0.21	0.12	0.01	0.01	0.03	0.03	0.13	0.10	0.03	0.03
2013	0.28	0.18	0.01	0.02	0.14	0.39	0.11	0.11	0.03	0.04
2014	0.20	0.14	0.01	0.01	0.02	0.02	0.09	0.11	0.02	0.02
2015	0.14	0.08	0.00	0.00	0.02	0.02	0.07	0.07	0.01	0.01
2016	0.16	0.12	0.01	0.01	0.02	0.03	0.10	0.12	0.02	0.03
2017	0.11	0.09	0.01	0.01	0.01	0.02	0.08	0.10	0.01	0.02
2018	0.10	0.07	0.00	0.01	0.02	0.03	0.06	0.06	0.01	0.01
2019	0.09	0.08	0.00	0.00	0.01	0.01	0.07	0.10	0.01	0.02
2020	0.11	0.08	0.01	0.01	0.01	0.02	0.08	0.11	0.02	0.03
2021	0.08	0.07	0.00	0.00	0.02	0.03	0.06	0.06	0.01	0.02
2022	0.08	0.08	0.00	0.00	0.01	0.01	0.06	0.07	0.01	0.01
2023	0.06	0.06	0.00	0.01	0.02	0.08	0.04	0.04	0.01	0.02

,

Table A2. Annual descriptive statistics of fisher household vulnerability sub-indicators under the sensitivity dimension.

Year	S1 Share of Regional Fishery Output in GDP		S2 per Capita Aquaculture Area		S3 Volatility of Cumulative Year-on-Year GDP Growth Rate		S4 Fishery-Dependent Adjusted Income Index		S5 Share of Aquaculture Output in Total Production *
	Mean	Std. Dev.	Mean	Std. Dev.	Mean	Std. Dev.	Mean	Std. Dev.	Mean	Std. Dev.
2011	0.01	0.02	54.46	45.67	0.38	0.31	0.85	0.08	0.81	0.17
2012	0.01	0.02	56.21	48.18	0.33	0.20	0.84	0.08	0.82	0.17
2013	0.02	0.02	57.85	52.72	0.22	0.19	0.83	0.10	0.82	0.17
2014	0.02	0.02	58.03	52.76	0.29	0.23	0.85	0.10	0.82	0.18
2015	0.01	0.02	58.52	53.44	0.32	0.31	0.85	0.09	0.82	0.18
2016	0.01	0.02	51.41	45.51	0.27	0.28	0.85	0.08	0.82	0.18
2017	0.01	0.02	51.93	45.94	0.40	0.46	0.85	0.08	0.83	0.17
2018	0.01	0.02	50.61	45.49	0.32	0.24	0.86	0.08	0.84	0.18
2019	0.01	0.01	49.78	45.16	0.32	0.20	0.85	0.07	0.83	0.18
2020	0.01	0.01	49.12	45.24	3.96	2.24	0.85	0.08	0.85	0.18
2021	0.01	0.01	49.32	46.80	4.40	3.10	0.86	0.07	0.85	0.19
2022	0.01	0.01	50.04	48.15	1.23	0.82	0.87	0.07	0.86	0.18
2023	0.01	0.01	53.31	54.38	0.84	0.51	0.86	0.08	0.86	0.18

Note: Indicators marked with an * are negative indicators, meaning that the larger the value, the lower the related level of the corresponding major indicator.

and

Table A3. Annual descriptive statistics of fisher household vulnerability sub-indicators under the adaptive capacity dimension.

Year	A1 Quantity of Aquatic Seedlings *		A2 Public Information Services for Fisheries *		A3 Specialization Rate of Fishery Workers *
	Mean	Std. Dev.	Mean	Std. Dev.	Mean	Std. Dev.
2011	16,035.78	57,602.74	342,684.10	572,240.67	0.55	0.14
2012	17,252.53	60,731.59	347,150.86	539,535.42	0.55	0.14
2013	20,414.80	64,495.78	341,784.86	498,040.33	0.55	0.14
2014	22,794.02	71,817.48	321,828.24	511,809.62	0.55	0.14
2015	27,471.70	86,486.40	325,287.03	526,006.54	0.55	0.13
2016	31,263.74	98,855.89	497,463.34	590,478.31	0.55	0.13
2017	45,037.66	158,209.53	476,195.45	665,858.69	0.55	0.12
2018	44,727.84	124,987.19	407,316.07	488,915.06	0.55	0.12
2019	39,878.50	114,852.50	335,598.28	395,687.25	0.55	0.14
2020	40,643.55	116,837.21	359,242.90	458,438.64	0.56	0.21
2021	34,710.15	105,170.20	319,904.24	373,717.65	0.55	0.15
2022	45,932.31	148,010.62	347,420.69	438,824.42	0.55	0.15
2023	46,463.38	140,012.87	466,063.45	1,012,104.13	0.55	0.15

Note: Indicators marked with an * are negative indicators, meaning that the larger the value, the lower the related level of the corresponding major indicator.

in Appendix A.

(1) 

Exposure Dimension

The exposure dimension measures the direct impact of natural disasters and economic shocks on households and production. The selected indicators include: (1) Disaster-Affected Population Density (E1): reflects the scale of the population affected and the potential disruption of livelihoods. (2) Direct Economic Loss Rate from Natural Disasters (E2): measures the severity of economic impacts at the regional level. (3) Direct Economic Loss Rate from Fishery Disasters (E3): captures the exposure of fishery production to external shocks. (4) Disaster-Affected Rate of Aquaculture Area (E4): indicates physical damage to aquaculture and associated economic risk. (5) Loss Rate of Aquatic Products (E5): directly reflects income disruption and short-term vulnerability. These indicators collectively quantify exposure from population, economic, and sectoral perspectives.

(2) 

Sensitivity Dimension

The sensitivity dimension reflects the degree to which households and regions are susceptible to external shocks. The selected indicators are: (1) Share of Regional Fishery Output in GDP (S1): a higher dependence on fisheries indicates greater vulnerability. (2) Per Capita Aquaculture Area (S2): represents production scale and potential exposure per capita. (3) Volatility of Cumulative Year-on-Year GDP Growth Rate (S3): captures macroeconomic stability; higher volatility increases sensitivity. (4) Fishery-Dependent Adjusted Income Index (S4): measures household income dependence on fisheries; higher values indicate greater sensitivity. (5) Share of Aquaculture Output in Total Production (S5): reflects production structure and its influence on vulnerability. These indicators assess how regional economic structure, production scale, and income reliance shape households’ responses to shocks.

(3) 

Adaptive Capacity Dimension

The adaptive capacity dimension measures the ability of households and regions to respond to and recover from shocks. The selected indicators are: (1) Quantity of Aquatic Seedlings (A1): a sufficient supply of fry enhances recovery and production flexibility. (2) Public Information Services for Fisheries (A2): access to information services improves risk awareness and adaptive responses. (3) Specialization Rate of Fishery Workers (A3): higher specialization supports technology adoption and effective risk management. These indicators capture the multidimensional nature of adaptive capacity, including productive resources, information access, and human capital, all of which reduce vulnerability and strengthen livelihood resilience.

This study adopts the Entropy-TOPSIS method to comprehensively and objectively evaluate the vulnerability of fishing households. This approach first determines the weights of each indicator using the entropy method and then calculates the composite vulnerability scores for each province using the Technique for Order Preference by Similarity to Ideal Solution (TOPSIS). The assessment was conducted using Stata 18.0, and the indicator system along with the entropy-based weights is presented in

Table 1. Indicator system for fishing household vulnerability variables.

Primary Indicator	Secondary Indicator	Indicator Composition	Weights
Exposure	E1 Disaster-Affected Population Density	Ratio of Natural Disaster-Affected Population to Resident Population	6.45%
	E2 Direct Economic Loss Rate from Natural Disasters	Ratio of Direct Economic Loss from Natural Disasters to Regional GDP	9.65%
	E3 Direct Economic Loss Rate from Fishery Disasters	Ratio of Total Direct Economic Loss to Regional Total Fishery Output Value	21.55%
	E4 Disaster-Affected Rate of Aquaculture Area	Ratio of Disaster-Affected Aquaculture Area to Total Aquaculture Area	9.30%
	E5 Loss Rate of Aquatic Products	Ratio of Aquatic Product Loss to Total Aquatic Product Output	11.21%
Sensitivity	S1 Share of Regional Fishery Output in GDP	Ratio of Regional Fishery Output Value to Regional GDP	9.13%
	S2 Per Capita Aquaculture Area	Ratio of Aquaculture Area to Resident Population	5.93%
	S3 Volatility of Cumulative Year-on-Year GDP Growth Rate	Standard Deviation of the Cumulative Year-on-Year GDP Growth Rate	13.02%
	S4 Fishery-Dependent Adjusted Income Index	Proportion of Income from Aquatic Product Sales and Fishery Wages in Total Annual Income per Capita Net Income of Fishermen	4.78%
	S5 Share of Aquaculture Output in Total Production *	Ratio of Aquaculture Product Subtotal to Total Aquatic Product Output	7.71%
Adaptive Capacity	A1 Quantity of Aquatic Seedlings *	Fish Fry Output (Freshwater & Marine)	0.23%
	A2 Public Information Services for Fisheries *	Websites + Mobile Info Users/Information Coverage Users + Telephone Hotlines/Public Information Releases + Journals + Materials/Distributed Technical Materials	0.13%
	A3 Specialization Rate of Fishery Workers *	Ratio of Professional Practitioners to Total Fishery Workers	0.90%

Note: Indicators marked with an * are negative indicators, meaning that the larger the value, the lower the related level of the corresponding major indicator. In the subsequent calculation of the comprehensive fishers’ vulnerability index, these values are entered with a negative sign. Missing values are imputed using the nearest neighbor method.

. The data were primarily sourced from provincial statistical bureaus, the National Bureau of Statistics, and the China Fisheries Statistical Yearbook. The specific calculation steps are as follows:

The specific calculation steps are as follows.

Step 1: Indicator Standardization. Given the heterogeneity in scale and direction across indicators, the original data are first normalized to ensure comparability. Appropriate standardization formulas are applied separately for positively and negatively oriented indicators.

Positive indicators:

$$z_{ij}={\displaystyle \frac{x_{ij}-min\left(x_j\right)}{max\left(x_j\right)-min\left(x_j\right)}}$$

Negative Indicators:

$$z_{ij}={\displaystyle \frac{max\left(x_j\right)-x_{ij}}{max\left(x_j\right)-min\left(x_j\right)}}$$

Specifically, represents the observed value of the j-th indicator for the i-th sample, and denotes the corresponding normalized value.

Step 2: Entropy Calculation and Weight Determination. After standardization, the objective weight of each indicator is determined based on information entropy theory.

Calculation of indicator weights:

$$p_{ij}={\displaystyle \frac{z_{ij}}{{\sum }_{i=1}^n{z}_{ij}}}$$

Calculation of information entropy for each indicator:

$$e_j=-k{\sum }_{i=1}^n{p}_{ij}\mathit{ln}\left(p_{ij}\right), \mathrm{where} k={\displaystyle \frac{1}{ln\left(n\right)}}$$

Calculation of redundancy and weights:

$$d_j=1-e_j, w_j={\displaystyle \frac{d_j}{{\sum }_{j=1}^m{d}_j}}$$

Step 3: Construction of Positive and Negative Ideal Solutions. Upon determining the weights for each indicator, the next step involves calculating the distance of each sample from the ideal solutions representing the best and worst possible states. This process begins with the formation of a weighted normalized decision matrix, followed by the determination of the positive ideal solution (the maximum value of each indicator) and the negative ideal solution (the minimum value of each indicator).

Step 4: Calculation of Distances and Composite Scores. The Euclidean distances between each evaluated object and the positive and negative ideal solutions are calculated separately.

Distance to the positive ideal solution:

$$D_i^{+}=\sqrt{\sum _{j=1}^m{w}_j{(z_{ij}-z_j^{+})}^2}$$

Distance to the negative ideal solution:

$$D_i^{-}=\sqrt{\sum _{j=1}^m{w}_j{(z_{ij}-z_j^{-})}^2}$$

Subsequently, the relative closeness of each sample is calculated and used as the final vulnerability score:

$${FV}_i={\displaystyle \frac{D_i^{-}}{D_i^{+}+D_i^{-}}}, 0\le C_i\le 1$$

A higher value of FV_i indicates greater vulnerability of fishing households.

4.1.2. Core Explanatory Variable

The core explanatory variable in this study is the digital economy policy, represented by the interaction term between the policy group dummy variable treated_i and the policy timing dummy variable time_t. Following the methodologies of Jin et al. (2022) [21] and Tao et al. (2022) [22], this study quantifies the intensity of provincial government support for the digital economy based on the frequency of 39 selected keywords (see

Table A4. Selected Keywords Related to the Digital Economy.

Category	Selected Keywords
Communication and Information Infrastructure	5G, ICT, Communication Infrastructure, Internet, Modern Information Networks, Information Infrastructure, Internet of Things (IoT)
Digital Economy and Digitization	Digital Economy, Digitization, Digital Information, Digital Industrialization, Industrial Digitization, Information Economy
Intelligent Technology and Smartization	Artificial Intelligence, Smartization, Smart Economy, Intelligent Manufacturing, Robotics
Data Technology and Governance	Big Data, Data Services, Data Sharing, Datafication, Data Governance
Information Technology and Cloud Computing	Information Technology, Information Industry, Cloud Computing, Cloud Technology, Cloud Platform, Cloud Services, Mobile Payment
Digital Application Scenarios	E-commerce, E-government, Smart Cities, Smartization, Smart Economy

) related to the digital economy extracted from provincial government work report texts. The textual data were web-scraped and cleaned, followed by the removal of stop words. Tokenization was performed using the Jieba segmentation library (version 0.42.1) in Python 3.12, and keyword frequencies were extracted using regular expressions.

Based on this, provinces with a cumulative keyword frequency of 58 or greater during the period of 2011–2016 were classified into the treatment group (treated_i coded as 1), while the remaining provinces were designated as the control group (treated_i coded as 0). The threshold of 58 was determined as the trimmed median of cumulative keyword frequencies across provinces over 2011–2016, excluding the lowest and highest 5% of values, ensuring a robust classification of the treatment group and mitigating the influence of extreme values. The policy timing dummy variable time_t takes a value of 0 for the year 2016 and earlier, and 1 for years after 2016.

4.1.3. Mechanism Variable

Table 2. Indicator system for mediating variables.

Primary Indicator	Secondary Indicator	Indicator Composition	Data Sources
Fishery Digital Enabling Capacity (FDEC)	Proportion of Digital Industry Employment	Ratio of Employees in Information Transmission, Software, and Information Technology Services to Total Urban Unit Employment	National Bureau of Statistics of China, Wind Database, China Internet Network Information Center, Provincial Departments/Bureaus of Finance, China Fisheries Statistical Yearbook
	Proportion of Digital Industry Enterprises	Ratio of Firms in Information Transmission, Software, and Information Technology Services to Total Number of Firms
	Education Development Level	Ratio of General Public Budget Education Expenditure to General Public Budget Expenditure
	Fishery Technology Extension Expenditure	/
	Transportation Accessibility Index	Road Length per Resident
Fisher Digital Literacy (FDL)	Professional Training of Aquaculture Extension Personnel (standardized)	/	China Fisheries Statistical Yearbook

presents the indicator system used to measure the mediating variables in this study.

Building on the mechanism through which the digital economy affects fisher household vulnerability, this study selects two mediating variables. First, fishery digital enabling capacity (FDEC) is employed to capture the digital support, industrial environment, technology extension, and infrastructure relevant to the fisheries sector, reflecting the industry-level enabling role of digitalization in fisheries development. This indicator integrates multiple variables, including the proportion of employment in the digital industry, the proportion of digital industry enterprises, the level of education development, expenditure on fishery technology extension institutions, and the transportation accessibility index, thereby providing a comprehensive measure of regional digital development and industry support capacity in fisheries. To consolidate these multiple indicators and reduce dimensionality, principal component analysis (PCA) is applied to extract a composite index, which is then used in subsequent empirical regressions. The selection of indicators is based on existing literature (Guo et al., 2020 [23]; Zhao et al., 2020 [24]; Zhang et al., 2025 [25]) while also incorporating sector-specific requirements of fisheries to ensure that the mediator adequately reflects the level of digital development in the sector.

Second, fisher digital literacy (FDL) is chosen as the second mediating variable to capture the digital capability at the individual household level. Given data availability, digital literacy is measured by the professional training of aquaculture extension personnel, which reflects the demand for professional training and knowledge dissemination in the fisheries sector. Higher values of this indicator indicate greater digital literacy among fishers, representing a stronger ability to acquire and apply digital technologies effectively.

4.1.4. Control Variables

Specifically, regarding economic fundamentals, fishery output value (denoted as FOV) is selected as a representative indicator to measure the level of regional fishery economic development. Per capita regional GDP (denoted as PCGDP) is also incorporated to reflect overall economic conditions, thereby controlling for the influence of macroeconomic factors on fishers’ resilience to risks. Regarding infrastructure and technological capacity, road mileage (denoted as RL) serves as a proxy for transportation accessibility and logistical convenience, while the intensity of R&D expenditure (defined as the ratio of research and experimental development expenditure to regional GDP, denoted as RDEI) reflects technological investment supporting the adaptive capacity of fisheries. In terms of market environment, the cumulative year-on-year growth rate of the consumer price index for aquatic products (denoted as APCPI-GR) is employed to measure the impact of price fluctuations on fishers’ livelihood stability. Regarding industrial structure, control variables include the proportion of aquatic product processing (the ratio of processing volume to total aquatic product output, denoted as PPAP), which reflects the degree of extension and value addition in the fishery industrial chain. For climate change, the number of extreme high-temperature days (defined as days within a year when the daily maximum temperature exceeds the 90th percentile, denoted as HTD) is included to capture the adverse effects of climate anomalies on fishery production. In terms of financial support, the proportion of loans to agriculture, forestry, animal husbandry, and fishery sectors (denoted as PLAF) is used as an indicator of fishers’ access to financial resources and the level of policy support. The data sources include the National Bureau of Statistics of China, the China Fishery Statistical Yearbook, the U.S. National Oceanic and Atmospheric Administration (NOAA), and the People’s Bank of China. Collectively, these control variables enable a comprehensive and multidimensional assessment of the mechanisms through which digital economy policies influence fishers’ vulnerability, thereby enhancing the robustness and explanatory power of the empirical model.

4.2. Model Specification

This study constructs the following DID regression model:

$${\mathrm{F}\mathrm{V}}_{\mathrm{i}\mathrm{t}}={\mathrm{\alpha }}_0+{\mathrm{\alpha }}_1{\mathrm{t}\mathrm{r}\mathrm{e}\mathrm{a}\mathrm{t}\mathrm{e}\mathrm{d}}_{\mathrm{i}}\times {\mathrm{t}\mathrm{i}\mathrm{m}\mathrm{e}}_{\mathrm{t}}+\sum {\mathrm{\alpha }}_{\mathrm{j}}{\mathrm{x}}_{\mathrm{i}\mathrm{t}}+{\mathrm{\eta }}_{\mathrm{i}}+{\mathrm{\mu }}_{\mathrm{t}}+{\mathrm{\epsilon }}_{\mathrm{i}\mathrm{t}}$$

(1)

To mitigate potential sample selection bias, this study applies the Propensity Score Matching (PSM) method to identify control group cities that are most comparable to the treatment group in terms of observable covariates. Based on the matched samples, the following PSM-DID regression model is constructed.

$${\mathrm{F}\mathrm{V}}_{\mathrm{i}\mathrm{t}}^{\mathrm{p}\mathrm{s}\mathrm{m}}={\mathrm{\alpha }}_0+{\mathrm{\alpha }}_1{\mathrm{t}\mathrm{r}\mathrm{e}\mathrm{a}\mathrm{t}\mathrm{e}\mathrm{d}}_{\mathrm{i}}\times {\mathrm{t}\mathrm{i}\mathrm{m}\mathrm{e}}_{\mathrm{t}}+\sum {\mathrm{\alpha }}_{\mathrm{j}}{\mathrm{x}}_{\mathrm{i}\mathrm{t}}+{\mathrm{\eta }}_{\mathrm{i}}+{\mathrm{\mu }}_{\mathrm{t}}+{\mathrm{\epsilon }}_{\mathrm{i}\mathrm{t}}$$

(2)

In Equations (1) and (2), FV_it and ${\mathrm{F}\mathrm{V}}_{\mathrm{i}\mathrm{t}}^{\mathrm{p}\mathrm{s}\mathrm{m}}$ denotes the fisher household vulnerability in the province i at year t. The key explanatory variable is the interaction term treated_i × time_t, where treated_i is the policy group dummy variable that equals 1 for provinces in the treatment group and 0 for those in the control group, and time_t is the policy time dummy variable that equals 0 for years up to and including 2016, and 1 for years after 2016. $\sum {\mathrm{\alpha }}_{\mathrm{j}}{\mathrm{x}}_{\mathrm{i}\mathrm{t}}$ represents a vector of control variables. ${\mathrm{\eta }}_{\mathrm{i}}$ denotes province fixed effects, ${\mathrm{\mu }}_{\mathrm{t}}$ denotes year fixed effects, and ${\mathrm{\epsilon }}_{\mathrm{i}\mathrm{t}}$ is the stochastic error term, assumed to be normally distributed.

To ensure comparability and improve estimation robustness, all continuous variables are standardized, except for the core explanatory variable. The standardization is conducted as follows:

$${\mathrm{X}}^{\mathrm{*}}={\displaystyle \frac{\mathrm{X}-\overline{\mathrm{X}}}{{\mathrm{\sigma }}_{\mathrm{X}}}}$$

(3)

where denotes the original variable, $\overline{\mathrm{X}}$ represents the sample mean, and ${\mathrm{\sigma }}_{\mathrm{X}}$ denotes the sample standard deviation.

To examine the underlying mechanisms through which digital economy policies affect fisher household vulnerability, this study adopts the classical approach to mediation analysis and specifies the following regression models for mechanism testing:

$${\mathrm{M}}_{\mathrm{i}\mathrm{t}}={\mathrm{\beta }}_0+{\mathrm{\beta }}_1{\mathrm{t}\mathrm{r}\mathrm{e}\mathrm{a}\mathrm{t}\mathrm{e}\mathrm{d}}_{\mathrm{i}}\times {\mathrm{t}\mathrm{i}\mathrm{m}\mathrm{e}}_{\mathrm{t}}+\sum {\mathrm{\beta }}_{\mathrm{j}}{\mathrm{x}}_{\mathrm{i}\mathrm{t}}+{\mathrm{\eta }}_{\mathrm{i}}+{\mathrm{\mu }}_{\mathrm{t}}+{\mathrm{\epsilon }}_{\mathrm{i}\mathrm{t}}$$

(4)

$${\mathrm{Y}}_{\mathrm{i}\mathrm{t}}={\mathrm{\gamma }}_0+{\mathrm{\gamma }}_1{\mathrm{t}\mathrm{r}\mathrm{e}\mathrm{a}\mathrm{t}\mathrm{e}\mathrm{d}}_{\mathrm{i}}\times {\mathrm{t}\mathrm{i}\mathrm{m}\mathrm{e}}_{\mathrm{t}}+{\mathrm{\gamma }}_2{\mathrm{M}}_{\mathrm{i}\mathrm{t}}+\sum {\mathrm{\gamma }}_{\mathrm{j}}{\mathrm{x}}_{\mathrm{i}\mathrm{t}}+{\mathrm{\eta }}_{\mathrm{i}}+{\mathrm{\mu }}_{\mathrm{t}}+{\mathrm{\epsilon }}_{\mathrm{i}\mathrm{t}}$$

(5)

In Equations (4) and (5), M_it represents the mediating variables, namely digital basic capability and digital deepening capability. The definitions of the remaining variables are consistent with those in Equation (1).

5. Empirical Results and Analysis

5.1. DID Regression Results

Table 3. DID-based regression analysis of digital economy policies’ impact on fisher household vulnerability.

	(1)	(2)
treatedi × timet	−0.3534 *** (−2.96)	−0.5923 *** (−3.36)
Controls	No	Yes
Province FE	Yes	Yes
Year FE	Yes	Yes
N	377	377
Adj. R Square	0.6211	0.6347

Note: *, ** and *** denote statistical significance at the 10%, 5%, and 1% levels, respectively. t-values are reported in parentheses throughout. Due to space constraints, regression coefficients for control variables and the intercept are omitted but are available upon request. The same applies hereinafter.

reports the regression results examining the impact of digital economy policies on fisher household vulnerability. Column (1) includes only the core explanatory variable. The results show that the estimated coefficient of treated_i × time_t is negative and statistically significant at the 1% level, preliminarily suggesting that digitalization advances help alleviate fisher household vulnerability. To account for potential confounding effects from exogenous factors such as economic fundamentals, infrastructure, technological capacity, and market environment, Column (2) introduces relevant control variables. The estimated coefficient of treated_i × time_t remains significantly negative, indicating that the mitigating effect of digital economy policies persists after controlling for these factors. This result confirms the robustness and reliability of the baseline findings.

To further address potential spatial dependence and inference concerns, we compute Moran’s I of the regression residuals based on the provincial adjacency matrix. The Moran’s I statistic is 0.069, with a corresponding z-value of 0.543 and p-value of 0.587, indicating no significant spatial autocorrelation. This suggests that cross-province spillovers are unlikely to bias the estimated effects, and the results are spatially robust. Regarding statistical inference, the regressions are estimated using heteroskedasticity-robust standard errors, which effectively account for within-panel heteroskedasticity and autocorrelation. In addition, a wild cluster bootstrap procedure clustered at the provincial level (1000 replications) yields a bootstrap p-value of 0.001 for the core coefficient, reaffirming its strong statistical significance and robustness.

To assess the robustness of the DID estimates, we conduct a leave-one-out Jackknife sensitivity analysis. As shown in Figure 2, the coefficients remain stable after omitting any single province and largely overlap with the full-sample estimate, indicating that no single province drives the results and confirming the robustness of the estimated policy effect.

5.2. Parallel Trends Test

This paper utilizes the event study approach to rigorously test the parallel trends assumption and to visually capture the dynamic impact of digital economy policies on fisher household vulnerability. The following econometric model is constructed for this purpose:

$${\mathrm{F}\mathrm{V}}_{\mathrm{i}\mathrm{t}}={\mathrm{\theta }}_0+{\sum }_{\mathrm{s}=-4,\mathrm{s}\ne -1}^3{\mathrm{\theta }}_{\mathrm{s}}·1({\mathrm{t}\mathrm{i}\mathrm{m}\mathrm{e}}_{\mathrm{t}}-{\mathrm{T}}_{\mathrm{D}}=\mathrm{s})·{\mathrm{t}\mathrm{r}\mathrm{e}\mathrm{a}\mathrm{t}\mathrm{e}\mathrm{d}}_{\mathrm{i}}+\sum {\mathrm{\theta }}_{\mathrm{j}}{\mathrm{x}}_{\mathrm{i}\mathrm{t}}+{\mathrm{\eta }}_{\mathrm{i}}+{\mathrm{\mu }}_{\mathrm{t}}+{\mathrm{\epsilon }}_{\mathrm{i}\mathrm{t}}$$

(6)

Here, 1(time_t − T_D = s)denotes the interaction term for the individual i at relative time s relative to the policy implementation year, indicating membership in the treatment group. This term is employed to estimate the policy effects for each year, with s = −1 omitted as the baseline period. The coefficients ${\mathrm{\theta }}_{\mathrm{s}}$ represent the treatment effects at various time points, which are used to rigorously test the parallel trends assumption and to trace the dynamic evolution of the policy impact.

In theory, if the parallel trends assumption held, the trends of the outcome variable in the treated and control groups would not have differed significantly prior to the policy implementation. Figure 3 presents the dynamic effects of the digital economy policy on fisher household vulnerability. The estimated coefficients for the pre-policy years did not reach the 10% significance level, indicating that the trends in the treated and control groups were generally consistent before the policy, thereby supporting the parallel trends assumption underlying the DID analysis. Further joint significance tests showed that the pre-treatment coefficients were jointly insignificant at the 5% level, providing additional evidence that there were no systematic differences between the groups before the policy.

Starting from the policy implementation year, the estimated coefficients became significantly negative and remained significant in subsequent years. The corresponding joint significance test for the contemporaneous and post-policy coefficients was significant at the 1% level, indicating that the digital economy policy effectively reduced fisher household vulnerability and generated a clear dynamic improvement effect. Overall, both the dynamic estimates and the joint significance tests confirm that the policy impact was robust and persistent.

5.3. PSM-DID and Continuous-Intensity DID Regression Results

To mitigate sample selection bias, this study employs a combined Propensity Score Matching and Difference-in-Differences (PSM-DID) approach to further identify the impact of digital economy policies on fisher household vulnerability. Specifically, propensity scores for each province are estimated using the kernel matching method, and treatment and control groups are matched based on the principle of score proximity. Subsequently, a covariate balance test is conducted on the matched samples. As shown in

Table 4. Covariate balance test results from propensity score matching.

Variable Name	Matching	Mean		Standardized Bias (%)	Bias Reduction (%)	t-Statistic	p > t
		Treated	Control
FOV	Before	0.1916	−0.3135	53.6	85.7	4.90	0.000
	After	0.0900	0.0178	7.7		0.74	0.461
PCGDP	Before	0.0064	−0.0105	1.7	−18.0	0.16	0.874
	After	0.0092	−0.0108	2.0		0.20	0.843
RL	Before	−0.0679	0.1111	−18.0	65.1	−1.69	0.092
	After	−0.0357	−0.0982	6.3		0.69	0.493
RDEI	Before	0.0332	−0.0544	9.2	−1.5	0.83	0.410
	After	0.0378	−0.0512	9.3		0.96	0.337
APCPI-GR	Before	0.0128	−0.0210	3.4	−58.6	0.32	0.750
	After	0.0140	−0.0397	5.3		0.55	0.580
PPAP	Before	−0.0902	0.1475	−21.8	58.1	−2.25	0.025
	After	−0.0682	−0.1679	9.1		1.08	0.280
HTD	Before	0.0071	−0.0115	1.9	−104.5	0.17	0.861
	After	−0.0172	0.0208	−4.0		−0.41	0.680
PLAF	Before	−0.1128	0.1846	−28.6	85.4	−2.83	0.005
	After	−0.0867	−0.1302	4.2		0.62	0.534

, the absolute standardized differences in all matched variables are below 10% after matching, indicating a substantial reduction in covariate imbalance and demonstrating good matching quality. In addition, the t-statistics for all variables are statistically insignificant post-matching, further suggesting that the covariate differences between the treatment and control groups have been effectively minimized. This enhances the comparability between groups and provides a solid foundation for the accurate identification of policy effects.

Using propensity score weights obtained via kernel matching, this study conducts weighted DID regression on the matched samples.

Table 5. PSM-DID and continuous-Intensity DID estimates of the impact of digital economy policies on fisher household vulnerability.

	(1)	(2)	(3)
treatedi × timet	−0.3538 *** (−2.73)	−0.5646 *** (−3.22)	−0.4797 *** (−2.61)
Controls	No	Yes	Yes
Province FE	Yes	Yes	Yes
Year FE	Yes	Yes	Yes
N	358	358	377
Adj. R Square	0.6691	0.6369	0.6292

Note: *, ** and *** denote statistical significance at the 10%, 5%, and 1% levels, respectively.

reports the PSM-DID regression results. The results indicate that the estimated coefficients of the digital economy policy on fisher household vulnerability are significantly negative at the 1% level, regardless of the inclusion of control variables. This suggests that the policy implementation has effectively reduced fisher household vulnerability and significantly mitigated associated risks.

Furthermore, to account for heterogeneity in policy exposure and potential biases arising from the binary specification of the key explanatory variable, this study implements a continuous-intensity DID analysis. Specifically, a continuous treatment variable is constructed based on the frequency of digital economy keywords in provincial work reports from the year prior to policy implementation, and interacted with the policy period dummy. Column 3 of Table 5 reports the results of the continuous-intensity DID, indicating that provinces with higher pre-policy keyword frequency experienced greater reductions in fisher household vulnerability, consistent with the baseline regression results. This analysis not only quantifies the effect of policy exposure intensity but also further reinforces the robustness and causal interpretation of the estimated policy effects.

5.4. Mechanism Identification Analysis: Mediation Effect Test

Table 6. Results of the impact mechanism.

	FDEC		FDL
	FDEC	FV	FDL	FV
treatedi × timet	0.1889 * (1.89)	−0.4349 ** (−2.05)	0.4808 *** (3.06)	−0.5450 *** (−3.09)
FDEC		−0.2095 ** (−2.20)
FDL				−0.0983 ** (−2.18)
Controls	Yes	Yes	Yes
Province FE	Yes	Yes	Yes
Year FE	Yes	Yes	Yes
N	377	377	377	377
Adj. R Square	0.9681	0.6378	0.6878	0.6894

Note: *, ** and *** denote statistical significance at the 10%, 5%, and 1% levels, respectively.

presents the estimated effects of digital economy policies on fishery digital enabling capacity and fisher digital literacy. Considering the time-lagged nature of policy effects, this study incorporates a two-period lag of FDEC in the empirical analysis to more accurately capture the dynamic impact of digital economy policies on the empowerment of the fishery system. The results indicate that digital economy policies have a significantly positive effect on FDEC, suggesting that such policies effectively promote the development of digital infrastructure, information transmission, and industrial digitalization support within the fishery sector. Further mediation analysis reveals that the two-period lagged FDEC exerts a significant negative influence on fisher household vulnerability, implying that FDEC serves as an important mediating mechanism through which digital economy policies reduce vulnerability. Specifically, by enhancing the degree of fishery digitalization and the accessibility of information, FDEC improves production efficiency and risk response capacity, thereby indirectly lowering fisher vulnerability.

Moreover, digital economy policies significantly enhance fisher digital literacy. When both digital economy policies and digital literacy are included in the regression model, digital literacy exhibits a significant negative effect on fisher vulnerability, while the direct effect of the policy remains significant. This finding indicates that digital literacy constitutes another critical mediating channel through which digital economy policies indirectly mitigate fisher vulnerability.

Overall, digital economy policies primarily improve fisher resilience and reduce vulnerability by strengthening fishery digital enabling capacity and enhancing fisher digital literacy. Fishery digital enabling capacity provides a digital foundation for policy transmission, whereas fisher digital literacy further facilitates the behavioral transformation and sustainability of policy effects at the individual level.

6. Robustness Tests and Heterogeneity Analysis

6.1. Placebo Test: Fictitious Treatment Group

To address potential confounding from unobserved factors and to further verify the robustness of the estimated policy effects, this study conducts a placebo test, following the methodologies of Chetty et al. (2009) and Ye et al. (2023) [26,27]. Specifically, all provinces are randomly perturbed using a sampling method to generate pseudo treatment and control groups, thereby constructing a placebo treatment variable treated_i. This variable is subsequently interacted with the policy shock variable time_t to produce a placebo core explanatory variable treated_i × time_t for robustness testing.

Theoretically, the placebo explanatory variable should have no causal relationship with fisher household vulnerability, and its regression coefficient is therefore expected to be statistically insignificant. To mitigate potential bias arising from random variation, the placebo test is repeated 1000 times. Figure 4 displays the kernel density distribution and corresponding p-values of the estimated coefficients derived from the 1000 placebo replications. The results indicate that most placebo estimates are statistically indistinguishable from zero, whereas the actual policy effect lies far outside this distribution, suggesting it is not a product of chance. This provides robust evidence that the baseline regression results are not driven by unobserved confounding factors, but instead reflect the genuine impact of the digital economy policy.

6.2. Placebo Test: Fictitious Treatment Timing

To assess the robustness of the baseline regression results and rule out potential confounding from unobserved time-varying factors, we advance the actual policy implementation by 1 to 4 periods and re-estimate the placebo effects at each fictitious treatment timing. The results (see Figure 5) indicate that the estimated placebo effects are uniformly insignificant, with 95% confidence intervals encompassing zero, suggesting no spurious effects prior to the actual policy implementation and further confirming the causal validity of the baseline estimates.

6.3. Robustness Check: Secondary Indicator Weights of Fisher Household Vulnerability

To assess the robustness of the secondary indicator weights in the comprehensive evaluation system of fisher household vulnerability, this study employed the Bootstrap resampling method based on the original sample (N = 386) to validate the entropy-TOPSIS weights. Specifically, 1000 bootstrap samples were generated with replacement from the original dataset, the secondary indicator weights were recalculated for each resample, and the mean values and 95% confidence intervals (CI) were computed across the 1000 iterations. Figure 6 presents the comparison between the original entropy-TOPSIS weights and the Bootstrap mean weights with their corresponding 95% CIs. The results indicate a high degree of consistency between the original and resampled mean weights. The core indicators (e.g., E3, S3, E5) exhibit minimal deviations and stable rankings, suggesting that the indicator weights remain robust under sample perturbations. Overall, the secondary indicator weights are stable and reliable, confirming the methodological soundness of the index construction and providing robust support for the comprehensive vulnerability assessment.

6.4. Regional Heterogeneity: Coastal and Non-Coastal Provinces

Given the variation in resource endowments across provinces for developing the digital economy, the effectiveness of digital economy policies may differ accordingly. Therefore, it is essential to examine whether the policy effects differ across regions. Accordingly, this study divides the sample into coastal and non-coastal provinces and separately estimates the policy’s impact on fisher household vulnerability.

Table 7. Regression results of regional heterogeneity analysis.

	Coastal Provinces	Non-Coastal Provinces
did	−0.7895 ** (−2.36)	−0.4425 *** (−2.63)
Controls	Yes	Yes
Province FE	Yes	Yes
Year FE	Yes	Yes
N	143	234
Adj. R Square	0.5015	0.7563

Note: *, ** and *** denote statistical significance at the 10%, 5%, and 1% levels, respectively.

reports the region-specific regression results. The empirical results show that digital economy policies significantly reduce fisher household vulnerability in both coastal and non-coastal provinces. Notably, although the policy effect is significant in both groups, the DID coefficient is more negative for coastal provinces, indicating a stronger reduction in vulnerability and a more pronounced policy impact in these areas. This discrepancy may stem from better digital infrastructure, easier access to digital resources, and more advanced industrial digitization in coastal provinces, which collectively enhance policy transmission and implementation effectiveness.

6.5. Economic Development Heterogeneity: Developed and Less-Developed Provinces

Considering the significant differences in economic development levels across provinces, the implementation effects of digital economy policies may exhibit heterogeneity across regions at different stages of development. Accordingly, this study divides the sample provinces into developed and less-developed groups based on the median per capita GDP and estimates the impact of digital economy policies on fisher household vulnerability for each subsample. The regression results (see

Table 8. Regression results of economic development heterogeneity analysis.

	Developed Provinces	Less-Developed Provinces
did	−0.2055 (−0.93)	−0.7318 *** (−3.15)
Controls	Yes	Yes
Province FE	Yes	Yes
Year FE	Yes	Yes
N	169	208
Adj. R Square	0.5429	0.6032

Note: *, ** and *** denote statistical significance at the 10%, 5%, and 1% levels, respectively.

) indicate that digital economy policies significantly reduce fisher household vulnerability in less-developed provinces, while the effect in developed provinces does not reach statistical significance.

In coastal regions, the policy effect is larger in magnitude, whereas in developed provinces, the baseline vulnerability of fisher households is relatively low and the digital infrastructure is more advanced, limiting the marginal effect of the policy and resulting in an insignificant effect. This suggests that digital economy policies exert a stronger mitigating effect in economically less-developed regions, effectively reducing fisher household vulnerability, while the marginal benefit of such policies in economically developed regions is relatively limited.

6.6. Policy Environment Heterogeneity: High and Low-Policy-Intensity Provinces

To further investigate how the policy environment moderates the role of the digital economy in reducing fisher household vulnerability, this study adopts the total number of digital economy-related policies enacted by each province as a proxy for policy intensity. Based on the median policy intensity, the sample is classified into high- and low-intensity groups, and separate regression analyses are conducted for each. The results of heterogeneity analysis is presented in

Table 9. Regression results of policy environment heterogeneity analysis.

	High-Policy-Intensity Provinces	Low-Policy-Intensity Provinces
did	−1.0051 ** (−3.25)	−0.4565 ** (−2.38)
Controls	Yes	Yes
Province FE	Yes	Yes
Year FE	Yes	Yes
N	182	195
Adj. R Square	0.6631	0.6275

Note: *, ** and *** denote statistical significance at the 10%, 5%, and 1% levels, respectively.

.

The heterogeneity analysis reveals that the mitigating effect of the digital economy on fisher household vulnerability is more pronounced in high-policy-intensity areas than in low-policy-intensity ones. This suggests that a supportive policy environment positively moderates the effectiveness of the digital economy. This finding highlights the importance of dense policy support in enhancing the resilience of the fishery industry chain, and suggests that governments should focus on the systematic and sustained implementation of digital economy policies to fully harness their agglomeration effects.

7. Research Conclusions and Policy Recommendations

The digital economy has emerged as a pivotal driver of economic restructuring and high-quality development. Against the backdrop of China’s strategic promotion of the Digital China initiative and the Maritime Power strategy, this study systematically investigates how digital economy policies affect fisher household vulnerability based on panel data from 29 provinces covering 2011–2023 in China. Empirical analyses using both DID and PSM-DID approaches consistently demonstrate that implementing digital economy policies significantly reduces fisher household vulnerability and strengthens the resilience of fishery workers. To explore the underlying mechanisms, this study incorporates two mediating variables. The results show that both channels play important roles in reducing fisher household vulnerability. Specifically, fishery digital enabling capacity significantly promotes information access, industrial digitalization, and operational efficiency within the fishery sector, thereby enhancing system-level digital empowerment and reducing vulnerability. Meanwhile, fisher digital literacy strengthens individual adaptive capacity and digital engagement, further mitigating vulnerability at the household level. These findings suggest that digital economy policies alleviate fisher vulnerability by simultaneously improving the digital foundation and fostering human-level digital competencies. Further heterogeneity analysis shows that digital economy policies significantly reduce fisher household vulnerability in both coastal and non-coastal regions, with stronger effects observed in coastal areas. This regional disparity may be attributed to more advanced digital infrastructure, more efficient resource allocation, and stronger industrial coordination in coastal provinces. Considering differences in economic development levels, provinces are further divided into developed and less-developed groups based on the median per capita GDP. The results indicate that digital economy policies significantly reduce vulnerability in less-developed provinces, while the effects in developed provinces are not statistically significant. These findings suggest that digital economy policies have greater mitigating effects in coastal and economically less-developed regions. Moreover, the policy impact is significantly stronger in regions with high policy intensity than in those with lower levels of policy support. This suggests that a robust policy environment amplifies the effectiveness of the digital economy in mitigating household vulnerability.

This study provides key insights into how digital economy policies can reduce fisher vulnerability in the context of digital transformation. Based on empirical evidence, we offer the following policy recommendations: (1) Enhance coordination and precision of digital economy policies. The results confirm that digital policies significantly reduce fisher vulnerability. To further improve effectiveness, top-level policy design should consider regional disparities and the specific needs of the fisheries sector. Policies should shift from broad-based coverage to targeted interventions, especially in traditional fishing regions, with sustained and coordinated policy delivery to maximize impact. (2) Promote digital applications across the fisheries value chain. Digital deepening capacity plays a crucial mediating role in reducing vulnerability. Governments should promote the adoption of digital technologies across all stages of the fisheries chain, with a focus on smart fisheries, blockchain traceability, digital finance, and information services. These measures can improve access to information, management efficiency, and financial inclusion, enhancing fishers’ resilience to external shocks. (3) Strengthen regional coordination and digital adaptability in non-coastal areas. The regional analysis indicates stronger policy effects in coastal provinces. To bridge this gap, it is essential to enhance digital infrastructure, institutional capacity, and resource integration in non-coastal regions. Local governments should develop localized ‘digital + fisheries’ strategies and invest in talent development and technology adoption. Additionally, they should improve the reach and effectiveness of digital economy policies to promote inclusive and sustainable fisheries development.