The 2021 Henan flood increased citizen demand for government-led climate change adaptation in China

Citizen petitioning and the Henan flood context

To examine the political aftermath of the flood, this study draws on a comprehensive and proprietary dataset of citizen queries submitted to the LLMB. Launched in 2008 and operated by People.cn—the official online arm of the People’s Daily—the LLMB is the largest centrally operated digital petition platform in China, covering all 32 provincial-level administrative units. All citizens can use the platform to report local misgovernance, request administrative support, or offer suggestions to government agencies.

When submitting a query, individuals choose a topic category and specify the target government office. Messages are subject to review and potential censorship by the platform before being posted publicly, after which local governments issue public responses. This dual public-facing structure, where both the citizens’ appeal and the government’s reply are visible, offers a unique lens into bottom-up engagement under authoritarian rule.

While detailed demographic information for individual LLMB petitioners is unavailable, regional variation in usage appears to be correlated with levels of urbanization and internet penetration³⁸. These structural factors likely shape who participates and what types of risks and demands are articulated. As such, the petition data should be understood as reflecting the expressed preferences of a politically engaged and digitally connected subset of the population, rather than a fully representative cross-section of Chinese society.

Submitting to LLMB constitutes a form of political behavior rather than merely expressing opinions. LLMB messages are formalized appeals concerning routine governance issues, such as water access, transportation, education, and environmental hazards^38,39. As a tool of “responsive authoritarianism,” the LLMB enables the central state to monitor local performance and absorb grassroots pressure without relinquishing political control. It thus serves as both a pressure-release valve and an instrument of bureaucratic accountability^33,39,40.

Crucially, the LLMB is analytically distinct from commercial blogs or microblogs, which have received substantial scholarly attention for their relationship to government censorship^41,42. Unlike social media platforms that amplify content through virality, LLMB enforces categorization, moderation, and hierarchical routing of complaints to government offices. Government-run microblogs, by contrast, are more often used for propaganda, information dissemination, or superficial public engagement rather than formal grievance redress^43,44.

The dataset used in this study includes all LLMB submissions between January 1 and October 4, 2021. Henan Province serves as the treatment group, while all other provinces serve as the control group. The time frame ends on October 4 to avoid contamination from a second major flood event in Henan that began on October 5, 2021.

To measure public demand for disaster-related risk mitigation, the dependent variable is defined as the daily percentage of petitions in each province that pertain to adaptation. In this study, climate adaptation requests refer to citizen petitions that express concern about flood risk, drainage, and infrastructure vulnerabilities, including both short-term appeals for immediate protection and longer-term calls for systemic resilience. While some petitions may reference damage repair, only those that imply anticipatory or structural risk reduction, rather than routine maintenance, are coded as adaptation-oriented. These petitions are interpreted as a proxy for adaptation-relevant concerns, based on the rationale that appeals concerning flood impact management most directly reflect citizen-identified needs for enhanced local resilience⁴⁵.

To identify relevant petitions, the study employs a keyword-based filtering approach. Messages are flagged as flood-related if they include one or more of the following terms: “flood water,” “heavy water,” “water disaster,” “flood disaster,” “torrent,” “flood waterlogging,” “rainstorm,” “flood control,” “flood prevention,” “flood discharge,” “flood drainage,” and “flood storage.” These terms were selected to capture citizen concerns about both immediate flood impacts and the need for anticipatory or preventive infrastructure.

To contextualize the Henan flood within the broader landscape of climate disasters during the study period, this study compares its severity to that of other major—but less destructive—floods occurring over the same timeframe. Information on climate disasters is compiled primarily from authoritative publications issued by the Ministry of Emergency Management (MEM), the central government agency responsible for disaster reporting and response in China. These official sources serve as the foundation for constructing a comprehensive and detailed list of climate-related disasters, particularly flooding events, across provinces. The dataset includes key impact metrics, such as the number of total deaths, the number of people affected, and estimated economic damages (in 2021 USD).

The EM-DAT database, maintained by the Centre for Research on the Epidemiology of Disasters at the Catholic University of Louvain in Belgium, remains a valuable global reference in disaster research. However, for the Chinese context, EM-DAT entries were occasionally found to lack granularity or to diverge from official domestic reporting. Therefore, EM-DAT is used selectively for cross-validation, while MEM records and provincial emergency bulletins are treated as primary data sources. In case of missing or ambiguous MEM data, supplementary information is drawn from other credible official reports and state media summaries.

Surge in adaptation demands in Henan

Figure 1 presents the log-transformed percentage of daily climate adaptation requests in Henan Province and the rest of the country from January 1 to October 4, 2021. These trends are visualized on a log scale to improve interpretability at the lower end of the distribution and to better capture meaningful changes in rare event data.

In the months leading up to the flood, Henan registered virtually no adaptation-related requests. Immediately following the onset of the July 2021 rainstorm, however, the percentage of climate adaptation requests in Henan spiked dramatically. While this initial surge later declined, levels remained markedly elevated through late summer and early fall, indicating a persistent shift in public attention and demand for government-led climate adaptation. In contrast, the rest of the country exhibited a comparably stable baseline throughout the period, despite six provinces also experiencing major flood events of lesser magnitude and impact than the 2021 Henan flood.

Spillover effects across provinces

The severe flood that struck Henan Province had measurable spillover effects, prompting a rise in climate adaptation demands in other regions of China (Fig. 2). From the onset of the rainstorm on July 17 through October 4, approximately 20% of climate adaptation requests (13 out of 67) submitted outside Henan explicitly referenced “Zhengzhou,” “Henan,” or the shorthand “720” and “7.20” commonly used to denote the disaster. One potential concern is that some of these messages might have been submitted by Henan residents posting to other provinces’ LLMB portals. However, IP address data confirm that all 13 messages originated from users physically located in the provinces where the petitions were submitted. These references suggest that the scale and destructiveness of the Henan flood prompted residents in other regions to reflect on their own vulnerabilities and to urge local governments to take preventive action, even in the absence of direct exposure to flooding. Illustrative excerpts from these requests are shown in Table 1.

At the same time, it is essential to acknowledge the limitations of content-based filtering: citizens affected by the Henan flood may not always explicitly reference it. Media coverage and emotional salience could still motivate individuals to request adaptation measures without directly naming the event. Such latent or unobservable spillover effects cannot be systematically detected, but their likely presence suggests that this analysis may underestimate the full influence of the Henan flood on public adaptation demands.

Distinctiveness of the Henan flood

Figure 3 visualizes trends in daily percentage of climate adaptation requests (log-transformed) alongside key disaster severity indicators across seven Chinese provinces affected by major floods from January 1 to October 4, 2021. Following established conventions⁴⁶, disaster severity is captured through three metrics: total fatalities, the number of people affected, and direct economic losses (converted to 2021 USD). Henan Province, which experienced the deadliest and most economically damaging flood, stands out with the largest and most sustained increase in climate adaptation requests. Shaanxi Province, which experienced two distinct flood events during the study period, exhibited a more noticeable increase in adaptation requests than other provinces, though the magnitude remained smaller than that observed in Henan.

Estimating causal effects of the Henan flood

To assess the timing and persistence of treatment effects and to test the parallel trends assumption, a dynamic difference-in-differences (DiD) model using event-time indicators is employed. This method, also known as an event study design, estimates how citizen demand for adaptation evolved on a weekly basis before and after the flood’s onset on July 24, 2021. The dynamic specification enables a flexible examination of treatment timing, including potential anticipatory behavior, and provides a visual diagnostic for pre-treatment trend violations^47,48.

Figure 4 presents the estimated weekly treatment effects relative to the week prior to the flood. Pre-treatment coefficients are centered around zero and statistically insignificant, supporting the plausibility of the parallel trends assumption. While one pre-treatment estimate is marginally significant, this is consistent with expected random variation rather than evidence of systematic pre-trend violations⁴⁷. Following the onset of the Henan flood, the results show a marked and sustained increase in adaptation-related requests in Henan, with several post-treatment coefficients reaching conventional significance thresholds. The pattern suggests that the flood catalyzed a meaningful and persistent shift in public demand for government-led climate adaptation.

Importantly, these estimates are likely conservative. As shown in earlier sections, the Henan flood generated spillover effects in other provinces. Through national media coverage and heightened public awareness, residents outside Henan may have also experienced increased concern about climate risks, even without direct flood exposure⁴⁹. These indirect effects reduce the contrast between treated and control units, potentially attenuating measured treatment effects. In addition, contemporaneous but less severe flooding in other provinces further narrows this contrast. As a result, the estimated effect size likely underrepresents the true magnitude of public political response to this mega disaster.

To complement the dynamic analysis, a standard DiD model was employed to estimate the average treatment effect of the Henan flood. Results indicate that the percentage of adaptation-related requests in Henan increased by approximately 0.39 percentage points relative to other provinces (p = 0.007). Although this numerical change appears modest, it is substantively significant given that Henan’s pre-flood baseline was effectively zero. The flood thus appears to have catalyzed the emergence of climate adaptation as a new domain of citizen petitioning. As noted earlier, spillover effects into the control group, driven by national media coverage and heightened risk awareness, likely attenuate the estimated effect. Accordingly, this result should be interpreted as a conservative, lower-bound estimate of the true impact.

Themes in adaptation appeals: a topic modeling analysis

Topic modeling of Henan climate adaptation petitions yielded six distinct themes, each grounded in tangible, place-specific risks associated with flood impacts. As shown in the keyword bar chart (Fig. 5), citizens focused primarily on infrastructural concerns. Topic 0 centers on river management and urban development, with keywords such as riverway, construction, and city. Topic 1 emphasizes major transit disruptions and neighborhood vulnerability, referencing “expressway”, “flood,” and “old city district.” Topic 2 focuses on residential services and drainage problems, with keywords like “water supply” and “basement.” Topic 3, dominated by terms such as “new river” and “management,” points to localized restoration or infrastructure projects. Topics 4 and 5 include broader socio-political framings, highlighting local governance, township priorities, and calls for state-led development.

Notably, terms such as “climate change”, “global warming”, or “climate warming” were never mentioned in citizen appeals. This underscores that the petitions reflect concrete calls for safety and infrastructure improvements in response to an extreme event, rather than ideational commitments to climate adaptation per se. Public demand for government-led disaster resilience was anchored in practical concerns rather than in the language of climate governance.