Understanding the timing of Chinese border incursions into India

The paper investigates when and why Chinese military incursions into Indian or disputed territory along the China–India border occur. The context is a decades-long territorial dispute, including a 1962 war and periodic flare-ups such as the deadly 2020 clashes. Given risks to civilians, great-power relations, and global supply chains, understanding drivers of these incursions is a priority. The authors compile an original monthly dataset (2005–2019) of reported Chinese incursions and 18 explanatory variables spanning bilateral China–India interactions, broader alliance and rivalry dynamics, domestic politics, and economic indicators. They formulate testable hypotheses from international relations theory (reciprocity, triangular reciprocity, diversionary conflict, election/tenure uncertainty, and domestic unrest) and assess correlates that predict incursions with 1–6 month lead times.

The review synthesizes theories relevant to China–India encounters and develops hypotheses. Reciprocity theory posits that hostile actions beget future hostility, implying that periods of China–India tension should raise incursion likelihood (Hypothesis 1). Triangular reciprocity suggests third-party relations matter: increased India cooperation with China’s rivals (notably the US and Japan, within arrangements like the Quad) may trigger Chinese activity (Hypothesis 2a). Conversely, dynamics with China’s allies, particularly Pakistan, can spill over: heightened India–Pakistan tensions or closer China–Pakistan ties could increase China–India border incidents (Hypothesis 2b). Broader Chinese territorial assertiveness elsewhere (e.g., Japan, Vietnam, the Philippines, Taiwan) may correlate with incursions into India (Hypothesis 2c). Domestic politics can create windows of vulnerability or uncertainty: unrest and elections in India may invite external pressure (Hypotheses 3a, 3b), while leadership tenure effects imply higher conflict risk early in a leader’s term due to uncertainty. Diversionary conflict theories emphasize that leaders might externalize during domestic troubles; for China, economic downturns and uncertainty—even more than protests or elections—may incentivize assertiveness (Hypotheses 4a, 4b).

Study design and data: The authors build a monthly panel for 2005–2019 on the China–India dyad. The dependent variable is a binary indicator of a reported Chinese incursion into Indian or disputed territory in a given month, coded from LexisUni and multiple independent journalistic/academic sources; government-only claims without third-party verification are excluded. The dependent variable is led by 1–6 months to model near-term risk windows. Independent variables: Drawing on theory and multiple sources, they construct measures in four domains: (1) China–India dyadic interactions: binary indicators of Chinese and Indian border construction and troop deployments near the LAC (coded from policy publications and LexisUni); ICEWS counts of conflictual events between India and China (with border-incursion events removed to avoid leakage); and a binary indicator for executive leader meetings (summits/visits) between India and China. (2) Alliance/rivalry network effects: ICEWS Goldstein-based average cooperation score (0–10) for India–US; Goldstein-based hostility score (0 to −10) for India–Pakistan; and a binary indicator for Chinese territorial disputes elsewhere (Japan, Vietnam, the Philippines, Taiwan) using ICEWS categories. (3) Indian domestic politics/unrest: leader tenure in months and indicator for elections expected within six months (REIGN); internal conflict battle deaths (monthly, UCDP GED); violent anti-government protests (ICEWS). (4) Chinese domestic politics/economy: Chinese leader tenure in months (REIGN); violent anti-government protests (ICEWS); economic policy uncertainty index (FRED series based on Davis et al. methodology); and consumer confidence index (OECD CCI; <100 pessimistic, >100 optimistic). To address temporal confounding, models include a linear time trend and a summer dummy (June–August) given seasonal border accessibility patterns. Statistical methods: They estimate binary logistic regressions predicting an incursion at leads 1–6 months with all covariates included. Multicollinearity diagnostics (correlations and VIF) show no issues for the final model set. Variables that risk leakage (e.g., direct mobilization/territorial occupation events) are excluded from certain constructs, and some highly collinear trade variables are omitted in final analyses.

Overall, results provide evidence consistent with several theoretical expectations and highlight lead-time-dependent effects:

• Reciprocity and dyadic tension: Higher China–India hostility (ICEWS) is positively associated with future incursions; e.g., at t+3 months, IN–CH hostility coefficient β≈0.06 (p<0.05).
• Border activities: Chinese construction near the border shows mixed timing effects—negative at t+3 (β≈−0.99; p≈0.07) but positive at t+5 (β≈1.06; p≈0.03), suggesting delayed risk after construction. Indian deployments are associated with lower risk at t+5 (β≈−1.36; p≈0.07) but higher risk at t+6 (β≈1.37; p≈0.05).
• Alliance/rivalry network effects: Greater India–US cooperation predicts higher incursion likelihood, notably at t+2 (β≈0.36; p<0.05) and marginally at t+1 (β≈0.31; p≈0.08). India–Pakistan hostility shows an inconsistent pattern, with a negative coefficient at t+4 (β≈−0.20; p<0.05), reflecting scale direction (less negative indicates reduced hostility), and a positive relationship at t+6.
• Indian domestic politics and unrest: Upcoming Indian elections are associated with higher risk at t+6 (β≈1.25; p≈0.07). Indian leader tenure increases risk at t+4 to t+6 (β≈0.01* to 0.02**), implying greater incursion odds later in an Indian leader’s term. Internal conflict deaths correlate with increased risk at t+2 (β≈0.02; p<0.05). Violent anti-government protests in India show a small, marginally negative association at t+2 (β≈−0.02; p≈0.09).
• Chinese domestic politics and economy: Longer Chinese leader tenure reduces risk at t+2 to t+5 (e.g., t+2 β≈−0.01*; t+5 β≈−0.02**), consistent with declining uncertainty over time. Lower consumer confidence (OECD CCI) is robustly linked to more incursions; coefficients are negative and significant at t+2, t+3, and t+5 (β≈−0.23 to −0.24; p<0.05). Expressed economic uncertainty (EPU) is generally not significant, with a marginal negative effect at t+1 (β≈−0.01; p≈0.07).
• Seasonality: Incursions are more likely in summer at short leads (t+1 β≈1.96; t+2 β≈0.96; both p<0.05) and less likely by t+6 (β≈−1.85; p<0.05). These results indicate heightened incursion risk when: China–India tensions are elevated; India–US cooperation increases; Indian leaders are later in tenure (and elections approach); Chinese leaders are early in tenure; and Chinese consumer confidence is low.

Findings address the central question of when Chinese incursions are more likely by identifying conditions that elevate near-term risk. They imply actionable indicators for policymakers: periods of Chinese economic weakness and early Chinese leadership tenure correlate with heightened probability of incursions; thus, India and partners should raise alert levels and emphasize deterrent signaling and crisis-management channels. Strengthening US–India ties, while strategically valuable, is associated with increased short-term incursion risk, suggesting a need for parallel risk-mitigation steps (e.g., clear communication, confidence-building measures). Mixed timing effects of border construction/deployments suggest close monitoring of LAC infrastructure activities and force postures, potentially aided by advanced surveillance and intelligence-sharing (e.g., with the US and Australia). Elevated tensions with Pakistan, as well as broader India–China hostility, also presage incursions, underscoring the interconnected nature of regional rivalries. Overall, the results reinforce theories of reciprocity and diversionary behavior, providing a lead-time window (1–6 months) for preventive diplomacy and military readiness.

The study presents one of the first comprehensive, data-driven analyses of Chinese incursions into Indian or disputed territory, leveraging a novel monthly dataset (2005–2019) and theory-informed covariates to identify correlates predicting incursions with 1–6 month lead times. Key contributions include evidence that: (i) early Chinese leader tenure and low Chinese consumer confidence are associated with higher incursion risk; (ii) greater India–US cooperation elevates risk; and (iii) Indian leader tenure and election timing also matter. Future research should translate these findings into operational security postures for India and partners, examine countermeasures and signaling strategies that reduce escalation, and extend the framework to other Chinese territorial disputes. Additional work could incorporate richer measures of economic leverage (e.g., boycotts, trade shocks) and enhance event validation across multiple independent sources.

Several limitations may affect interpretation and generalizability: (1) Incursions are coded only when corroborated by multiple independent sources; unverified government-only reports are excluded, potentially leading to undercounting. (2) The Line of Actual Control is poorly demarcated, complicating classification of events. (3) The analysis focuses on Chinese incursions into India; potential Indian incursions into China were largely unverifiable by third parties and thus excluded, which may limit dyadic symmetry analysis. (4) Some theoretically relevant variables lack reliable monthly measures; for example, Chinese internal conflict battle deaths are near zero and omitted. (5) Certain proxies are used due to data gaps (e.g., Chinese territorial disputes elsewhere via ICEWS), and some variables (trade flows) were excluded from final models to avoid multicollinearity. (6) While time trends and seasonality are modeled, unobserved confounders and measurement error in media-based event data (ICEWS, LexisUni) remain possible. (7) Reported lead-time effects may vary with alternative model specifications or different coding thresholds.