Border Effect of Complex Network: An analysis on the cooperation network of movie stars across different regions

Wen Zhou, Dong Zhang, Chen Chen, Dianbo LiuarXiv 2018 · OpenReview

Abstract

People always carry out social activities within certain geographic space. The geographic boundary likes national boundaries and provincial boundaries functions as dividing policies, distributing administrative resources, defining the scope of application of laws and regulations or even the boundary of cultural region, producing fundamental influence on lifestyle. Network analysis is a powerful tool to abstract complex system. In this article, we used network method to understand changes of cooperation network in movie industry and effect of geographical boundary.

Introduction

In an increasingly globalized and digitally connected world, the relevance of geographic and administrative boundaries remains a subject of debate. While previous research on the "border effect" has utilized macro-level data like trade volumes or snapshot spatial data from mobile check-ins to demonstrate that borders still restrict human interaction, these studies often fail to capture the longitudinal evolution of networks in response to specific policy shifts. This paper addresses this gap by analyzing the professional collaboration network of the Chinese entertainment industry across Mainland China, Hong Kong, and Taiwan from 1990 to 2009. By treating the 1997 Hong Kong handover as a natural experiment, the study explores how political uncertainty and subsequent integration policies physically reshape a professional social network.

Methodology

The entertainment industry is modeled as an undirected graph where nodes represent actors and edges signify co-starring roles in film or television productions. Two primary analytical techniques are employed:

  1. Cross-regional Stars Cooperation Index: To measure the true probability of cross-regional cooperation, the observed network is compared against a randomized null model. Using a degree-preserving randomization technique, the authors calculate the ratio of observed versus expected collaborations.
  2. Exponential Random Graph Models (ERGMs): Recognizing that network edges are interdependent, ERGMs are used to predict link formation based on node attributes (age, seniority, popularity) and the homogeneity of regional attributes—specifically, whether two actors originate from the same region.

Data Description

The study utilizes a custom-built Chinese Star Cooperative Database, compiled from Movie_1905 and Baidu Encyclopedia. The dataset encompasses 4,585 stars (3,297 from Mainland China, 617 from Taiwan, and 617 from Hong Kong) and 117,715 cooperation ties. Data is segmented into five periods (1990–1993, 1994–1997, 1998–2001, 2002–2005, and 2006–2009) to track dynamics surrounding the 1997 handover.

Key Results and Analysis

The Substitution Effect

Analysis of the cooperation index reveals that policy uncertainty creates a chilling effect on professional collaboration. Leading up to and immediately following the 1997 handover, institutional ambiguity led to a significant decline in Mainland-Hong Kong (MH) cooperation. Interestingly, the data suggests a "substitution effect": as MH collaborations dipped, Mainland-Taiwan (MT) collaborations spiked to fill the market demand for content.

Figure 3: Observed vs. expected collaboration percentages for Mainland-Hong Kong (MH) and Mainland-Taiwan (MT). Note the inverse relationship around the 1997-2000 period.

As shown in Figure 3, the MH cooperation index reached a nadir in 1999 before recovering. Conversely, the MT index peaked at over 0.73 in 1997, remaining elevated until 2000, effectively mirroring the decline in Hong Kong-based collaborations.

Regional Homophily and Policy Integration

The ERGM analysis quantifies the "border effect" through the coefficient of regional homogeneity. A higher positive value indicates a stronger preference for within-region cooperation, signifying a more robust barrier to cross-regional interaction.

Figure 4: Trends in the homogeneity of regional attributes coefficient. The Mainland-Hong Kong barrier peaks during the 1998-2001 handover period before declining.

The coefficient for Mainland-Hong Kong stars rose from 1.1 (1990-1993) to a peak of 1.81 during the 1998-2001 period. This peak aligns with the height of political transition. Following the implementation of integration policies in the 2000s, this coefficient dropped to 1.59 by the 2006-2009 period, quantitatively demonstrating that cross-border collaboration became more accessible over time as political environments stabilized.

Conclusion and Limitations

The study demonstrates that geopolitical boundaries and policy shifts significantly influence the structure of professional networks. However, some limitations exist. The dataset does not distinguish between film and television series, nor does it categorize production types (e.g., purely mainland-funded vs. official co-productions). Furthermore, comparing ERGM coefficients across different time periods is statistically complex due to changing network sizes and structures. Future research utilizing more granular data and advanced temporal network models could further refine our understanding of how borders shape human interaction.

Summary updated

Abstract

People always carry out social activities within certain geographic space. The geographic boundary likes national boundaries and provincial boundaries functions as dividing policies, distributing administrative resources, defining the scope of application of laws and regulations or even the boundary of cultural region, producing fundamental influence on lifestyle. Network analysis is a powerful tool to abstract complex system. In this article, we used network method to understand changes of cooperation network in movie industry and effect of geographical boundary.
BibTeX
@misc{zhou2018border,
  author = {Wen Zhou and Dong Zhang and Chen Chen and Dianbo Liu},
  title = {{Border Effect of Complex Network: An analysis on the cooperation network of movie stars across different regions}},
  year = {2018},
  eprint = {1812.09657},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1812.09657},
}
Figure