Venture capital (VC) plays a crucial role in fostering innovation and economic growth, particularly for startups. Understanding the factors driving VC investment decisions is essential for regional development. While existing research has explored investor and investee characteristics, there's a gap in understanding the relative importance of city-level factors. This study addresses this gap by examining the influence of various city-level factors on intercity VC investments in China, a significant player in the global VC market. The study leverages data from 2018 (to mitigate the effects of the 2019 COVID-19 pandemic) from the CVSource database, focusing exclusively on intercity investments to avoid local bias. The research utilizes two distinct analytical methods: Multiple Linear Regression (MLR) and Random Forest (RF), to capture both linear and non-linear relationships between city-level factors and VC investment flows. The combination of these methods provides a robust and comprehensive assessment of the relative importance of various factors.

Existing literature highlights various factors influencing VC investments, categorized into investor characteristics (reputation, network, experience), investee characteristics (management team, technology, innovation), and external factors (economic conditions, financial markets, innovation ecosystem, location, policy, and cultural factors). Studies have employed diverse methodologies, including interviews, surveys, and regression analyses to explore these factors. While some research has identified specific influential factors, a comprehensive analysis comparing the relative importance of city-level factors remains lacking. This study aims to fill this gap by focusing on city-level determinants and employing both MLR and RF to capture a complete picture of the factors influencing intercity VC investments in China.

The study uses data on intercity VC investment deals in China during 2018 from the CVSource database. The dependent variable is the total amount of VC investment received by each city. Independent variables are selected based on existing literature and categorized into five dimensions: economy (GDP, GDP growth rate, tertiary industry share), finance (number of newly established enterprises, number of VC firms, bank deposits, marketization index), innovation (number of universities, student population, R&D investment, number of patents), location (presence of international airport, airline network centrality, high-speed rail network centrality, distance to major VC centers), and policy (administrative level, presence of economic and technological development zones and high-tech industrial zones, Baidu Index). Logarithmic transformations are applied to most variables to address issues of collinearity and uneven data distribution. The Variance Inflation Factor (VIF) is used to identify and eliminate highly correlated variables (VIF > 10). Stepwise regression guided by the Akaike Information Criterion (AIC) is employed to further refine the MLR model. The Lindeman, Merenda, and Gold (LMG) method is used to calculate variable importance in the MLR model. For the RF model, hyperparameters (mtry and ntree) are tuned using a training set (80% of the data), and variable importance is assessed using the percentage increase in mean squared error (%IncMSE). The performance of the RF model is evaluated on a test set (20% of the data). Finally, spatial heterogeneity is analyzed by comparing variable importance across the eastern, central, and western regions of China.

Both MLR and RF models reveal that economic and financial factors are the most significant drivers of intercity VC investments. Specifically, the number of VC firms (VCF), the number of newly listed enterprises (IPO), and GDP consistently rank among the most important variables in both models. The number of universities, indicating a strong innovation ecosystem, also demonstrates significant influence. The distance to major VC centers (DIS_Center) shows a negative correlation, highlighting the impact of geographical proximity. The relative importance of other variables, such as the marketization index, Baidu Index, and presence of high-tech industrial zones (HTIZ), is lower but still significant. A comparison of the LMG values (MLR) and %IncMSE values (RF), standardized to sum to 100%, shows some differences in the ranking of variables. For instance, in the MLR model, VCF and IPO have a higher relative importance compared to GDP, while in the RF model, GDP has the highest importance. This difference may be attributed to the RF model’s ability to capture non-linear relationships between variables. Spatial heterogeneity analysis reveals that the financial market plays a more crucial role in the eastern region, economic development in the central region, and innovation capacity in the western region. The eastern region's higher importance of financial market factors aligns with the national-level analysis. The relative immaturity of the economy and VC industry in the central and western regions explains the differing factor importance in those areas.

The findings confirm that the attractiveness of a city to intercity VC investment is strongly influenced by a combination of economic strength, a robust financial market, and a vibrant innovation ecosystem. The significant influence of variables like VCF and IPO highlights the importance of a supportive financial environment, including the presence of established VC firms and successful exits through IPOs. The positive impact of the number of universities underscores the value of a skilled workforce and technological innovation in attracting VC investments. The negative impact of distance to established VC centers suggests that, despite advancements in communication and transportation, geographic proximity remains advantageous for VC firms seeking easier access to information and management of investments. The spatial heterogeneity indicates that regional development strategies should be tailored to local contexts. Policies aiming to attract intercity VC investments should focus on building robust financial markets in the East, bolstering local economies in the Central region, and fostering innovation in the West.

This study contributes to a deeper understanding of the factors driving intercity VC investments by using both MLR and RF to assess the relative importance of city-level attributes. The findings underscore the importance of economic development, a strong financial environment, and innovation capacity, while acknowledging regional variations. Future research should incorporate time-series data to examine dynamic changes in factor importance and investigate the interaction between city-level factors and VC firm and entrepreneur characteristics.

The study uses cross-sectional data from 2018, limiting the ability to capture temporal dynamics in factor importance. Future research should incorporate time-series data to address this limitation and include unforeseen events, such as economic crises or pandemics, to capture their influence on VC investment decisions. The focus on city-level characteristics neglects the influence of VC firm and entrepreneur attributes. Future studies should incorporate these factors to provide a more comprehensive understanding of the drivers of VC investment decisions.