A study on the public attention in the Belt and Road Initiative

： This paper constructs residents' attentiveness index of the Belt and Road Initiative by big data, and combines social and economic factors to study the spatial and temporal distribution of initiative attentiveness, we can find that （ 1 ） the attentiveness of the initiative will be greatly affected by large government conferences, and there is a mismatch between the areas covered by the initiative and the areas. （ 2 ） both per capita GDP and the proportion of fiscal investment in GDP have a significantly positive impact on the initiative attentiveness of residents, while CPI has a significantly negative Population size and unemployment rates do not have a significant impact on the initiative's attentiveness. (3) the impact of initiative attentiveness has obvious regional differences. This paper breaks the habitual cognition of the attention of the initiative, and combines the social and economic conditions, and puts forward some suggestions, such as expanding the boundary of the Belt and Road, taking multiple measures to mobilize the social forces to build the Belt and Road, and treating the different policies.


1.Introduction
The Belt and Road initiative, which inherits the civilization of the Silk Road for a thousand years and initiates national rejuvenation in the new era, is the common voice of China and countries along the Silk Road for intercommunication and friendly exchanges, and an important initiative of China to promote the restructuring of the global value chain (Dai, 2019) [ 1 ] , which has attracted strong attention both at home and abroad.In September and October 2013, Chinese President Xi Jinping proposed the "Silk Road Economic Belt" and the "21st Century Maritime Silk Road" during his visits to Kazakhstan and Indonesia.In recent years, in order to promote communication and exchanges among countries and regions along the Belt and Road, governments at all levels have focused on the Five-Pronged Approach, a number of policies have been introduced.As of January 6, 2023, China has signed more than 200 documents with 151 countries and 32 international organizations in the area of building the Belt and Road together.In the past seven years, the Belt and Road initiative has been actively practiced, and the influence of the initiative has been actively practiced and its influence is growing.However, there are still doubts in the international community about the topic of cooperation-centered initiatives proposed by China.How to evaluate and analyze the global attention of the Belt and Road initiative, grasp the appropriate strength of voice in global governance (Qu, 2019) [2] , and build a new pattern of diversified and accurate discourse communication (Wan, 2020) [ 3 ] , is an important proposition in the new era.
However, considering that the vast regions covered by the Belt and Road Initiative are significantly different in many ways.We focused our research on the attention of provincial residences to the Belt and Road Initiative within China.
Research and analysis is a direct way to evaluate policy concerns.Gulnar Govli (2018) conducted a study on the awareness of the Belt and Road among 300 people in Kashgar, and found that people's awareness of the Belt and Road initiative was relatively low and their attention was poorly sustained, but most people expressed a willingness to learn more about and support it [4] .Ru (2018) studied the current situation of cross-cultural communication in the context of the Belt and Road documentary, and found that international students have limited recognition and doubt about the credibility of the image of China portrayed in the documentary, and then explored the issue of cultural exchange and communication in the context of value pluralism [5] .
The news media, think tank studies and search engines are indirect ways to evaluate and analyse policy concerns.
The following studies have been conducted from the perspective of the news media: Wu (2014) analyses Philippine society's attitude towards the initiative and cooperation in the socio-economic field by comparing the similarities and differences between the positions of Chinese and Filipino media on the Belt and Road and territorial disputes [6 ] .Qin(2019) extracted the focus of policy attention through a textual keyword clustering analysis of the Belt and Road news topics on the Chinese government website [ 7 ] .In order to detect the public opinion of Central Asian countries on the Belt and Road initiative in a timely and accurate manner, Xi (2020) used the TEM model to extract the subject words of news documents and selected representative documents to assist in the interpretation of news topics [8] .
Research on the Belt and Road initiative from the perspective of think tank research includes: Tian (2020) conducted a survey and analysis of major think tank research in the United States and found that there is an overall bias in the perception of the Belt and Road initiative by American think tanks.They generally regard the Belt and Road Initiative as a strategy and a tool for China's expansion, and raise many questions about its future [9] .In an analysis of academic concerns related to the Belt and Road Initiative, Ren (2020) found that the communication effects of the Belt and Road Initiative are clearly lacking and that the theoretical construction of the Initiative needs to be further strengthened [10] .Wang (2019)  studied the topics related to the Belt and Road through literature data from China Knowledge Network and Web of Science databases, using methods such as co-word analysis and cluster analysis [11] .Wang (2019) [12] have also analyzed the academic concerns of the Belt and Road Initiative based on literature.
Research on the Belt and Road Initiative from search engines includes: Zhang (2015) studied the attention and selection tendencies of mainstream media in Xinjiang and Fujian based on Baidu search engine data [13] .Feng (2018) used the Baidu search index to study the spatial and temporal distribution of the the Belt and Road initiative, but did not conduct an impact factor analysis [14] .Yu (2019)  found that the Belt and Road was a hot spot for residents in 2016, 2018 and 2019 [15] .
From the perspective of the supply and demand of the system, previous studies on the Belt and Road have focused on optimising the supply and refining the toplevel design, while few have studied the public opinion from the perspective of the demand of the system.Although the questionnaires are relatively accurate in terms of the level of interest in the policy, it is difficult to overcome the disadvantage of a small sample size in dealing with large audience.Although the news media can reflect social concerns in a timely and comprehensive manner, China's traditional media have strong governmental attributes, and their core function is to ensure ideological leadership, management and discourse, focusing on the media's role in guiding policies (Wang, 2019) [16] .In traditional media, residents are only passive recipients of information, and it is difficult to analyse residents' concerns and attitudes towards policies from the perspective of traditional media.The number of studies using search engines as a tool to study the attention of the Belt and Road Initiative is still relatively small.The research is limited to a few keywords such as "the Belt and Road", which does not provide a comprehensive coverage of the concept of the initiative, and is limited to study the socio-economic impact of the initiative.
The search engine data is a suitable quantitative method to evaluate the attention of the Belt and Road.Therefore, the thesis uses the seven keywords related to the Belt and Road Initiative in Baidu search index, and to analyse the inter-provincial spatial and temporal distribution of attention to the initiative in relation to socio-economic factors.The paper's marginal contribution is mainly reflected in the following three aspects: (1)  Using the big data method, crawl the residents' search behaviour data on the concepts related to the Belt and Road Initiative from Baidu's search engine.(2) A factor analysis of the Belt and Road Initiative attention from various socio-economic factors reveals that the Initiative is significantly influenced by financial investment and government action.
The remainder of the thesis is organised as follows: Section 2 presents the model setting, data sources, and descriptive statistics; Section 3 conducts panel data regressions, conducts heterogeneity analysis, discusses endogeneity and robustness tests, and explains the impact mechanisms; and the last section presents conclusions and policy implications.

Study design 2.1 Measurement model setting
Panel data is a record of changes in the object of study in both the temporal and spatial dimensions, extend the sample size and provide a higher degree of freedom for statistical tests, provide richer information on more individual (or unit), and effectively address the problem of omitted variables (Wooldridge, 2002) [17 ] .As a result, panel data models have become a common econometric model for exploring macroeconomic social phenomena.Combined with the research hypotheses underlying the theory of this paper, the panel model for the benchmark can be set as follows.
Where BRA is the attention of the Belt and Road initiative in 30 provinces in China, where  is the natural logarithm of per capita GDP,  is the share of fiscal input in GDP,  is the consumer price index,  is the natural logarithm of population,  is the unemployment rate; i is the 30 provinces in China (excluding Hong Kong, Macao, Taiwan and Tibet), and � � ����� ������ .

2.2.1，The Belt and Road initiative attention
According to the 2019 CNNIC China Internet Survey, China's Internet users have reached 829 million, and Baidu is the largest search engine in China, accounting for 76.42% of China's search engine market in 2019, with the number of users covering most of the Internet-using population.Baidu has started to count the attention index since 2011, based on the search volume of web search keywords.Baidu search index can record the search volume of users and media in PC and mobile, which can effectively represent the attention preference of China's residents, and is an important reference for public opinion, as well as a data source for many policy attention studies.Therefore, the thesis collected the 7 keywords provincial data through the Baidu search index between 2011-2019.

2.2.2， Explanatory variables
Combining previous studies, it can be found that socioeconomic factors such as government financial investment, CPI, population size and unemployment rate can affect the residents' attention to the initiative.In terms of socioeconomic factors, residents with a lower quality of life are more likely to work for a living and are less likely to be sensitive and concerned about policies, while those with a higher quality of life are more likely to be concerned about current affairs, hence the CPI is chosen as the explanatory variable.In addition, the population size of a region and the employment status of its residents are also likely to have an impact on residents' attention to initiatives, so an attempt was made to include the population size and the unemployment rate as explanatory variables.

Data sources
Data on the attention of residents for the seven keywords included in the "Belt and Road" initiative were obtained from the Baidu index.To examine the influencing factors of public attention, the thesis collected variables of provincial economic and social development from 2011-2019.Specifically, regional economic development through GDP per capita, local fiscal expenditure as a proportion of GDP, consumer price index CPI, population size and unemployment rate data were sourced from wind information.To better reflect the consistency of the data indicators, GDP per capita and population size are taken as pairs.

Analysis of factors influencing the level of interest of residents in the Belt and Road Initiative
In order to analyse the influencing factors of the initiative attention, the thesis constructs a panel data model with socio-economic variables for 30 provinces.Firstly, factors that may have an impact on the level of interest in the initiative are added to the model step by step to find the explanatory variables with higher significance, and then the model with the best fit is identified.The results of the model are then interpreted through panel quantile regression for different sample distributions.Finally, an endogenous discussion of the model and robustness checks were carried out.

Factor analysis based on panel data
The stepwise regression results of the baseline model are shown in Table 3.As can be seen from the results in Table 3, the model with per capita GPD, fiscal investment as a share of GDP and CPI as explanatory variables has a higher goodness of fit and is able to explain the Belt and Road Initiative relatively well.The model with the addition of population and unemployment variables does not have a significantly better fit and the significance of the additional variables is not significant.In terms of specific Belt and Road Initiative concerns.
Holding all other parameters constant, for every 1% increase in GDP per capita, initiative attention increases by 3.12 units, with the parameter significant at the 99% confidence level.This indicates that an increase in GDP per capita output has a substantial boosting effect on the increase in policy concern, and that the positive effect is statistically significant.Combined with the quantile regressions in Table 4, it can be seen that the effect of this factor on initiative concern has diminishing utility as GDP per capita increases.This suggests that the higher the economic contribution per capita, the more likely to pay attention to the policy by residents.In China, it has long been a consensus that policy has a significant impact on the economy.The Belt and Road Initiative is a huge business opportunity for individuals and companies.In terms of finding and seizing opportunities, we can see that groups with greater capita GDP generation capacity are more concerned about initiative and more able to take full advantage of it.Policies that are uniform across the country may be interpreted and applied differently in the east and west, which in part explains the real dilemma that the gap between regions is more likely to be widened.
In terms of fiscal expenditure as a share of GDP, if all other parameters are held constant, each 1% increase in fiscal expenditure as a share of GDP increases the initiative's attention by 10.659 units, with the parameter significant at the 99% confidence level.Combined with the quantile regressions in Table 4, it can be seen that the effect of this factor on initiative attention has increasing utility as fiscal expenditure increases as a proportion of GDP.This indicator shows government investment largely responsible for attracting public attention to the Belt and Road Initiative, and the Belt and Road Initiative system is still relatively dependent on the government's financial drive.
In terms of CPI, if all other parameters are held constant, each unit increase in CPI is associated with a 7.95 unit decrease in initiative concern, with the parameter significant at the 95% confidence level.Combined with the quantile regression in Table 4, it can be seen that as the CPI increases, the effect of this factor on concern for the initiative has a reverse incremental utility, but the significance of the parameter is limited.This may be due to the fact that at higher inflation, the standard of living of the population is affected.The more life stressful, the easier for people to ignore sensitivity for policy.
In terms of population size, the impact is not significant.Since within 2011, the development of the internet has allowed the public to use internet access devices such as smartphones and computers in a easier way, so internet conditions are basically no longer a limiting condition for residents to pay attention to the policy.The index of concern, as an indexation product, also removes to some extent the effect of initiative concern due to the size of the population.In terms of unemployment, it can be seen that the regression coefficient between unemployment and initiative concern is not significant, indicating that employment status is not a significant indicator of policy concern.

Discussion of the heterogeneity of the model
In order to further discuss the heterogeneity of the influencing factors of the Belt and Road initiative, the thesis discusses the attention of the initiative in different socio-economic environments by comparing whether the initiative is along the Belt and Road route or not, and by dividing it into three geographical regions: The analysis of the factors along or not along the route shows that all the influencing factors of the initiative are significant, but the absolute values of the coefficients of the influencing factors in the regions along the route are larger than those in the provinces not along the route, reflecting the higher socio-economic impact elasticity of the provinces along the route compared to the provinces not along the route.The analysis of the factors for the three regions, East, Central and West, shows that there is a significant difference between the three variables of GDP per capita, financial investment as a proportion of GDP and CPI in terms of the level of interest in the Belt and Road Initiative.
The positive effect of GDP per capita and fiscal investment to GDP is significantly greater in the east than in the central and western regions, while the western region lags behind the central and eastern regions, and there is a clear imbalance between the east and the west in the CPI parameter, with the eastern region having a greater negative impact on the initiative, followed by the central region and the western region having the lowest.To a certain extent, this parameter also indicates that the eastern region is more resilient in terms of policy concern due to the pressure of consumption, while the central and western regions have relatively low consumption levels and the negative impact of the price index on policy concern is relatively low.

Endogeneity Discussion and Robustness Tests of the Model
There may be a degree of endogeneity in the analytical model of the factors influencing initiative concern for the following reasons: (1) There may be a reciprocal causal relationship between some of the independent variables and the dependent variable.For example, the higher the GDP per capita, the higher the initiative concern of residents may be.On the other hand, initiatives represent certain development opportunities, and attention to initiatives will in turn contribute to an increase in the creation of GDP per capita in residential areas.Thus, in the model, GDP per capita may have a causal relationship with initiative concern.(2) Omitted variables.From the analysis of the internal and external factors of the Belt and Road Initiative, the Initiative may be influenced by both exogenous and exogenous factors.However, the analysis of socio-economic factors in the province may not cover all factors such as the level of education per capita and the level of internet usage, so there is a possibility of omitting variables.
To further verify the robustness of the model, the thesis conducts robustness tests through group sampling regression and replacement regression methods.Firstly, the concern is divided into three groups of high, medium and low concern, and the data from the provinces with medium concern are selected for regression can be obtained from the model (16), and it can be seen that there is no significant change in the parameter coefficients as well as significance of the model compared with the overall regression in the previous paper and the regression divided into three regions, East and West.Roodman (2009)  proposed the use of linear GMM, which can effectively avoid reducing the validity of Hansen's test and is more suitable for dynamic estimation of short panel models [18] , hence the robustness test of the linear GMM model method chosen for model (17).The fitted results from the three GMM models show that the regression coefficients and their significance are also relatively similar to the previous fits under different methods and different regions, and they are all highly significant.Therefore, the robustness tests indicate that the models are relatively stable and are scientifically explanatory.

Conclusions and policy recommendations
Combining the distribution of mechanisms, the study of the spatial and temporal distribution of initiative concerns and the analysis of factors, the following conclusions can be drawn.
(1) The Belt and Road Initiative attention shows its distribution has multiple characteristics.Compared to the economically developed coastal regions, residents along the Belt and Road do not pay more attention to the initiative, and are less likely to make full use of it to develop themselves and narrow the gap between regions.Therefore, it is an important proposition to mobilise the attention and participation of less economically developed regions in the policy formulation process.
(2) The Belt and Road Initiative relies more on government investment for its attention.China has achieved good results in promoting the Belt and Road Initiative through major conferences and financial investments.However, the government is playing a leading role in this process.This solution is reasonable when the government takes the initiative and creates the right infrastructure in the initial implementation.However, the full effectiveness of a policy or initiative also depends on the widespread interest and deep involvement of the members along the route.
(3) The price level is an important factor influencing the residents' lives and initiative concerns.From the findings of the thesis, it can be seen that the consumer price level could affect the initiative concerns of the residents, and may further affect the political participation behaviour of the residents.It is important to maintain price stability and safeguard the living standards of the people in a good socio-economic environment.
The above findings imply that attention to the Belt and Road Initiative is important to its implementation, and that attention to the Initiative is influenced by local socioeconomic factors.In order to effectively enhance the attention to the Initiative and promote its overall coordination, the following recommendations are made.
(1) The Belt and Road Initiative should break the limitations of geographic space and involve the participation of more subjects and larger regions.All provinces and regions in China can participate extensively in the construction and development of the Belt and Road, and each region should combine its own advantages, seize the opportunity to combine with the general policies of the country, participate in the Belt and Road initiative.
(2) The Belt and Road Initiative is an all-round policy system promoted by the central government and widely responded to by governments at all levels, but this government system should not just be a call from the central government, it requires local governments, local enterprises and residents to put forward richer and more specific implementation plans, taking into account their own endowments.The government should strive to shift the design of the Belt and Road policy system from one supported by fiscal policy to one that combines fiscal and monetary policies, and make extensive use of various methods and tools such as taxation, transfer payments and differential interest rates to mobilise social initiative and jointly build an ecologically complete policy system.
(3) From the perspective of public opinion, the Belt and Road initiative should be formulated in a way that fully incorporates public opinion information.Policies and initiatives should be developed with public opinion.Building a moderately prosperous society requires the mobilisation of policy attention and participation from the economically underdeveloped regions.This also requires us to promote and guide the policies of the less developed regions, especially those along the Belt and Road, so that local people can learn more about the policies of and use them to improve their living standards and achieve healthy and coordinated socio-economic development.

Table 1
Descriptive statistics of variables

Table 5
Comparison of results of concern for initiatives along or not Robust t-statistics in parentheses *** p<0.01, ** p<0.05, * p<0.1

Table 6
Comparison of concern results by East, Central and

Table 7
Robustness tests of the model