Study on the Level of Talent Attractiveness of the Yangtze River Delta Urban Agglomerations Using Bayesian Quantile Regression Method

Innovation is the first driving force to lead development, and the carrier of innovation drive is talent, and talent has been an important strategic resource to lead social development in the 21 st century. And different levels of cities should focus on the strategy of attracting talents. In this paper, based on the index data of the Yangtze River Delta Urban Agglomerations in 2019, we construct the evaluation talent attractiveness level score and use it to build a Bayesian quantile regression model to map the focus that cities of different levels should focus on talent attraction policies. The results show that the marginal gain of talent attraction is different for different level cities in different dimensions. By quantifying the level of talent attractiveness of different cities through the objective assignment method, we find the differences in the spatial distribution of talent attractiveness of different cities in the Yangtze River Delta region and provide theoretical guidance for the integration of the Yangtze River Delta. At the same time, by exploring the differences in the talent attractiveness of cities of different levels that should be focused on, we find the general rules and provide reference and guidance for other cities.


Introduction
The 21 st century is the era of the knowledge economy, the economic form has changed to knowledge-based and information-based. Modern theoretical re-Theoretical Economics Letters search has also begun to recognize the importance of human capital to urban development. Lucas & Rossi-Hansberg (2002) have shown that the concentration of human capital drives economic growth by increasing local productivity.
The development of cities cannot be separated from the concentration of industries and human capital, and the concentration of talents undoubtedly greatly improves the development efficiency of cities. A country or a city with sufficient talent reserves will have a long momentum of economic development, so the research on talent attraction becomes especially important. In the previous research on talent attraction in cities, due to the data acquisition and the different development status of each city as well as the natural environment and other factors, most of the research objects choose a single big city, while a city needs to effectively develop policies to attract talents, most of them need to find their own positioning according to their actual situation, and the policies introduced should be focused according to their own situation to ensure their own competitive advantage in talent attraction to a greater extent, to ensure their competitive advantage in talent attraction. Therefore, the study of talent attraction in cities with different levels of development is a worthwhile problem to investigate the direction of talent attraction policies in cities with different levels of development and to provide a reference for cities whose policies are affected by the lack of data guidance, and the quantile regression method can be a good solution to this problem, by selecting suitable cities, the quantile regression method can portray to some extent the level of talent attraction in cities with different levels of development. As one of the six major city clusters in the world, the Yangtze River Delta city cluster is not geographically heterogeneous and contains 5 first-tier cities, 9 second-tier cities, 9 third-tier cities, 3 fourth-tier cities, and 1 fifth-tier city, so this paper selects the Yangtze River Delta city cluster as the object of investigation to explore Therefore, this paper selects the Yangtze River Delta city cluster as the object of investigation to explore the direction that cities at different levels of development should focus on to improve the level of talent attraction.

The Importance of Talent and the Factors that Affect It
The early academic research on talent is mainly about talent mobility, the traditional talent mobility models include (March & Simon, 1958) decision participation talent mobility model and (Mobley, 1982) intrinsic mobility relationship analysis model, etc., while the research on talent mobility gradually expands to talent attraction, exploring the impact of economic and social environment on talent clustering. Porter (1998) pointed out the influence of industrial clusters on talent attraction through a comprehensive and systematic analysis of talent concentration, Audretsch & Feldman (2004) pointed out that if a country or a region has well-developed basic industries, it must have strong competitiveness in Y. Zhang et al. DOI: 10.4236/tel.2021.116072 1142 Theoretical Economics Letters talent attraction. Palivos & Wang (2008) argues that factors that increase the attractiveness of talent include not only the internal and external large-scale market economy, local government incentives, but also salary levels, public resource availability, and the effects of knowledge outflow and spillover. Qian (2010) examined two types of talent: human capital in terms of educational attainment and innovation class in terms of occupational skills, in terms of the relationship between the geographical distribution of talent in China and its innovation, entrepreneurship and regional economic performance. Correlation and multivariate analyses found that universities are an important factor influencing the distribution of talent, and that wage levels, service facilities, and openness also contributes to attracting talent, but to varying degrees. Weng & McElroy (2010) used one-way ANOVA, hierarchical regression and structural equation modeling to analyze the data. The results found that there were significant differences in the regional attractiveness of the four industry clusters that factors in the macro HR environment directly influenced talent growth and regional attractiveness, and that talent growth partially moderated the relationship between the three levels of the HR environment. It explains why talents are more willing to gravitate toward industry clusters and how the HR environment affects the flow of talents. Gao (2012) constructed an evaluation index system for talent attraction in Shanghai from three perspectives: social and cultural atmosphere, institutional environment and living environment. It provides a theoretical basis for the adjustment of talent policy and the shift of focus in Shanghai in the 12 th Five-Year Plan period. Liu et al. (2015) and others take Shenzhen as an example to explore the factors influencing the attractiveness of financial talents in Shenzhen, which is of reference. Using the research method of rooting theory, the results of the study show that the city's financial industry environment, cultural atmosphere, living environment, and talent policy have a significant impact on the attractiveness of financial talents in Shenzhen, and the financial industry environment is the fundamental influencing factor of the attractiveness of financial talents in the city. Finally, several suggestions are proposed to enhance the attractiveness of Shenzhen to financial talents. Song & Chen (2006)  The results confirmed that the talent attractiveness of cities in the case of "with" high-speed railroad was significantly higher than that in the case of "without"

Research on Different Cities
high-speed railroad, and Wuhan was the city that benefited most from high-speed railway among the three types of cities, followed by Zhuzhou and Guangzhou.

Variable Selection and Description
To measure the attractiveness of a city's talent, we need to take a comprehensive consideration from the perspective of what the talent needs. Through the summary of previous literature, in this paper, the talent needs to consider the natural ecological environment, the level of economic development, the level of public services, the scientific and technological living environment and the scientific and technological innovation environment of the city when choosing a city to settle, and start to design indicators and build an evaluation system, as shown in Table 1. In the selection of methods to quantify the level of talent attractiveness of cities, considering the subjective assignment methods bring subjective correlation error, so this paper selects two objective assignment methods entropy method and principal component method to compare the results horizontally, the specific steps of the entropy method are as follows.
First, standardize the raw data and process the positive and negative indicators separately: Here, ij x is the value of the jth indicator of the ith city, The fourth step is to calculation of information entropy redundancy: The fifth step is to calculate the weight value of each indicator: The sixth step is to calculation of the Talent Attractiveness Index for each city: By calculating the above steps, the talent attractiveness index of each city in the Yangtze River Delta Urban Agglomerations was obtained and ranked, and the results are shown in Table 2. Table 2 shows that most of the top-ranked cities are first and second-tier cities and the lower-ranked cities are mostly alternating between third and fourth-tier cities, which also show that talent attraction is closely related to the comprehensive development level of a city. In terms of spatial distribution, Shanghai ranks first, while the cities in the other three provinces in Anhui Province  The specific steps of the principal component method are as follows.
First using the characteristic roots, and their corresponding unit eigenvectors, find the load matrix is: The final composite evaluation model, the larger the model, the * Z higher the evaluation level: The data were tested by KMO value test and Bartlett's sphericity test and the results are shown in Table 3.
The KMO value is greater than 0.8 and the Bartlett test p-value is less than 0.01, which is suitable for principal component analysis. The talent attractiveness index of each city in the Yangtze River Delta Urban Agglomerations is derived and ranked by the principal component analysis method, and the results are shown in Table 4.
From the above table, it can be seen that the ranking order derived from the principal component method does not differ much from the results derived from the entropy method, which can also support the feasibility of the constructed index system evaluation.

Methods
Because it is going to explore the trend of talent attraction in different levels of cities subject to different factors, Quantile regression was first proposed by (Koenker & Bassett, 1978), and quantile regression can be a good solution to such problems by portraying the different effects of independent variables when   There are many ways to minimize the loss function, the common linear programming generally has difficulties, while the Bayesian approach is easier and more robust in nature, so we use the Bayesian quantile regression method, which has the following basic steps: first, assume that the quantile regression model error term obeys the asymmetric Laplace distribution, find out the likelihood function, and then obtain the sampling distribution of the regression parameters based on the Bayesian method of MCMC algorithm, and then perform the convergence test on the sampling results.

Empirical Analysis
By  Table 5. Figure 1 shows the sampling trajectories of the natural ecosystem coefficients at each quantile and the posterior probability density plots.
The first row of Figure 1 shows that the sampling values of the natural ecological habitat coefficients fluctuate around a certain numerical, indicating that the Markov chains constituted by sampling the natural ecological habitat coefficients at each quantile converge, which can be used as a basis for the significant results obtained from Bayesian quantile regression, combined with Table 5, which shows that the coefficients of Bayesian quantile regression are significant at each quantile.
From the regression results and the graph of coefficient changes (Figure 2), it can be seen that the coefficient of the natural ecological environment is always greater than 0, and the trend is slightly weakened as the quantile increases, but the overall trend is not significant, indicating that the talent attractiveness of cities with different development levels is closely related to the natural ecological environment, the better the environment, the higher the level of talent attractiveness and the marginal gain of natural ecological environment on attractiveness is weakened as the development level increases, but the overall change is not significant; the economic development level is greater than 0 in all quantile points, but the gap between high and low quantile points is larger, and the overall trend is decreasing. The economic development level is greater than 0 in all  Note: ***, ** and * indicate that the test was passed at the 1%, 5% and 10% significance levels, respectively, and the numbers in the parentheses represent the standard error.
Y. Zhang et al. DOI: 10.4236/tel.2021.116072 1151 Theoretical Economics Letters quartiles, but the difference between the high and low quartiles is larger, and the overall trend is downward, indicating that the economic development level is a positive gain effect on the talent attractiveness, but the improvement of the economic level in cities with low development level is more significant than that in cities with high development level; the high quartile of the public service level coefficient is different from the low quartile, and the overall trend is upward, indicating that The overall trend is upward, which means that the improvement of public service level has a positive effect on the attractiveness of talents, but the improvement of social life environment in cities with low development level has a relatively lower effect on the attractiveness of talents compared with cities with high development level. The coefficient is higher than other aspects, which indicates that the environment of science and technology development is extremely important in the talent attractiveness of cities of all development levels.     Note: ***, ** and * indicate that the test was passed at the 1%, 5% and 10% significance levels, respectively, and the numbers in the parentheses represent the standard error. to the number of registered population" and "fixed-asset investment in urban municipal utilities construction (RMB million)"; the social living environment increases the variables "urban residents' consumption price index" and "the proportion of tertiary industry in GDP"; the variables "number of patents granted"

Robustness Test
and "R & D personnel (people)" are added to the environment of science and technology innovation.
From the results in Table 6 above, we can see that the positive and negative coefficients of each variable at different quartiles and the trend of change are consistent with the previous paper, thus it is reasonable to analyze the influence of different factors on the attractiveness of talent at different quartiles in this paper.

Conclusion
In terms of the spatial distribution of city clusters, Shanghai has the highest level From the results, when talents choose to go to cities with lower development level, the ratio of considering natural environment and economic development is higher than that of going to cities with better development level, while when they go to cities with higher development level, they will pay more attention to the level of public services and social entertainment and cultural environment, so cities with low development level give relatively higher priority to the construction of economy and environment when making policies, while cities with high-level development give the construction priority of services and social and recreational life is relatively higher.
Y. Zhang et al.