*4.3. Correlation between Intercity-MAI and Intercity Population Flow*

*4.3. Correlation between Intercity-MAI and Intercity Population Flow*  To explore the relationship between intercity-MAI and intercity population flow, the results have been shown in Table 8 and Figure 7. As we could notice, the average value of intercity-MAI is 1.00, Guangzhou-Shenzhen has the highest intercity-MAI at 5.19; and Shenzhen-Chengde has the lowest index of 0.04. For the individual urban agglomeration, the intercity-MAI among Beijing, Tianjin, and Shijiazhuang has the highest top three values in BTH. The same level of intercity-MAI To explore the relationship between intercity-MAI and intercity population flow, the results have been shown in Table 8 and Figure 7. As we could notice, the average value of intercity-MAI is 1.00, Guangzhou-Shenzhen has the highest intercity-MAI at 5.19; and Shenzhen-Chengde has the lowest index of 0.04. For the individual urban agglomeration, the intercity-MAI among Beijing, Tianjin, andShijiazhuang has the highest top three values in BTH. The same level of intercity-MAI can be found in YRD for Shanghai, Hangzhou, and Suzhou. In PRD, such level interactions are observed betweenGuangzhou, Shenzhen, and Foshan.

**Table 8.** The Pearson coefficient between intercity-MAI and intercity population flow.


Under the correlation analysis of these two indexes, a moderately positive correlation can be observed in the study area (See Table 8). For the three urban agglomerations, PRD has the highest correlation among them and the correlation in BTH and YRD represents a relatively lower level. There are 59 pairs of cities that have a high-high correlation pattern (high intercity-MAI and high intercity population flow), there into 15 pairs in BTH, 24 pairs in YRD, and 20 pairs in PRD; 147 pairs of cities exhibited the low-low correlation pattern, of which, in BTH, YRD, and PRD, are separately 46, 69, and 32. These two kinds of correlation patterns occupy 75% of the total. Although relevant correlation coefficients of intercity-MAI are relatively limited, it can capture the interaction trend of population flow at an acceptable level.

population flow at an acceptable level.

observed between Guangzhou, Shenzhen, and Foshan.

can be found in YRD for Shanghai, Hangzhou, and Suzhou. In PRD, such level interactions are

**Table 8.** The Pearson coefficient between intercity-MAI and intercity population flow.

**Region Three UAs BTH YRD PRD** 

Sig(2-side) 0.0000 0.0000 0.0000 0.0000

Under the correlation analysis of these two indexes, a moderately positive correlation can be observed in the study area (See Table 8). For the three urban agglomerations, PRD has the highest correlation among them and the correlation in BTH and YRD represents a relatively lower level. There are 59 pairs of cities that have a high-high correlation pattern (high intercity-MAI and high intercity population flow), there into 15 pairs in BTH, 24 pairs in YRD, and 20 pairs in PRD; 147 pairs of cities exhibited the low-low correlation pattern, of which, in BTH, YRD, and PRD, are separately 46, 69, and 32. These two kinds of correlation patterns occupy 75% of the total. Although relevant correlation coefficients of intercity-MAI are relatively limited, it can capture the interaction trend of

**Figure 7.** The scatter plot of intercity-MAI and intercity population flow. **Figure 7.** The scatter plot of intercity-MAI and intercity population flow.

#### **5. Discussion 5. Discussion**

#### *5.1. Evaluation of MAIs 5.1. Evaluation of MAIs*

The massive population migration is the specific phenomenon and the inevitable driving force promoting the urbanization of population in China and many developing countries. The collection of urban MAIs can obtain the public intention of migration based on individual search actions and can offer exploration of population migration. Depending on the MAIs, we analyzed the correlation relationship between local-MAI and external-MAI; a high correlation has been discovered. It implied that the city with relatively higher local-MAI has a higher external-MAI. Migration may be active in the high-high cities, such as Shanghai, Beijing, and Shenzhen. According to the dynamic monitoring survey of China's migration population in 2015, the proportions of floating population inside these three cities separately reached 40.26%, 38.02%, and 67.51%, which are much higher than the average value of China at 18.00%. Besides, they have separately occupied 12.22%, 11.99%, and 8.64% (ranked top 3) of the whole inflow population of the three urban agglomerations, which has the most dynamic The massive population migration is the specific phenomenon and the inevitable driving force promoting the urbanization of population in China and many developing countries. The collection of urban MAIs can obtain the public intention of migration based on individual search actions and can offer exploration of population migration. Depending on the MAIs, we analyzed the correlation relationship between local-MAI and external-MAI; a high correlation has been discovered. It implied that the city with relatively higher local-MAI has a higher external-MAI. Migration may be active in the high-high cities, such as Shanghai, Beijing, and Shenzhen. According to the dynamic monitoring survey of China's migration population in 2015, the proportions of floating population inside these three cities separately reached 40.26%, 38.02%, and 67.51%, which are much higher than the average value of China at 18.00%. Besides, they have separately occupied 12.22%, 11.99%, and 8.64% (ranked top 3) of the whole inflow population of the three urban agglomerations, which has the most dynamic migration in China. The predominant roles of them in attracting population outside the city are declared. Active migration movement can be detected to support the hypothesis derived from the correlation relationship between local-MAI and external-MAI.

Analyzing the correlation of external-MAI with UAM, the reasonability of external-MAI can be verified through the high correlation with conventional statistics analysis. Based on the push-pull theory of migration, in the cities with higher urban pulling force, more influx of population can be observed. Through the calculation of UAM, which depicted urban pulling force, the city with higher external-MAI can observe higher UAM. The feature of external-MAI coincides with the setting of push-pull theory. Further, exploring the relationship between external-MAI and the indexes which reflect urban migration attractiveness, there are significant correlations between tertiary industrial output-value and urban residents' per capita disposable income with the external-MAI in the whole study area. Most of the cities with higher job opportunities and income levels receive more migration attentions from the outside area. This finding coincides with the dynamics monitoring survey of migration population suggesting a migration reason in Table 1 (*work and trade* as the predominant migration reason), which can represent the ability of MAI indexes in capturing the impact of migration reasons. With respect to the indexes described urban living conditions, there are no significant correlations between the population migration attention and unemployment rate or per capita living area, this results may correspond to the great exception of potential migrants for their future urban condition, which can be explained by the Todaro migration model from the perspective of development economics. Todaro migration model argues that the migration of population is based on the "expected profit" of migrants. The hardships in urban life have not obtained enough attention from potential migrants, particularly for the rural-urban migrants who lack the necessary information as they enter a new different world [43]. Further, the more schools a city has, the more public migration attention it receives. The positive correlation existing between the education indexes (the number of primary schools, secondary schools, and universities) and external public migration attention exposes that the educational opportunities also intensify the level of urban migration attention. In PRD, the focus of educational concern only derives from the consideration of secondary schools; significant correlation has not been observed between the number of the other two levels of schools in the area, which may be attributed to the relatively lower education level of Guangzhou Province (the administrative province that PRD belongs to) than the other two urban agglomerations.

Besides, we further adopted the neoclassical theory in population migration to explore the reasonability of MAIs. The per capita disposable income of urban residents, which has been viewed as the direct index depicting the possibility for migrants to improve economic benefit, has been adopted to conduct the correlation analysis with external-MAI; the results have shown that the external-MAI has significant positive correlation with the per capita disposable income of urban residents (with the correlation coefficients 0.650, 0.945, 0.752, and 0.780 separately in the whole study area, in BTH, in YRD, and in PRD). The reasonability of MAIs can be further identified.

#### *5.2. Relationship with Migrants*

With the correlation analysis of external-MAI and urban migration population, we could observe a significantly positive correlation. The resource endowment gap between different urban areas (e.g., economic development level, environmental quality, promotion of opportunities for individuals, etc.) triggers personal develop exception and forges migrant needs in flowing among diverse regions [22,43]. Collecting information about the targeted city by employing the search query engine is an efficient and low-cost approach to supplement requisite information before deploying actual migration to external areas. As noted as the correlation results of external-MAI and urban inflow migrants in the study area, we can accept the hypothesis that the migration-related search queries from individual users were able to positively reflect urban inflow migrants. External-MAI can be applied to reflect urban inflow migrants on the annual scale.

The high correlation between local-MAI and the floating population inside cities was a signal to prove their close relationship. Nowadays, the floating population inner city has become an influential part in enacting urban planning and policy. Generally speaking, the floating population lived with relatively weaker urban amenities than the local population [54,55]. The desire of improving current conditions was more intensive for them, which was delineated by the high demand for new job opportunities, study chance, and the possibility of improving living quality, etc. Driven by such basic needs, the corresponding search query can be brought into the cyber space and raises the high correlation between local-MAI and floating population inside the city.

Intercity migration has already become one of the significant migration models in current China. We analyzed the correlation between intercity-MAI and intercity population flow in 2015; a similar positive correlation can be observed as 0.57 (*p*-value 0.00) in the whole area. The results show that the representativeness of intercity-MAI for population flow between different cities was effective, but the correlation relationship was relatively limited. It might be caused by two main reasons: (1) The selection of search keywords cannot cover every reason for migration flows. A unique keyword system may exhibit some deviation in reflecting the driving force of every population flow interaction; (2) migration movement has a lagging feature. It may happen a few months, years, or a much longer

time after the search action. It also may be canceled or indefinitely delayed after information acquisition through searching, which makes the relatively lower correlation between intercity-MAI and intercity population inflow. Generally speaking, the correlation between intercity-MAI and population flow is still on an acceptable level. It can be a supplement for the population flow research of insufficient data. In future work, the construction of a targeted search keywords system for objective population flow can be adopted to remedy such drawbacks.

Besides, for the three MAI indexes, the different correlation coefficients in the three urban agglomerations revealed that regional disparity exists. We calculated the variance (VAR) and coefficient of variation (CV) of relevant correlation coefficients of three MAI indexes, as shown in Table 9. It can be seen that all the VARs are lower than 0.01 and that all the CVs are lower than 10%. The robustness of external-MAI, local-MAI, and intercity-MAI in reflecting population can be partly ensured in the study area.


**Table 9.** The variance (VAR) and coefficient of variation (CV) of different types of MAI.

Furthermore, we have tested the significance of slope of the three trend lines, which were separately fitted based on external-MAI and inflow population, local-MAI and floating population, and intercity-MAI and population flow to identify whether MAI indexes could steadily reflect the migration situation in different urban agglomerations. All the significances of slopes have been rejected by significant testing at a significant level of 0.05 (sig = 0.43, 0.19, and 0.86 for external-MAI, local-MAI, and intercity-MAI). The null hypothesis could be accepted as there is no significant deviance between the slopes. Although the representations of MAI are diverse in different regions, the deviances are nonsignificant.
