The identification of urban relative poverty based on mobile phone metadata:A case study of Guangzhou
China won the battle against poverty in 2020,signifying the resolution of absolute poverty in rural areas,but the relative poverty phenomenon in urban areas persists.How to solve the intricate problem of relative poverty within cities,as well as to alleviate the problem of unbalanced and inadequate socio-economic development?It will become an important topic in the next stage of poverty research.The prerequisite for eliminating urban relative poverty is the accurate identification of urban relatively poor individuals as well as their concentration spaces.With the rapid development of Internet technology,multi-scale and multi-type open big data,such as remote sensing images and street view images,contain integrated information reflecting the built environment,improving the identification accuracy of research in poverty space.However,there are some limitations in accurately identifying relatively poor individuals,such as the difficulty of image data in revealing the socio-economic characteristics of poor individuals.Mobile phone data record information on the socio-economic attributes of urban residents,which can identify the urban poverty space more precisely by locating relatively poor groups.Therefore,taking Guangzhou as an example,this paper combined mobile phone metadata with questionnaire data to identify the relative poverty space within the city.Starting from individual poverty,this paper collected information on mobile phone usage,income and consumption of urban residents through a questionnaire survey.Then,this paper analyzed the correlation between individual mobile phone usage and economic poverty,establishing a poverty index as the dependent variable based on income and consumption data,and using mobile phone usage indicators as the independent variables.After that,a multiple linear regression model between mobile phone indicators and economic status was constructed.Finally,based on the mobile phone metadata,the poverty index of each town or sub-district was calculated to identify the relative poverty space within the city.In addition,based on the analysis of the model,this paper explored the differences in mobile phone usage behavior across urban spaces as well as the reasons behind them.A total of 233 valid questionnaires were collected in this study,and the representativeness of the sample was verified by analyzing the income structure of the questionnaires.In the multiple linear regression model constructed based on the questionnaire data,six significant mobile phone indicators-mobile phone cost,Pinduoduo usage time,online shopping time,office and study APP usage time,takeaway APP usage time,and mobile game duration were selected through the stepwise regression analysis.The model indicated that the mobile phone variable group explained 21.9%of the dependent variable.After obtaining the poverty index,the dependent variable of the model,for each town and sub-district in Guangzhou based on the model and mobile phone metadata,it was divided into five classes-lowest income,lower income,middle income,higher income and highest income for further analysis.The spatial distribution of relative poverty reveals that Tianhe District and the eastern part of Yuexiu District are the areas with relatively high economic levels,while the northern part of Nansha District,the eastern part of Zengcheng District and the southern part of Conghua District are relatively poor areas.Moreover,it is found that the relatively poor space in the central city is concentrated on the border of the Yuexiu,Liwan and Haizhu districts.Identifying poverty through mobile phone usage helps generate new insight into relative poverty.At the individual level,higher income groups have higher mobile phone costs and more diverse digital consumption,while lower income groups have a lower frequency of mobile phone use.The differences in mobile phone usage reveal the existence of information deprivation,which comes from differences in information-seeking abilities and habitual preferences between groups.The digital divide is one of the many barriers faced by relatively poor groups in the mobile Internet era due to the different development levels of information technology.At the spatial level,for our study area,there is a spatial distribution pattern of high center and low periphery,and the relative poverty spaces in the central city are concentrated in the old residential areas,urban villages and the old industrial zones.In general,the distribution of poverty inferred by our model is generally consistent with that based on traditional census data,which also reflects the feasibility of identifying relative poverty through mobile phone usage.This study verified the feasibility and applicability of mobile phone metadata in the field of poverty identification,and enriched the methodological system of spatial identification for urban relative poverty,which is of practical significance for the targeted promotion of poverty governance.However,there are certain shortcomings,such as the low accuracy of mobile phone metadata in identifying sub-districts with complex residential structures and widely varying internal environments.In future research,it can be tried to improve the model by combining mobile phone metadata with other types of big data.
relative povertymobile phone metadatamultiple linear regressionpoverty identificationpoverty space
国家科技期刊平台
NSTL
万方数据
