Transnational Transition to Adulthood: New insights from social media data

社交媒体数据代表了新的数据源，为社会科学提供了新的机遇（和挑战）（Savage & Burrows 2007; McCormick et al. 2015）。该项目将调查Facebook作为人口统计研究的潜在数据源所提供的机会。特别是，我提出了一种社交媒体数据应用程序，该应用程序通过利用个人在Facebook（以及其他社交媒体）上留下的帖子的数字痕迹来研究欧洲的移民运动。正如Raymer等人（2013）所述，欧洲的国际移民数据“在可用性、定义和质量方面都不一致”。欧洲国家确实用不同的系统和设计分别收集移民数据。我将根据他们在Facebook上指定的地点的变化，估算欧洲国家之间的人口流动（按性别、教育程度等）。这样的估计将补充现有但不准确的官方移民流量估计。该项目还将通过记录“向成年过渡”的出现及其驱动因素（与工作或学习相关的原因），对社会人口学文献做出理论贡献。我建议利用贝叶斯推理，这将允许估计具有层次结构的复杂模型的参数。它还将为参数和其他感兴趣的量提供不确定度的度量。统计模型的贝叶斯分析将需要应用并可能扩展尖端的计算方法，也可能导致使用R和Python等语言开发新的软件，这些软件将可转移到其他应用程序中。

Social media data represent new data sources providing new opportunities (and challenges) for theSocial Sciences (Savage & Burrows 2007; McCormick et al. 2015). The project will investigate theopportunities offered by Facebook as a potential data source for demographic research.In particular, I propose an application for social media data which studies migration movements inEurope by exploiting the digital traces that individuals leave as posts on Facebook (and potentially othersocial media). As it is stated by Raymer et al. (2013), international migration data in Europe are"inconsistent in availability, definition, and quality". European countries do collect separately migrationdata with different systems and designs. I will produce estimates of the flows of people (by gender,education etc.) migrating between European countries based on the change in the location that theyspecify on Facebook. Such estimates will complement existing but inaccurate official estimates ofmigration flows. The project will also have a theoretical contribution to the socio-demographicliterature, by documenting the emergence of a 'transitional transition to adulthood' and its drivers (workor study-related reasons). I propose to utilise Bayesian inference which will allow estimation of the parameters of a complexmodel with hierarchical structure. It will also provide measures of uncertainty for the parameters andother quantities of interest. Bayesian analysis of statistical models will require applying and possiblyextending cutting-edge computational methods, and may also lead to development of new software inlanguages such as R and Python, which will be transferrable to other applications.