of coordinated efforts, which must comply with respondent confidentiality and privacy requirements.
Nonsurvey Data Collection
Multimodal data collection, involving complements and substitutes for traditional government surveys, is necessitated by the fact that much of what is interesting and important about social capital takes place at the level of neighborhoods or communities, where general population surveys need to be augmented or, in some cases, replaced by data sources that allow for more targeted studies.
It has become commonplace to emphasize the potential—for solving problems in government, the private sector, and in scientific research—of the ever-growing volume of data created and captured digitally. Some kinds of information, such as the structure and density of people’s online relationships and connections or their patterns of cellphone communication, are next to impossible to discern using conventional survey methods. However, while alternative data collection and analysis methods are no doubt flourishing, establishing the statistical validity of estimates based on “big data” sources is in its infancy. In addition, most unstructured digital data are generated by and owned by private sector entities where models for methodological transparency and privacy and confidentiality protection are undeveloped. These are but two reasons, among several, that a survey-centric approach—for which problems of data accuracy, quality, representativeness, and confidentiality have largely been contained or solved—will continue to play a central role in social science research for the foreseeable future.
Beyond social media, private-sector data generated by people’s purchasing and other online activities and by automated payroll systems has created private-sector alternatives (or, in some cases, complements) to such key economic indicators as the Consumer Price Index (e.g., the Web-based MIT Billion Prices Index) and employment statistics (e.g., ADP employment reports). The emergence of big data, coupled with advances in computational science analytic techniques, raises the possibility of developing indicators of citizens’ civic engagement and other social behaviors and attitudes that are less burdensome than surveys.
The statistical agencies are of course aware of the changing data landscape and are considering measures to adapt and take advantage to modernize their programs. Even so, the magnitude of upcoming changes argues that the statistical agencies be involved even more closely in these developing areas and engage in parallel data studies.