|
Abstract:
|
Information from social networks seems an opportunity for production of official statistics. National statistical offices (NSOs) have begun experiencing with it. Therefore, it is necessary to inquire about its representativeness. NSOs collect socio-demographic information through censuses and surveys. This allows for comparisons between them and, when a sample is not representative, correction of results. This is not possible in social networks since user profiles are not public. Consequently, it is difficult to determine whether a sample is representative. Here we make explicit biases attributable to use of social network information in Mexico, from ENDUTIH, 2017-2019. Beyond bias detection, we can now reduce them using techniques developed for random sample surveys but using predicted socio-demographic information of network users instead. To us ENDUTIH is only a first step in using mixed network-survey strategies. If successful, we can consider other topics included in ENDUTIH, as well as from any survey conducted by ONS, to produce representative results from social media. Therefore, any topic is open for study by combination of surveys and social networks.
|
