Bart Buelens

Disentangling mode-specific selection and measurement bias in social surveys

A large-scale mixed-mode experiment linked to the Dutch Crime Victimization Survey was conducted in 2011. The experiment consisted of two waves; one wave with random assignment to one of the modes web, paper, telephone and face-to-face, and one follow-up wave to the full sample with interviewer modes only. The objective of the experiment is to estimate total mode effects and more specifically the corresponding mode effect components arising from undercoverage, nonresponse and measurement. In this paper, mode-specific selection and measurement bias are defined, and estimators for the bias terms based on the experimental design are introduced and discussed. The proposed estimators are applied to a number of key survey variables from the Labour Force Survey and the Crime Victimization Survey.

Big Data as a Source for Official Statistics

Abstract More and more data are being produced by an increasing number of electronic devices physically surrounding us and on the internet. The large amount of data and the high frequency at which they are produced have resulted in the introduction of the term ‘Big Data’. Because these data reflect many different aspects of our daily lives and because of their abundance and availability, Big Data sources are very interesting from an official statistics point of view. This article discusses the exploration of both opportunities and challenges for official statistics associated with the application of Big Data. Experiences gained with analyses of large amounts of Dutch traffic loop detection records and Dutch social media messages are described to illustrate the topics characteristic of the statistical analysis and use of Big Data.

Measurement Error Calibration in Mixed-Mode Sample Surveys.

Mixed-mode surveys are known to be susceptible to mode-dependent selection and measurement effects, collectively referred to as mode effects. The use of different data collection modes within the same survey may reduce selectivity of the overall response but is characterized by measurement errors differing across modes. Inference in sample surveys generally proceeds by correcting for selectivity—for example, by applying calibration estimators—and ignoring measurement error. When a survey is conducted repeatedly, such inferences are valid only if the measurement error remains constant between surveys. In sequential mixed-mode surveys, it is likely that the mode composition of the overall response differs between subsequent editions of the survey, leading to variations in the total measurement error and invalidating classical inferences. An approach to inference in these circumstances, which is based on calibrating the mode composition of the respondents toward fixed levels, is proposed. Assumptions and risks are discussed and explored in a simulation and applied to the Dutch crime victimization survey.

Comparing Two Inferential Approaches to Handling Measurement Error in Mixed-Mode Surveys

Abstract Nowadays sample survey data collection strategies combine web, telephone, face-to-face, or other modes of interviewing in a sequential fashion. Measurement bias of survey estimates of means and totals are composed of different mode-dependent measurement errors as each data collection mode has its own associated measurement error. This article contains an appraisal of two recently proposed methods of inference in this setting. The first is a calibration adjustment to the survey weights so as to balance the survey response to a prespecified distribution of the respondents over the modes. The second is a prediction method that seeks to correct measurements towards a benchmark mode. The two methods are motivated differently but at the same time coincide in some circumstances and agree in terms of required assumptions. The methods are applied to the Labour Force Survey in the Netherlands and are found to provide almost identical estimates of the number of unemployed. Each method has its own specific merits. Both can be applied easily in practice as they do not require additional data collection beyond the regular sequential mixed-mode survey, an attractive element for national statistical institutes and other survey organisations.

Data Mining for Official Statistics: Challenges and Opportunities

We present our vision on the use of data mining for official statistics, illustrate this with some examples, sketch a general framework, and provide directions for future research.