Page 170 - Contributed Paper Session (CPS) - Volume 3

CPS1983 Chong N. et al.

Data analytics for better statistics
Chong Ning, Benjamin Ng, Jeremy Heng
Ministry of Manpower, Singapore

Abstract
Modern technology has revolutionized the way official statistics is
produced and consumed. With large amounts of data available, Singapore’s
Ministry of Manpower seeks to tap on data analytics to improve the quality of
statistics produced and formulate better policies. In the Manpower Research
and Statistics Department, data analytics is used in three areas to aid in the
statistical production process: (1) automated-classification system, (2)
sentiment analysis and (3) predictive modelling.
An automated-coding system is developed to automate the conversion of
raw occupation and industry data into standard occupation and industry
codes. It ensures consistency among interviewers and respondents who may
have their own understanding of occupation and industry definitions. A
speech-to-text analytics tool facilitates sentiment analysis to provide insights
into the behaviour of interviewers and respondents. By analysing telephone
conversations, it is able to flag out uncooperative respondents and under-
performing interviewers for follow-up action. Lastly, a fieldwork predictive
model is used to predict the optimal dates and times to conduct survey
interviews with various demographics of respondents, thereby reducing the
likelihood of non-response and refusal cases.
Through these initiatives, the Ministry is able to improve operational
efficiency and data quality. The paper discusses the challenges facing official
statistics, the idea behind the data analytics initiatives, how they are able to
tackle the challenges, and ultimately take official statistics into the future.

Keywords
Efficiency; Quality; Prediction; Optimization; Classification

1. Introduction
Singapore’s official labour statistics is produced and compiled by the
Manpower Research and Statistics Department (MRSD) of the Ministry of
Manpower. The department conducts regular national surveys to collect a
wide range of labour-related data from households, individuals and
businesses. The survey data is cleansed and processed into usable information
which is then analysed to provide labour market insights. The statistics and
accompanying publications are eventually disseminated to the public and

159 | I S I W S C 2 0 1 9