Chair: Marina Signore
Room: S4A Mariacki
Time: 11:30 - 13:00
Date: 27 June
Session 26 - papers & presentations
|Title: <<< Geospatial mobile data to increase the quality of usual place of residence >>>
The next population and housing census in Estonia at the end of 2020 is intended to be register-based. One of the biggest challenges on the way to register based census is the difference between registered and actual places of residence. By law, everybody is obliged to ensure that their correct usual residential address is entered in the Population Register, but there are different reasons why people don’t do that (for example school places, different benefits etc). The difference between registered and actual places of residence affects the breakdown of the lowest level of the place of usual residence and all household and family characteristics. To solve this problem Statistics Estonia (SE) has started pilot project for testing possibility to use mobile positioning data. The pilot study consists of four steps. Firstly SE asked volunteers to participate in the project and to fill in written consent to use their mobile positioning data. In addition the actual place of residence of volunteers was collected. Secondly the set of potential addresses was created for each participant by linking the addresses of respondent and his/her close relatives from Population Register and Land Register. The anchor points of the first and second place of residence are estimated based on the mobile positioning data as a third step. Finally anchor points and other auxiliary information is used to build a model for selecting the most probable place of residence from the set of addresses. Results of the pilot study will be presented.
|Title: <<< Mixed Mode Effects of Web and Telephone Surveys on Measuring Employment Status >>>
Web questionnaires are increasingly used to complement traditional data collection, leading to different combinations of mixed survey modes. The flexibility of mixed mode surveys provides many advantages such as broader coverage of the population, less nonresponse issues, lowered expenditures, and compensation for the decreasing availability of other data sources, i.e. fixed-line telephone numbers. However, the increased usage of web data raises concerns whether web questionnaires lead to systematic measurement errors, since responses given to web questionnaires may be significantly different compared to other survey modes.
Based on the Luxembourgish Labor Force Survey, the largest recurring household survey in Luxembourg, we investigate if reports on participants’ employment status differ in web and telephone data. Analysis of the raw data reveals significant differences in sample composition (e.g. respondents’ personal characteristics such as age or nationality) as well as in key variables for measuring employment status. People in the web sample report, for instance, more often to be in employment, work more hours on average, and have more often a full-time job.
In order to investigate whether differences in employment status are caused by sample composition or survey mode effects, we match web and telephone samples according to variables that lead to dissimilarities in sample composition. We identify these variables by a combination of automatic variable selection via random forest and a theory driven selection. Based on the selected variables, we then apply a Coarsened Exact Matching to enable a comparison among web and telephone samples.
After matching, we show that most differences between web and telephone samples disappear, which leads to the conclusion that survey mode has a minor effect on measuring employment status. The results of the present study therefore support the implementation of mixed survey modes in labour market related official statistics such as the Labour Force Survey.
|Title: <<< Innovation in questionnaire design: the electronic questionnaire ‘authoring’ tool >>>
The Office for National Statistics is committed to moving all business surveys online in the next few years, as well as making the Census online mode the default method of data collection. So far ONS has moved five business surveys online with a high degree of success (which has been evidenced by quicker returns and fewer errors when compared to the paper versions); however the process for moving these surveys has been slow. Previously, designing the online version of paper questionnaires was very time consuming; methodologists would create the initial questionnaire templates using Google Slides while seeking feedback from subject matter experts. The agreed content would then be passed to user experience designers and developers who were responsible for highlighting problems with the design of certain questions. For example, designs might not be suitable for users who need to use assistive technology, such as screenreaders or zoom text, to help them complete the questionnaire. Subsequent redesigns were sometimes necessary. The questionnaire would then undergo pre-testing before being dispatched to respondents. In order to speed up the design and testing process an ‘authoring’ tool has been developed, this enables methodologists to quickly design iterations of questionnaires in the same format and style as they appear in the final version. The tool has in-built question types and answer options, such as checkboxes, radio buttons and currency values. It allows the user to add routing and validation to the questionnaire as they are building it which significantly speeds up the design process. The tool is also being developed in conjunction with a question bank, providing a unique opportunity to harmonise business survey questions, ensuring quality and consistency across business surveys. This paper will address the benefits and challenges of using the tool and discuss the future of questionnaire design in ONS.
|Title: <<< Transition to WEB data collection in household surveys at SURS – what we have learned so far >>>
At the Statistical Office of the Republic of Slovenia (hereinafter SURS) 20 surveys of households or individuals are conducted on a regular basis. More than 100,000 persons or households are surveyed every year (5% of the population of persons and more than 10% of the population of households). The predominant data collection mode is still telephone. Due to huge technological development of the ICT in the last 10 years and increased demand for internet-based services, and also pressure to reduce data collection costs, SURS launched a 13-month project in 2014 to set up processes and standards for WEB data collection for surveys of households and individuals. Since 2016 at SURS several household surveys have been transferred to the WEB survey mode: Consumer Survey, Mobility Survey, ICT Usage in Households and by Individuals Survey, and Household Energy Consumption Survey. The article aims to present some of the methodological challenges when transferring data collection to a new survey mode: questionnaire testing for WEB surveys, re-interviewing to possibly assess the measurement error, communication with the respondents who are the non-respondents in the presented WEB surveys, analysis of the costs, and impact of the WEB survey data collection on some key statistics.