Use of mobile positioning data to calibrate travel behaviour surveys

Objective and implementation

In the frame of quarterly sample surveys, the national travel behaviour of people aged 15 years and over living in Austria is portrayed. Each quarter, about 3 500 selected domestic residents aged 15 and over are surveyed by telephone or - since 2020 - also via a web questionnaire (population 2020 of about 7.5 million people).

The surveys concern holiday and business trips with (overnight trips) and without overnight stays (same-day trips) in Austria and abroad. In addition to the profile of domestic and international trips (length of stay, travel destination, month of departure, purpose of travel, type of accommodation, booking, means of transport and expenses), the profile of travellers and non-travellers (socio-demographic characteristics, reasons for not traveling) is recorded.

With regard to the interpretation of the results, it should be noted that they are based on retrospective surveys over a previous reporting period, which can lead to further bias due to memory problems.

Due to the relatively small sample size (3 500 people per quarter), there are high fluctuations regarding individual destination countries but also other aggregates (e.g. expenditure). Thus, in the past (particularly when comparing 2018 to 2019), implausible changes have occurred, which was possible due to variations caused by random samples.

In order to further improve the quality of the results, the use of additional or new data sources was evaluated („Big Data“- e.g. mobile positioning data) and alternative ways of collecting data have been analysed for some time now.

Since the mobile phone market is almost saturated and there are no roaming fees in EU-countries, which are the most important destination countries for people living in Austria, the information from active Austrian SIM cards abroad is used to compensate these statistical variations and thus stabilize the results.

The objective of the project was to check the plausibility of the results on holiday and business trips from the sample survey  with mobile phone data or to improve their quality. In this regard it was also important to analyse the inaccuracy of the results that arise from the use of the mobile phone data.

For this purpose, the reporting years 2018 and 2019 were used jointly as a basis for aligning the figures from the mobile phone data with the surveys´ figures. A linear model was estimated for this, which had the logarithmized monthly values of the survey as a dependent variable and the logarithmized monthly values of the mobile phone data as an independent variable. This model was partially estimated for selected countries (Germany, Italy, Switzerland incl. Liechtenstein, Slovakia, Slovenia, Czech Republic, Hungary, France, Spain, Croatia, Greece) and jointly estimated for all other countries.

The adjusted values of this model served as the basis for the provision of key figures, which were aggregated per quarter.