Online Program

Return to main conference page

All Times ET

Program is Subject to Change

Tuesday, June 15
Tue, Jun 15, 11:30 AM - 1:00 PM
TBD
Topics in the Collection, Production, and Estimation of Short Term and other Business Statistics

What Do the Business Tendency Surveys Measure, and What Is Their Information Content? (308072)

Veronika Ptácková, University of Economics 
*Paul Anthony Smith, University of Southampton 

Keywords: Business Tendency Survey, rotation group bias, panel conditioning

Business tendency surveys (BTS) collect qualitative information which can be provided with relatively little respondent effort. At least part of the information is prospective, providing an assessment of what will happen in the near future. Liu et al. (2011) used a matched dataset with BTS data linked to quantitative data from official surveys in the UK, and found little consistency in the information from the two sources. The BTS aims to measure something in the future, but it is possible that a) respondents are conditioned by what is happening at the time of data collection, so the current situation influences their responses; b) the time for data to be collected, processed and published causes a lag in published information so that the survey actually measures something in the recent past.

In this paper, we will use Czech BTS data in four sectors: industry, construction, some selected services, and trade. We also use the results from a voluntary questionnaire asking about respondents’ approach to answering questions in the BTS. We investigate two questions: Firstly, we examine the information content of the BTS. From the voluntary survey, we will know how the respondent understands the questions in the survey. In Czechia, respondents are legally required to fill in the survey. Secondly, we will use these data sources to assess the presence of rotation group bias (RGB) and panel conditioning in the BTS using suitable models accounting for the complex data structure. Panel conditioning might be influential in the BTS where responses are opinion-based rather than numeric. If RGB is detected we will also investigate the use of models to adjust the outputs from the BTS and improve its quality.