Home > Statistics > Consumer Confidence > 2020 > January > 2. Description of level revision of time series

2. Description of level revision of time series

In May 2019, Statistics Finland published time series for the statistics on consumer confidence that started from January 2018. The time series were based on previous time series of the Consumer Survey, the figures of which were revised to take into account the changes made to the data collection. Due to feedback from data users, we decided to form comparable time series until 1995.

The statistics on consumer confidence published by Statistics Finland in May 2019 differs from the earlier Consumer Survey in terms of both the data collection method and age limits. In place of previously used telephone interviews, the data are collected with a combination of a web questionnaire and a telephone interview, which will be referred to as mixed-mode data collection. A panel design is in use, while earlier the data was collected from one-time samples. In addition, the population of the statistics on consumer confidence has been limited to persons aged between 18 and 74 instead of the previous population aged 15 to 84.

In order to continue comparable time series for the statistics on consumer confidence, the weights used in the sample survey were recalibrated starting from 2000. Unlike in the past, the level of education of the respondents was taken into consideration when creating the weights. The survey results were generalised with the help of the new weights to the entire population aged 18 to 74 for the entire time series. In addition, questions that previously were asked as household-specific in the survey, e.g. concerning plans to purchase a car or dwelling, were changed with a weighting coefficient to describe the purchasing plans of individual persons.

How the data collection method affects the results

The changed data collection method could not be taken into account with the help of weights. The effects of the change in the data collection method on balance figures were estimated by utilising the data collected in mixed-mode data collections in February, March and April of 2019 that included 1,443 observations. The data from the mixed-mode data collection were compared to the data collected in telephone interviews for the Consumer Survey over the same period that included 3,071 observations. The aim was to examine to what extent the data collection method affects the results.

The consumer confidence indicator (CCI/A1) is the main key figure of the statistics on consumer confidence: the average of the balance figure of the four components. The balance figure is the weighted difference between the positive and negative response rates that varies between -100 and +100.

The values of the CCI and its components by data collection method are presented in Figure 1. Based on this, the mixed-mode data collection systematically produces more negative balance figures than mere telephone interviews.

The statistical significance of the differences between the data collection methods can be estimated using the 95 per cent confidence intervals presented in the line segment in Figure 1. Even though the balance figures show that the views concerning the respondent’s own economy (B1, B2) are more negative in the mixed-mode data collection than in mere telephone interviews, the confidence intervals are overlapping and, thus, the differences are likely to not be statistically significant. For other components, the differences in the balance figures are higher than for the views concerning the respondent’s own economy. The biggest difference in the components of the indicator between the data collection methods is in the balance figure for spending on durable goods (E1), where the difference is around seven units.Figure 1. Consumer Confidence Indicator (CCI) and its components by two different data collection methods

Figure 1. Consumer Confidence Indicator (CCI) and its components by two different data collection methods

It is known based on research literature that the data collection method affects the results. The respondents tend to give more pessimistic views on the web form than in telephone interviews. Especially for sensitive topics, the form has been seen as enabling more truthful responses than the telephone interview as the respondent is not under pressure to answer in a socially acceptable way. In this respect, a mixed-mode data collection can be seen as an improvement in terms of the reliability of the results. In addition, several data collection methods serve different types of responders better.

Forming of the time series

In May 2019, in connection with the release of the statistics on consumer confidence, the old time series of the Consumer Survey were revised to take into account the changes caused by the data collection method starting from 2018. Existing time series were revised to take into account the effect of the data collection methods on key figures that are presented in Figure 1 and Appendix table 2.

Revising the time series further back than 2018 required methodological work to calculate the weights and to determine the magnitude of level revisions in the longer term. When the weights had been calculated for the monthly data starting from 2000 using the new method, the magnitude of the level revision could be estimated with the help of re-weighted time series with uniform populations.

It was decided that the level revision would be backcasted in the time series until 1995 in order to avoid a clear break in the time series caused by the changed data collection method. Estimates on the level revision are based on the data from the mixed-mode data collection in early 2019 described above.

The time series for the statistics on consumer confidence are now available starting from 1995. Uniform time series describe the population aged 18 to 74 and they try to take into account the changes caused by the data collection method in recent years. We could only assess the changes caused by the data collection method reliably for the balance figures. Thus, the revised time series only include data on key balance figures, the confidence indicator, inflation percentages and a few relative shares.

Source: Consumer Confidence 2020, January. Statistics Finland

Inquiries: Pertti Kangassalo 029 551 3598, Tuomas Parikka 029 551 3276, consumer.confidence@stat.fi

Director in charge: Jari Tarkoma

---

Updated 27.1.2020