logo
25logo hrv-eng


Variability

This presentation of results is intended for the general public and therefore is not overloaded with technical details, including variability. However, it should be noted that results obtained in samples are not exact indicators but are estimates of the range of values within populations. If the research methodology is implemented correctly on an unbiased sample, then the accuracy of the estimates of population values depends on the variability of the phenomenon that is being examined and the sample size.

To illustrate, if the result (expressed as a percentage) obtained in a simple random sample of 1000, at proportion of 50% variability is highest and population value (in 95% cases) is in interval 50% +/- 3.1%. Confidence interval symmetrically reduces as proportion increases or decreases, and by result of, e.g., 10% (or 90%) is 10 (90)% +/- 1.9%).

Since possible sources of bias (refusals, non-responding, the social desirability of responding) should not be overlooked in the interpretation of survey results, small differences should not considered, regardless of whether these are percentages or averages. Although small differences in large samples can reach the level of statistical significance (especially if they are numerous), they are more theoretically than practically relevant (and in the scientific literature, along with statistical significance, effect size is often considered). Therefore, attention should be paid to the prominent differences, groupings, clear and consistent trends, as well as the review of specific results and interpretation of comparisons.

 

Comparisons

As a result of the mentioned reasons and because the compared results are not part of the same research (although they were all conducted on representative samples), comparisons are illustrative and are shown graphically, without specifying the exact figures.

Some of the results from this study were compared with the findings of previous surveys conducted by the Institute (surveys from 2005, Pilar barometer of 2008 and 2009). All the studies were conducted on independent, equivalently selected representative samples of the adult population.

In addition, some findings are compared with the results of international research, the European Social Survey (ESS) and the World Values Survey (WVS). Details of the participating countries, samples, content and conduct of the study can be found on their websites (ESS, WVS), that allow online and independent analysis.

For comparisons on this website, results from EU member states that participated in the ESS twice were selected; in 2010 (22 countries, including Croatia, conducted by the Pilar Institute) and in 2012 (21 countries, with results from Pilar’s 2014 barometer. In 2012, the Institute could not participate due to financial reasons). In the last round (2010–2014), nine EU member states (excluding Croatia) participated in the WVS and their results have also been compared with the results of Pilar’s 2014 barometer.

Note: In graphs and tables commas are decimal delimiters.