2019 Volume 3 Issue 1 Pages 107-128
In the case of company surveys, companies that have failed on the survey theme are reluctant to respond, resulting in sampling bias. We tried to estimate this bias in the case of data utilization by asking the same question to the employee and comparing it with the result of the company survey. Since employees are asked anonymously, they will answer without concern even if the company at work fails on the theme. Thus, the difference between the results of the two surveys can be interpreted as the result of sampling bias. We found that the success ratio rate of data utilization was about 7.1 percentage points higher in the company survey than in the employee survey. If this difference is caused by sampling bias, successful companies answered about 30% more than those that did not. Converting to a correction weight, when performing regression analysis etc., a successful company should be multiplied by 1/1.3 or about 0.77. With this degree of correction, it is unlikely that the results of regression analysis will be significantly altered. However, this survey is a business survey conducted by the government in which the bias is relatively small. It is expected that bias will be greater in the more sensitive thematic surveys conducted by the private sector.
