In this work, we reflect on some questions about the measurement problem in economics and, especially, their relationship with the scientific method. Statistical sources frequently used by economists contain qualitative information obtained from verbal expressions of individuals by means of surveys, and we discuss the reasons why it would be more adequately analyzed with soft methods than with traditional ones. Some comments on the most commonly applied techniques in the analysis of these types of data with verbal answers are followed by our proposal to compute with words. In our view, an alternative use of the well known Income Evaluation Question seems especially suggestive for a soft computing approach, since it would facilitate an empirical estimation of the corresponding linguistic variable adjectives. A new treatment of the information contained in such surveys would avoid some questions incorporated in the so called Leyden approach that do not fit to the actual world.