how to evaluate the representation of market data in compensation surveys

Compensation survey market data is not a representative source of information in the sense of a representative sampling assuring that inferences and conclusions can safely extend from the sample to the population as a whole.

Salary reports should therefore always be used with care!

We recommend to use salary comparisons in combination with other sources, e.g. multiple compensation surveys, data from labour agreements, an internal analysis of employees' and new hires' pay based on an analytical job evaluation, etc.


The population of a survey (i.e. the participating companies and employees) changes with every issue. Hence vendors are faced with the challenge to maintain a certain level of consistency of their benchmark salaries between the years. To do this, the providers use different approaches to analysing and preparing their salary reports.

  • Market values from statistical models
    This approach uses statistical data calculated through regression analysis and / or significance testing etc. to show a possible market salary and is used in some global surveys.
  • Market values from experience
    This approach often combines data and experience from candidate interviews to show market salary ranges and is often used by local and global recruiting agencies alike.
  • Market values from descriptive statistics
    This approach represents salary information through percentiles (quartiles and median values, etc.).
    • Based on pure data input
      In this case data is collected and described as is, there is no editing of the population or removing of data points. Vendors that follow this approach face volatile results in an annual comparison since the population rarely ever remains stable. This problem becomes bigger the smaller the population is.
    • Based on edited data input
      Automated computer programmes or human analysts edit the data by for example removing (low or high earning) "outliers" in order to keep median values relatively stable. A deviation of median values by +/- 5% might be allowed to show between the years.

Would you like to know more?

Register now for our free basic version
and evaluate the enterprise edition for free!

evaluate now!