What is the minimum participation percentage to achieve statistically valid results?

There is not a simple answer to this question. The proper answer goes beyond just the statistical aspect of things – i.e., it’s not just a statistical calculation using a mathematical formula. In fact, all the usual rules about sample size requirements don't really apply when it comes to an employee survey for an organization that wants to produce reports for all its various departments and work groups.

Statistical sampling tables indicate that the required sample size to produce a statistically valid response at the 95 percent confidence level with a +/-5 percent margin of error (which is considered the universal default standard) asymptotes at about 385. This means that the required sample size reaches an upper limit of only 385 regardless of how large the population is, even a million or more. This is why national public opinion polls often are based on a sample of only 350-400 people to make a statistically valid statement about an entire country! So, for a very large organization, the minimum participation percentage could be extremely small.

But the required sample size for an organization as a whole ignores what would be required to produce a statistically valid result for a specific division, department or work group. For example, a sampling table will indicate that the required sample size for a group of 200 is 132 (66%), for a group of 100 is 80 (80%) and for a group of 10 is 10 (100%). In other words, the smaller the group, the proportionately larger percentage of that group is required. To ensure a “statistically valid” sample for every part of the organization, no matter how small, one would need to achieve 100% participation, which is a virtual impossibility. From a more practical perspective, the response rate should be as high as possible to maximize the representativeness of the data for the whole of the organization and all its piece parts, as well as to minimize questions one might otherwise get about the sentiment of those who didn't respond to the survey. Achieving a 70 percent participation target for the organization as a whole puts most minds at ease regarding concerns about data representativeness and having heard from the lion’s share of employees, even though it’s massive overkill for most organizations, at least from a statistical sampling perspective. But a 70 percent response rate will come up well short of the sample required for many groups within an organization that are relatively small. And that’s simply the practical reality of employee surveys.

In sum, every client should do their best to achieve the highest possible participation to maximize data integrity (realizing perfection is not possible) and to minimize concerns about non-respondents. But at the end of the day for an employee survey, one simply takes what one can get. The 70 percent target participation for most clients is an ambitious but achievable target. Some organizations achieve 80 to 90 percent or better, although the larger the organization, the harder it is to get a high participation rate since there are more people to corral. And for very large organizations, most are quite happy with achieving 60 percent participation overall.