Data Bias and its Impact on Decision Making

Good decision making involves evaluating information quickly and being able to draw accurate conclusions.

However, data which is biased - not representative of the population - or which is taken out of context is likely lead to poor decision making.

What are some of the common types of data bias? And how can data bias be avoided?

Make strong decisions based on accurate, unbiased information

Types of Data Bias

Selection Bias

This occurs when data is not selected randomly and results in a sample which is not representative of the whole population.

For example, a retail company analysing sales trends in December when it has a strong seasonal trade during Christmas.

Self-Selection Bias

This occurs when individuals select themselves to be part of a sample.

For example, questionnaires which offer individuals the chance to be entered into a prize draw if they participate in a survey. Those who are not interested in the prize draw will likely not complete the survey.

Observer Bias

This occurs when an individual's expectations, opinions or prejudices influence the data which is recorded.

For example, a restaurant reviewer expecting locations to be dirty may notice more rubbish.

Cognitive Bias

This occurs when personal experience and preference affects how data is presented and distorts the context.

For example, a management report detailing that sales have increased 20% compared to the previous year signals strong performance. However, that view is adjusted if the context is that market growth exceeded 30%.

Confirmation Bias

This occurs when individuals see data which confirms their beliefs and ignore (consciously or sub-consciously) data which does not agree with their beliefs.

For example, hiring a candidate because they went to a school with a good reputation and overlooking other more important factors, such as their suitability to the role.

Survivorship Bias

This occurs where a sample only contains items which have survived another previous event.

For example, a school only lets students sit a final exam if they pass a previous mock test and then claims that 95% of students pass the final exam first time. This claim ignores the students not allowed to sit the exam.

Keys to Avoiding Data Bias

• Remain sceptical of any data used

• Ascertain how the data has been collected and its source

• Determine if there are any data outliers which have skewed the results

• Vary sampling techniques

• Use more than one indicator to make a fully informed decision

• Try to locate evidence that disproves initial presumptions

• Discuss the results with someone else - especially if they favour a different conclusion

Points to note:

More data doesn't automatically mean better data - even Artificial Intelligence (AI) models using enormous data sets have been recognised to be susceptible to data bias, with a recent hiring algorithm favouring people from one ethnicity over another.

This document is a simplified helpsheet and careful research should be completed if you are unsure.

Need more information? Contact us today to find out more.

Verifiable Accounts - Professional Financial Accountants providing Tax Preparation and Accounting Services