What Is a Stratified Random Sample?

Definition & Examples of Stratified Random Samples

Stratified random sampling produces a Representative Focus Group

Nicole S. Young / E+ / Getty Images

A stratified random sample is a sample consisting of distinct but homogenous subgroups known as strata.

Learn how a stratified random sample is used in market research, the types of samples you can derive, and how it compares to a simple random sample.

What Is a Stratified Random Sample? 

A stratified random sample is a sample obtained by dividing a larger, typically heterogeneous population into distinct but homogenous subgroups known as strata and then selecting sampling units from each stratum for inclusion in the sample.

A stratified random sample is considered probabilistic because every method used to select the sample population provides a reasonably reliable way of estimating how representative the sample population is to the larger population from which the sample was selected. In other words, a probabilistic sample permits a researcher to estimate the odds that the sample selected does or does not represent the larger population from which it was drawn.

How a Stratified Random Sample Works

A sample is a mini-representation of a larger population. Samples can be determined informally or formally. But samples that are systematically developed according to certain scientific methods, such as stratified random samples, are generally perceived as being more useful for making generalizations about the larger population. Businesses can use such generalizations in market research to gauge consumer needs and wants and develop an appropriate marketing strategy.

The aim of stratified random sampling is to select participants from various strata within a larger population when the differences between those groups are believed to have relevance to the market research that will be conducted. For instance, the results of a market research survey could be influenced by subject attributes such as age, gender, work experience level, racial or ethnic group, economic situation, or education level. In a stratified random sample, these potentially influential characteristics can be reasonably assumed to reflect the pattern of the characteristics in the overall population.

For example, let's say that you run a fintech firm and are interested in developing a new retirement savings app to help customers save more for retirement. You want to conduct a survey and learn how much Americans aged 23 and over contributed to their retirement accounts last year so that you can better tailor the app to customer investing habits. But that population amounts to millions who would be too difficult to survey individually. Plus, consumers of different age groups are likely to have different investing habits that affected their retirement savings contribution last year.

So, you decide to gather a stratified random sample with a sample size of 5,000 and the following five age strata: Silent Generation, Baby Boomers, Generation X, Millennials, and Generation Z.

When you research the demographics for the overall U.S. population last year, you learn that 6% belonged to the Silent Generation, 25% were Baby Boomers, 24% were Generation X, 23% were Millennials, and 22% were Generation Z. If you want to achieve a proportional stratified random sample where each stratum is proportional to its percentage of the overall population, you would gather the stratified random sample of 5,000 as follows: 300 from the Silent Generation (6% of 5,000), 1,250 Baby Boomers (25% of 5,000), 1,200 from Generation X (24% of 5,000), 1,150 Millennials (23% of 5,000), and 1,100 were from Generation Z (22% of 5,000).

You would then administer the survey to a sample comprising those sampling units to make more meaningful inferences about how much each age stratum contributed toward retirement last year, which are insights you can incorporate into your investing app.

Stratified random sampling is typically used when there is interest in the differences between subgroups and the larger population.

Types of Stratified Random Samples

Once you have divided an overall population into strata, there are two main ways to select units from each stratum for inclusion in the sample:

  • Proportionate stratified random sample: In this type of stratified random sample, each stratum represents the same percentage of the sample as it does of the larger population. In the earlier example, the sample that weighted each age stratum in the sample according to its percentage in the U.S. population was a proportionate stratified random sample. This type of sample requires access to data on the larger population.
  • Disproportionate stratified random sample: The percentage of each stratum in the larger population is not taken into account in this type of sample. For example, if you identified the five age strata and selected 1,000 people from each stratum for the survey, you'd have a disproportionate stratified random sample. This type of sample is useful when the research objective is broad or you don't have access to data on the larger population.

Stratified Random Sample vs. Simple Random Sample

Unlike a stratified random sample that contains sampling units from each distinct stratum that have a known, non-zero chance of being selected, a simple random sample is one without subgroups. Instead, every unit of the sample has an equal chance of being included in the sample. Going back to the earlier example, let's say that you randomly administer the survey to 5,000 people without regard for their differentiating characteristics. Among this simple random sample, certain age strata may be disproportionately represented, and others not represented at all.

As such, stratification is likely to produce more representative outcomes that can be used to make more accurate inferences about the larger population. Simple random sampling doesn't necessarily represent the larger population.

And, as a stratified random sample contains subgroups that can be studied together, this type of sampling can also reduce the cost per observation in a firm's market research study.

However, both types of samples have their place in market research. Whereas a stratified random sample is more complex and is used when differences between sampling units of the larger population are myriad and are significant to the results of the study, a simple random sample is more straightforward and often used when there are few differences to account for in the larger population (or in some cases so many differences that stratification isn't feasible), or those differences aren't relevant to the study. For example, if you're conducting a broad survey of new menu options for a pizza joint, you might consider a simple random sample, as characteristics like age, gender, and race may not affect the results.

Stratified Random Sample Simple Random Sample
Divides a larger, heterogeneous population into strata Undivided, typically homogeneous sample
Accurately represents the larger population Not necessarily representative of the larger population
More complex to carry out More straightforward
Lower cost per observation Potentially higher cost per observation

Key Takeaways

  • A stratified random sample contains distinct, homogenous subgroups and can be used to make inferences about a larger population for market research.
  • It's achieved by dividing a large, heterogeneous population into subgroups called strata and then selecting units from each stratum for inclusion in the sample.
  • Proportionate stratified random samples contain strata in proportion to their percentage in the larger population, while disproportionate samples don't take the share of each stratum in the overall population into account.
  • A stratified random sample results in more accurate and potentially less costly observations than simple random samples.