# Stratified Sampling

Stratified sampling is a probability sampling method and a form of random sampling in which the population is divided into two or more groups (strata) according to one or more common attributes.

Stratified random sampling intends to guarantee that the sample represents specific sub-groups or strata. Accordingly, application of stratified sampling method involves dividing population into different subgroups (strata) and selecting subjects from each strata in a proportionate manner. The table below illustrates simplistic example where sample group of 10 respondents are selected by dividing population into male and female strata in order to achieve equal representation of both genders in the sample group.

Stratified sampling can be divided into the following two groups: proportionate and disproportionate. Application of proportionate stratified random sampling technique involves determining sample size in each stratum in a proportionate manner to the entire population.

In disproportionate stratified random sampling, on the contrary, numbers of subjects recruited from each stratum does not have to be proportionate to the total size of the population. Accordingly, application of proportionate stratified random sampling generates more accurate primary data compared to disproportionate sampling.

## Application of Stratified Sampling: an Example

Suppose, you dissertation aims to explore the leadership styles exercised by medium-level managers at Bayerische Motoren Werke Aktiengesellschaft (BMW AG). You have selected semi-structured in-depth interviews with managers as the most appropriate primary data collection method to achieve the research objectives.

Application of stratified random sampling contains the following three stages.

1. Identification of relevant stratums and ensuring their actual representation in the population. Apart from gender as illustrated in example above, range of criteria that can be used to divide population into different strata include age, the level of education, status, nationality, religion and others. Specific patterns of categorization into different stratums depends aims and objectives of the study.

In our case, BMW Group employees are employed across four business segments – automotive, motorcycles, financial services and other entities[1]. Accordingly, each segment can be adapted as stratum to draw sample group members.

2. Numbering each subject within each stratum with a unique identification number.

3. Selection of sufficient numbers of subjects from each stratum. It is critically important for samples from each stratum to be selected in a random manner so that the relevance of bias can be minimized.

As it is illustrated in the table below, following the procedure described above results in the sample group of 16 respondents, BMW Group medium level managers that proportionately represent all four business segments of the company.

 Automotive Motorcycles Financial services Other entities N Manager ü N Manager ü N Manager ü N Manager ü 001 Hudson 001 Conrad ü 001 Guzman 001 Sparks 002 Bass ü 002 Braun 002 Craig 002 Atkinson ü 003 Richmond 003 Gentry 003 Green ü 003 Montes 004 Tucker 004 Hartman ü 004 Ballard ü 004 Mcguire 005 Chavez ü 005 Levine 005 Cox 005 Spencer ü 006 Riddle 006 Griffin ü 006 Dunlap ü 006 Davies 007 Mckinney 007 Valentine 007 Patrick 007 Bradford ü 008 Terrell ü 008 Mcdonald 008 Gardner ü 008 Collins 009 Hayes 009 Brown ü 009 Carpenter 009 Chen 010 Escobar ü 010 Kaufman 010 Vasquez 010 Hess ü

1. Stratified random sampling is superior to simple random sampling because the process of stratifying reduces sampling error and ensures a greater level of representation.
1. Thanks to the choice of stratified random sampling adequate representation of all subgroups can be ensured.
1. When there is homogeneity within strata and heterogeneity between strata, the estimates can be as precise (or even more precise) as with the use of simple random sampling.