Updated for the 2026-2027 CFA® Level I curriculum.
Sampling begins with a practical question: how will you select the observations used to represent a larger population?
The method you choose affects which observations enter the sample, how precisely you can estimate population characteristics, and whether selection bias may influence the result. For CFA Level I, you should be able to compare five sampling methods and identify the most appropriate one for a given investment problem.
Quick Answer
Probability sampling methods use a random selection process and give population members a known chance of entering the sample. These methods include simple random, stratified random, and cluster sampling. Non-probability methods, including convenience and judgmental sampling, rely on access or analyst choice. They are often easier to use, but their selection bias can be difficult to measure.
Key Takeaways About Probability and Non-Probability Sampling
A population is the complete group an analyst wants to study.
A sample is the smaller group used to estimate characteristics of that population.
Probability sampling uses a random selection mechanism.
Simple random sampling selects directly from the full sampling frame.
Stratified random sampling draws observations from every defined stratum.
Cluster sampling selects only certain groups and then studies observations within them.
Convenience sampling selects observations that are easiest to access.
Judgmental sampling relies on the analyst’s view of which observations are most useful.
A larger sample can reduce random sampling error, but it cannot automatically remove selection bias.
What You Need to Know for CFA Level I
For CFA Level I, focus on:
Classifying a method as probability or non-probability sampling.
Comparing simple random, stratified random, and cluster sampling.
Distinguishing convenience sampling from judgmental sampling.
Identifying the method described in an investment scenario.
Explaining how the sampling design affects sampling error.
Recognizing weak population coverage and possible selection bias.
Understanding when population weights may be needed.
Population, Sample, and Sampling Frame
Before comparing the methods, keep three terms separate:
Population: The full group the analyst wants to understand.
Sample: The subset of the population included in the analysis.
Sampling frame: The list or source used to identify units that can be selected.
Suppose an analyst wants to estimate the average credit spread of all publicly traded corporate bonds in a market.
The full set of eligible bonds is the population. The bonds actually selected are the sample. The database used to identify available bonds is the sampling frame.
A weak or incomplete sampling frame can exclude part of the population before the selection process even begins. Random sampling cannot represent bonds that were missing from the frame.
Probability vs Non-Probability Sampling
Area | Probability Sampling | Non-Probability Sampling |
|---|---|---|
Selection process | Uses a random mechanism | Based on access or analyst choice |
Selection probability | Known and greater than zero | Unknown for at least some population members |
Statistical inference | More defensible when correctly designed | More limited |
Sampling error | Can often be estimated | Usually difficult to estimate reliably |
Main concern | Random sampling error or poor execution | Selection bias and weak population coverage |
CFA methods | Simple random, stratified random, cluster | Convenience, judgmental |
Probability sampling provides a structured way to select observations and assess uncertainty. Its quality still depends on a suitable sampling frame and correct execution.
Non-probability sampling may be useful when time, access, or cost is limited. The analyst should interpret the findings carefully because the sample may systematically differ from the target population.
What Is Simple Random Sampling?
Simple random sampling gives every population member an equal chance of selection. Every possible sample of the chosen size also has an equal chance of being drawn.
Common selection tools include:
Random number generators.
Randomly selected database identifiers.
Random draws from a complete population list.
Simple Random Sampling Example
An analyst has a complete database of 2,000 corporate bonds and uses a random number generator to select 100 bond identifiers.
The random process limits deliberate selection and supports standard statistical inference.
Strengths
The method is straightforward to understand.
Each population member has the same chance of selection.
It reduces the analyst’s ability to favor particular observations.
Standard sampling-error calculations can usually be applied.
Limitations
The method requires a complete sampling frame.
Small but important subgroups may be underrepresented by chance.
Data collection may be expensive when selected observations are widely dispersed.
What Is Stratified Random Sampling?
Stratified random sampling divides the population into non-overlapping groups called strata. The analyst then draws a random sample from every stratum.
Strata are usually based on characteristics that matter to the analysis, such as:
Market capitalization.
Credit rating.
Industry.
Geographic region.
Investment style.
Stratified Random Sampling Example
An analyst divides a market into large-cap, mid-cap, and small-cap companies. The analyst then randomly selects companies from each group.
This process ensures that all three capitalization groups appear in the sample.
Strengths
Every important subgroup receives representation.
Sampling error may decline when observations within each stratum are relatively similar.
The analyst can produce separate estimates for each subgroup.
The method can improve precision compared with a simple random sample of the same size.
Limitations
The analyst needs accurate information to assign observations to strata.
Poorly designed strata can add work without improving the estimate.
Sampling and weighting become more complex when groups are sampled in different proportions.
If small-cap firms make up 10% of the population but 30% of the sample, the analyst may need to apply population weights when estimating an overall market average.
What Is Cluster Sampling?
Cluster sampling divides the population into naturally occurring groups called clusters. The analyst randomly selects some clusters and studies all observations, or a further sample of observations, within the chosen clusters.
Common clusters include:
Brokerage branches.
Geographic regions.
Stock exchanges.
Employer retirement plans.
Administrative districts.
Cluster Sampling Example
A researcher wants to survey retail investors across a country. The researcher randomly chooses 10 brokerage branches and then surveys clients at those branches.
Only the selected branches enter the study.
Strengths
Cluster sampling can reduce travel and data-collection costs.
It is useful when a list of clusters is available but a complete list of individuals is not.
Observations can be collected from a smaller number of locations.
Limitations
Sampling error can be high when observations within each cluster are similar.
The selected clusters may differ from the clusters that were not selected.
Statistical analysis may need to account for dependence among observations in the same cluster.
Stratified Random Sampling vs Cluster Sampling
Both methods divide a population into groups, but they use those groups differently.
Feature | Stratified Random Sampling | Cluster Sampling |
|---|---|---|
Type of group | Stratum | Cluster |
Groups included | Every stratum | Only randomly selected clusters |
Selection within groups | Random observations from each stratum | All or some observations from chosen clusters |
Main purpose | Improve precision and subgroup representation | Reduce collection cost and improve practicality |
Preferred group structure | Similar observations within each stratum | Each cluster broadly resembles the overall population |
Main risk | Incorrect strata or weighting | High similarity within clusters |
A useful exam cue is the number of groups included:
Every group represented: usually stratified sampling.
Only some groups selected: usually cluster sampling.
Simple Random Sampling vs Stratified Sampling
Simple random sampling selects observations directly from the complete population. Stratified sampling first divides the population into relevant groups and then samples from each group.
A simple random sample works well when the population is reasonably uniform and a complete sampling frame is available.
Stratified sampling is often more appropriate when the population contains distinct subgroups that should all appear in the sample. It can also improve precision when the strata explain meaningful differences among observations.
What Is Convenience Sampling?
Convenience sampling selects observations that are easy to reach or readily available.
Examples include:
Surveying investors who visit one financial website.
Using companies with easily accessible financial data.
Interviewing nearby colleagues.
Analyzing securities already stored in an internal database.
Main Advantage
Convenience sampling is fast and inexpensive.
Main Limitation
The available observations may differ systematically from the target population. Because selection probabilities are unknown, the analyst cannot reliably measure sampling error or claim that the sample represents the full population.
A very large convenience sample can still be biased when the selection process repeatedly excludes the same types of observations.
What Is Judgmental Sampling?
Judgmental sampling allows the analyst to select observations believed to be representative, typical, or especially informative.
Examples include:
Choosing securities considered typical of an industry.
Interviewing experienced portfolio managers.
Selecting firms believed to have comparable business models.
Studying countries viewed as representative of a region.
Analyst judgment can improve relevance in exploratory work or specialist research. The conclusion remains dependent on the assumptions used to select the observations.
Conscious preferences and unconscious biases may influence which observations enter the sample.
How Can You Identify the Sampling Method in a Question?
Look for the action used to select the observations.
Question Wording | Likely Sampling Method |
|---|---|
“Randomly selects from the complete population list” | Simple random sampling |
“Divides the population into categories and samples from each” | Stratified random sampling |
“Randomly selects several branches or regions” | Cluster sampling |
“Uses the observations that are easiest to obtain” | Convenience sampling |
“Selects observations based on professional opinion” | Judgmental sampling |
Pay close attention to whether every group is represented. This detail often separates stratified sampling from cluster sampling.
Worked Investment Example
An analyst wants to estimate average financial leverage across a market containing:
60% large-cap companies.
30% mid-cap companies.
10% small-cap companies.
The analyst is concerned that a simple random sample may include too few small-cap firms.
A stratified random sample is suitable because the analyst can:
Divide the market into large-cap, mid-cap, and small-cap strata.
Randomly select companies from every group.
Calculate a leverage estimate for each stratum.
Weight the results according to the population proportions.
This design guarantees representation from the small-cap group while keeping random selection within each stratum.
How Does the Sampling Method Affect Sampling Error?
Sampling error is the difference between a sample statistic and the population parameter that occurs because the analyst observes only part of the population.
A well-designed stratified sample may reduce sampling error when the strata capture meaningful differences across the population.
Cluster sampling may produce higher sampling error when observations within a selected cluster are highly similar. Adding more observations from the same cluster may provide less new information than sampling across several independent groups.
Convenience and judgmental sampling raise a broader concern. Their selection process can create systematic bias, which may persist even as the sample becomes larger.
Sample size improves precision most effectively when the selection process already provides reasonable population coverage.
Common Exam Traps
Common mistakes include:
Classifying convenience or judgmental sampling as probability sampling.
Assuming a probability sample is automatically representative.
Saying stratified sampling selects only some strata.
Saying cluster sampling draws observations from every cluster.
Confusing a stratum with a cluster.
Assuming a large convenience sample must be unbiased.
Ignoring population weights in a disproportionate stratified sample.
Confusing random sampling with random assignment in an experiment.
Assuming that a larger sample corrects an incomplete sampling frame.
Practice Question
An analyst wants to estimate average bond spreads while ensuring that investment-grade and high-yield bonds are both represented. The analyst divides the population by credit category and randomly selects bonds from each category.
Which sampling method is being used?
Cluster sampling
Convenience sampling
Stratified random sampling
Correct Answer: C
The analyst divides the bond population into credit-rating strata and draws a random sample from every category. This process is stratified random sampling.
Option A would involve selecting only certain clusters and studying bonds within those chosen groups.
Option B would select bonds primarily because they were easy to access.
Continue Your CFA Level I Prep With KeyPoint
Use structured lessons, practice questions, mock exams, and progress tracking to focus on the time you have left
FAQs About Sampling Methods
What Is the Difference Between Probability and Non-Probability Sampling?
Probability sampling uses a random selection process and gives population members a known chance of entering the sample.
Non-probability sampling relies on availability or analyst choice. Selection probabilities are unknown, which limits the analyst’s ability to measure sampling error and generalize the results.
What Is the Difference Between Stratified and Cluster Sampling?
Stratified sampling draws observations from every stratum. Cluster sampling selects only some clusters and then studies observations within the chosen clusters.
Stratified sampling usually aims to improve representation and precision. Cluster sampling often aims to lower collection costs.
When Is Stratified Sampling Better Than Simple Random Sampling?
Stratified sampling may be more useful when the population contains distinct subgroups that should all appear in the analysis.
It can also improve precision when observations within each stratum are similar and the differences between strata are meaningful.
Can a Large Convenience Sample Be Representative?
A large convenience sample may still exclude important parts of the population. Increasing its size adds more observations chosen through the same access-based process.
A representative sample requires suitable population coverage, not only a high observation count.
Is Judgmental Sampling Always Unreliable?
Judgmental sampling can be useful for exploratory research, specialist interviews, and situations where expert selection adds practical value.
Its results require careful interpretation because the analyst cannot objectively calculate each population member’s probability of selection.