Written and maintained by CASRAI Editorial Board
Published
On this page
A quantitative research design is the plan that fixes how numeric data will be collected and compared so that a study can answer its question. The main types are descriptive, correlational, quasi-experimental and experimental designs, and they differ mainly in what the researcher controls and therefore in what the results are allowed to claim: descriptive and correlational designs support statements about associations, while experimental designs are the ones library methods guides describe as able to establish cause and effect. Choosing the design is a decision about the claim you need to make, made before any data are collected.
What is a quantitative research design?
Quantitative research is described by the University of Texas at Arlington Libraries as “the exploration of numeric patterns”, using methods such as questionnaires, structured observations or experiments. USC Libraries characterise it as concerned with precise measurement and replicable, controlled work used to predict events. The research design is the structure that sits between the research question and the statistics: it states who or what is measured, how many times, whether anything is manipulated, and how groups are formed.
This page covers designs inside quantitative research only. For the wider sorting of designs by data type, purpose and observational versus experimental logic, see categories of research design; for the contrast with non-numeric work, see qualitative vs quantitative research.
Types of quantitative research design
Descriptive design
A descriptive design measures variables as they are, without trying to influence them. Sheffield Hallam University Library says descriptive designs help to understand the current state of a phenomenon and are useful when little is known about it; variables are not controlled. USC Libraries add that subjects are usually measured once, that the design establishes only associations between variables, and that samples often run to hundreds or thousands of subjects. The typical product is a set of frequencies, means and percentages for a defined population.
Correlational design
A correlational design examines how two or more variables are related. UC Merced Library notes that with this design “it is not possible to tell which variable is affecting the other(s)”, and Sheffield Hallam repeats the standard warning that correlation does not imply causation. The California State University, Sacramento library describes the relationship as possibly positive or negative. The researcher does not assign or manipulate anything, so the same observed association could reflect either direction of influence or a third factor. See types of variables for the terms used to name those roles.
Quasi-experimental design
A quasi-experiment looks like an experiment but the researcher cannot assign participants to groups. UC Merced puts it as “like an experiment, but the groups cannot be assigned”, using pre-existing groups. Sheffield Hallam says it aims at cause and effect but assignment to groups is not random, with groups formed from predetermined criteria such as age. Sacramento State describes it as a design where random assignment had no role in determining who received which level of treatment, and says it generally has less validity than an experiment. Sheffield Hallam adds that matching participants on relevant characteristics increases the likelihood that observed differences are due to the independent variable. The dedicated comparison experimental vs quasi-experimental design goes further on this boundary.
Experimental design
In an experiment the researcher controls conditions to examine how one variable affects another. Sheffield Hallam states that true experiments seek to establish cause-and-effect relationships and that the researcher controls all variables except the one manipulated. USC Libraries describe subjects being measured before and after a treatment and say an experimental study establishes causality; they also note the sample may be very small and purposefully chosen. Sacramento State defines the control group as the group of subjects not exposed to the experimental treatment.
How the designs compare: what each can claim
| Design | Researcher manipulates a variable? | Groups | Strongest claim the sources support |
|---|---|---|---|
| Descriptive | No | Usually a single sample, measured once (USC) | Describes the current state of a phenomenon; associations only |
| Correlational | No | No assignment | Variables are related; direction of influence cannot be told (UC Merced) |
| Quasi-experimental | Yes, but groups are not assigned | Pre-existing or predetermined groups | Aims at cause and effect; less validity than an experiment (Sacramento State) |
| Experimental | Yes, all other variables controlled | Treatment and control groups | Cause and effect can be explored or established (UC Merced, Sheffield Hallam, USC) |
The data table below the main text sets out how five university library guides name these designs, because the vocabulary is not uniform. UT Arlington lists three designs (descriptive non-experimental, quasi-experimental and experimental) and does not name correlational as a separate type, and it observes that studies do not always explicitly state which kind of design they use. UC Merced, Sheffield Hallam and Sacramento State each give correlational its own entry. When you read a paper, judge the design by what was controlled and how groups were formed, not by the label the authors chose.
A worked illustration: same numbers, different claims
Illustrative composite example; the figures are invented for teaching and do not come from a real study. Suppose a school introduces an extra tutoring programme and a test is scored out of 100. Five students who received tutoring score 70, 75, 80, 85 and 90. Five who did not score 65, 70, 75, 80 and 85.
- Tutored group mean: (70 + 75 + 80 + 85 + 90) / 5 = 400 / 5 = 80.
- Untutored group mean: (65 + 70 + 75 + 80 + 85) / 5 = 375 / 5 = 75.
- Difference in means: 80 − 75 = 5 points.
- Mean of all ten scores, as a descriptive summary: (400 + 375) / 10 = 775 / 10 = 77.5.
The arithmetic is identical under every design; what changes is the sentence you may write. As a descriptive design, you report that the ten students averaged 77.5. As a correlational design (tutoring status recorded alongside scores), you report that tutoring and scores are associated, without saying which drives which. As a quasi-experiment using two existing classes, you may report a 5-point difference between groups while acknowledging that the classes may have differed before the programme. Only if students had been randomly assigned to tutoring or no tutoring, in an experimental design, would the 5-point difference be presented as an effect of the programme, and even then a significance test and sample size would be needed before concluding anything. For planning those, see sample size for a quantitative study.
How to choose a design
The criteria in the sources reduce to three questions, asked in order.
- Do you need to claim that one variable affects another? If not, a descriptive or correlational design is enough, and Sheffield Hallam points to descriptive designs when little is yet known about the phenomenon.
- Can you manipulate the variable? If the variable cannot be manipulated (age, an existing policy, a prior condition), you are limited to descriptive, correlational or quasi-experimental work.
- Can you assign participants to groups? If you can assign, and control the other variables, an experimental design is available. If groups already exist or are formed by predetermined criteria, the design is quasi-experimental.
Match the research question to the design first; quantitative research question examples by study design shows how question wording differs for each type.
Assumptions and common mistakes
- Inferring causation from association. USC Libraries advise against inferring causality, especially in nonrandomised designs.
- Treating statistical significance as practical importance. USC warns that results can be statistically significant but “humanly insignificant”.
- Expecting depth from numbers. USC notes quantitative results provide less detail on behaviour, attitudes and motivation.
- Ignoring pre-existing group differences. In a quasi-experiment, differences between groups may predate the treatment; matching on relevant characteristics (Sheffield Hallam) reduces but does not remove this. See internal vs external validity.
- Calling a design by the wrong name. A study without random assignment is not a true experiment, whatever the label in the abstract.
USC lists strengths too: quantitative designs allow a broader study with more subjects and greater objectivity and accuracy of results.
How to report your design (APA)
In an APA-style Methods section, state the design in plain terms near the start, then describe participants, how groups were formed, the measures, and the procedure. The APA’s reporting standards for quantitative articles are summarised in Journal Article Reporting Standards (JARS); check the current APA Publication Manual for the exact items required. An illustrative sentence (not drawn from a real study):
We used a quasi-experimental, nonequivalent-groups design. Two existing classes were compared on end-of-term test scores; participants were not randomly assigned to conditions.
Name the design honestly. If assignment was not random, say so in the design statement, and carry that limit into the interpretation. For more on numeric data you may collect, see quantitative data examples.
FAQ
What are the main types of quantitative research design?
Descriptive, correlational, quasi-experimental and experimental. Some guides, such as UT Arlington’s, list only three types and omit correlational as a separate name.
Which quantitative design shows cause and effect?
The experimental design, where the researcher manipulates a variable and controls the others. Quasi-experiments aim at causal questions but, without random assignment, carry less validity.
What is the difference between correlational and experimental design?
A correlational design measures how variables are related without manipulating them, so direction of influence cannot be told. An experimental design manipulates one variable under controlled conditions.
Can a descriptive study use hypothesis tests?
Descriptive designs mainly summarise a phenomenon as it is; USC states they establish only associations. Which tests you can defend depends on your design and sampling.
Does quantitative research always need a large sample?
No. USC says descriptive samples are often hundreds or thousands, while an experimental sample may be very small and purposefully chosen.
Sources
- USC Libraries, Organizing Your Social Sciences Research Paper, Quantitative Methods: libguides.usc.edu/writingguide/quantitative (read 2026-11-12)
- UC Merced Library, Research Methods, Approaches: libguides.ucmerced.edu/research-methods/approaches (read 2026-11-12)
- Sheffield Hallam University Library, Quantitative research: libguides.shu.ac.uk/researchprojects/quantitative-research (read 2026-11-12)
- California State University, Sacramento Library, Quantitative research: csus.libguides.com/res-meth/quant-res (read 2026-11-12)
- UT Arlington Libraries, Quantitative and qualitative research: libguides.uta.edu/Quantitative_and_qualitative_research/Quant (read 2026-11-12)
Data: How five university library guides name quantitative research designs
Cross-walk compiled from five university library methods guides read on 2026-11-12, showing which designs each guide names, whether correlational design is listed separately, and what each says about causation.
Period covered: 2026.
| Library guide | Designs named | Correlational listed as its own design | Quasi-experimental named | What it says about causation |
|---|---|---|---|---|
| USC Libraries (Quantitative Methods) | Descriptive; experimental | No | No | Descriptive establishes only associations; an experimental study establishes causality |
| UC Merced Library (Approaches) | Observational; correlational; experimental; quasi-experimental | Yes | Yes | Correlational cannot tell which variable affects the other; in experiments causation can be explored |
| Sheffield Hallam University Library | Descriptive; correlational; true experimental; quasi-experimental | Yes | Yes | Correlation does not imply causation; true experiments seek cause and effect; quasi aims at it without random assignment |
| CSU Sacramento Library | Descriptive; correlational (called comparative); quasi-experimental; experimental | Yes | Yes | Quasi-experiments generally have less validity than experiments |
| UT Arlington Libraries | Descriptive non-experimental; quasi-experimental; experimental | No | Yes | Notes studies do not always state which design they use |
Sources
- USC Libraries - Quantitative Methods (accessed 2026-11-12)
- UC Merced Library - Research Methods: Approaches (accessed 2026-11-12)
- Sheffield Hallam University Library - Quantitative research (accessed 2026-11-12)
- CSU Sacramento Library - Quantitative research (accessed 2026-11-12)
- UT Arlington Libraries - Quantitative and qualitative research (accessed 2026-11-12)
Compiled by CASRAI from the sources above. Free to reuse under CC BY 4.0 with attribution.







