Research and evaluation questions are a small part of the LCSW and LMSW exams, but they are among the most predictable. They test vocabulary applied to practice: which design answers the question, what a measure can and cannot show, what an evaluation actually found, and what a researcher owes the people in a study. Learn a few dozen terms and their traps, and these items stop being guesses.
What each outline covers
The 2026 outlines treat this material differently by exam. From our map of the four outlines:
| Topic | Bachelors | Masters | Advanced Generalist | Clinical |
|---|---|---|---|---|
| Progress and effectiveness of intervention | Yes | Yes | Yes | Yes |
| Evaluating agency programs | Yes | Yes | Yes | Yes |
| Methods for evaluating practice effectiveness | Yes | Yes | Yes | No |
| Basic and applied research design | No | Yes | Yes | Yes |
| Quality assurance | No | No | Yes | Yes |
| Program objectives and outcomes | No | No | Yes | Yes |
| Program evaluation and its impact | No | No | Yes | Yes |
| Research ethics | No | No | Yes | Yes |
So the LCSW exam carries more of this material than the LMSW exam: quality assurance, program evaluation and research ethics are on the Clinical outline and not the Masters one, and the Advanced Generalist exam carries them too. Research ethics sits under Values and Ethics; the rest is in content area III. The outlines' own examples name the concepts to know: needs assessment, formative and summative evaluation, cost-effectiveness and cost-benefit analysis, outcomes, data collection and analysis, reliability and validity, and qualitative and quantitative methods. ASWB's suggested references for every exam include Rubin and Babbie's Research Methods for Social Work (10th edition, 2025).
Measurement: reliability and validity
Reliability is consistency: the measure gives similar results under similar conditions. The kinds to know are test-retest (the same person, a second time), inter-rater (two observers, the same behavior) and internal consistency (the items of one scale hanging together).
Validity is accuracy: the measure captures what it is meant to. Content validity asks whether it covers the whole idea; criterion validity whether it agrees with an established measure or predicts a later outcome; construct validity whether it relates to other measures the way the theory behind it predicts.
Know the relationship between the two exactly. A measure can be reliable and still not valid, like a scale that is always two pounds off, but an unreliable measure cannot be valid, because a result that changes at random cannot be accurate.
Two more distinctions appear in vignettes. Qualitative methods, such as interviews and observation, explore meaning and experience; quantitative methods count and compare. And a standardized instrument has fixed items, administration and scoring, with published evidence of its reliability and validity, while a scale built for one client, such as a self-anchored rating of distress, is quick and personal but has no such evidence behind it.
Research designs, and what each can claim
- Experimental designs assign people to groups at random, which makes the groups comparable and is what lets a study attribute a difference to the intervention.
- Quasi-experimental designs compare groups formed without random assignment, so differences between the groups could explain the result.
- One-group pretest-posttest designs measure the same people before and after. They can show change, but not that the program caused it.
- Correlational studies show that two things move together. They cannot show which causes which, or rule out a third factor.
- Basic research builds knowledge for its own sake; applied research answers a practical question about a program or population.
The reasons a before-and-after change may not be the intervention's doing are the threats to internal validity, catalogued by Donald Campbell and Julian Stanley in 1963. The ones to know: history (something else happened at the same time), maturation (people change on their own), testing (taking the measure changes the score), instrumentation (the measure itself changed), regression to the mean (people selected for extreme scores drift back toward average), selection (the groups differed from the start) and attrition (the people who dropped out differed from those who stayed).
Sampling has its own short list. Probability samples, such as simple random, stratified and cluster samples, give everyone a known chance of selection and support generalizing. Convenience, purposive and snowball samples do not, however large they are.
Evaluating your own practice: single-subject designs
Single-subject, or single-system, designs evaluate the work with one client, family or group. The target is measured repeatedly, first during a baseline phase (A) and then during the intervention phase (B).
- AB: baseline, then intervention. Simple and practical, but other events could explain the change.
- ABC: a second, different intervention follows the first, and C marks it.
- ABA and ABAB: the intervention is withdrawn and, in ABAB, reintroduced. If the target changes each time, the case for the intervention is stronger, but withdrawing something that is helping can be unethical, especially when the target is risky, such as self-harm or aggression.
- Multiple baseline: the intervention starts at different times across clients, behaviors or settings, so improvement that follows each start points to the intervention without withdrawing it.
- Changing criterion: the target is raised in steps, and performance should follow each step.
The data are usually analyzed visually, comparing the level, the trend and the variability across phases. Good practice measures throughout, not only at the end, and treats the client as a partner in choosing what to measure. ASWB's own sample questions tie evaluation to the end of the work too: one recall item asks what the termination process is MOST often associated with, and the key is evaluating goals and accomplishments.
Evaluating programs
Needs assessment comes first, and Jonathan Bradshaw's 1972 taxonomy of social need is a common frame: normative need (what experts or standards say people should have), felt need (what people themselves say they need), expressed need, or demand (what they actually ask for, such as a place on a waiting list), and comparative need (a group lacks a service that similar groups receive).
The kinds of evaluation. CDC's evaluation framework defines formative evaluation as checking whether a program is feasible and appropriate before it is fully implemented, with the aim of improving it; process evaluation as whether it was carried out as planned; outcome evaluation as whether it achieved its intended outcomes, which cannot by itself establish causality; and impact evaluation as comparing outcomes with what would have happened without the program. Summative evaluation is the judgment at the end: did it work, and should it continue?
Outputs and outcomes. Outputs are what a program does, such as sessions held and people served. Outcomes are what changes for the people it serves. A report full of outputs, or of satisfaction ratings, has not shown that anything changed. In a logic model the chain runs from inputs to activities to outputs to outcomes.
Costs. A cost-benefit analysis puts the outcomes in money and compares them with the costs. A cost-effectiveness analysis leaves the outcome in its own units and compares what each unit costs, such as the cost per client housed, across programs aiming at the same outcome.
Quality assurance is different from evaluation: it checks, continuously, whether services meet set standards, through case-record review, utilization review, and external reviews such as audits and accreditation.
Research ethics
The NASW Code's standard on evaluation and research, 5.02, sets out what the exam expects. Social workers should monitor and evaluate policies, programs and practice interventions. Where appropriate they need participants' voluntary, written informed consent, with no penalty for saying no and no undue inducement to say yes, and they should consult the appropriate institutional review boards. Participants keep the right to withdraw at any time without penalty. Social workers must protect them from unwarranted distress and harm, keep their data anonymous or confidential, and report findings accurately, with nothing fabricated or falsified.
These items work like the other ethics questions: the option that protects the participant beats the one that protects the study. Behind the Code sits the Belmont Report of April 18, 1979, with three principles, respect for persons, beneficence and justice, and three applications: informed consent, the assessment of risks and benefits, and the fair selection of subjects. Two distinctions come up often. Data are anonymous when even the researcher cannot link them to a person, and confidential when the researcher can but will not disclose. And recruiting your own clients needs particular care, because people who depend on you for a service may not feel free to say no.
How these items are written
Some research items ask for a definition and others are short vignettes, and the vignettes reward the same habits as the rest of the exam; how to read an ASWB vignette sets out the method.
- Ask what the design can claim. A one-group before-and-after change does not prove the program caused it, and a correlation does not prove either variable caused the other.
- Separate outputs, satisfaction and outcomes. Only outcomes show that people's lives changed.
- Prefer measuring to assuming. When a vignette asks how a worker can know whether the work is helping, look for the option that measures it, along the way rather than once at the end.
- Put participants first. Consent, voluntariness and confidentiality outrank a cleaner study, and a participant who wants to stop can stop.
- Read the qualifier. BEST means best on the facts given, so for a busy caseload it may be a practical design rather than the most rigorous one; the qualifier words explain the difference.
Our Practice Evaluation and Research chapter has a lesson for each of these topics. The LCSW exam cheat sheet puts this material in context, and how to pass the LCSW exam and how to pass the LMSW exam map each full outline.
Read at source on October 22, 2026: ASWB's August 2026 Examination Guidebook (content outlines, suggested references and sample items); the NASW Code of Ethics, standard 5.02; the Belmont Report (National Commission for the Protection of Human Subjects of Biomedical and Behavioral Research, April 18, 1979), read in the Internet Archive's copy of the HHS page; and CDC's Approach to Program Evaluation (updated August 18, 2024). The threats to internal validity follow Campbell and Stanley's Experimental and Quasi-Experimental Designs for Research (1963), and the needs taxonomy Jonathan Bradshaw's "The concept of social need" (New Society, 1972). Outline coverage is counted from the 2026 outlines as mapped in our course. ASWB Master is not affiliated with, endorsed by or approved by the Association of Social Work Boards.