HOW SINGAPORE EDUCATION WORKS · Coding open-ended university responses
Find your next learning step
To analyse short open-ended questionnaire responses, read the complete answers, define what your labels mean, apply those labels consistently and keep a route back to the original text. Count only after deciding whether your unit is a person, an answer or a coded passage. A keyword tally alone cannot tell you what respondents meant.
Full chapter index · Practice and checking · Sources and next reading · How Singapore Education Works
This guide uses a small descriptive codebook exercise: several labels may apply to one answer. It is not a universal recipe for qualitative research or a substitute for a supervisor's methodological instructions. The fictional data below let you inspect every decision without handling anyone's personal information.
The choices at a glance
| What you notice | Possible issue | Useful next move |
|---|---|---|
| Counts every keyword | Meaning and negation are being missed | Read the whole response against an inclusion rule |
| Totals exceed the number of people | Some people have multiple codes | Report assignments and people separately |
| New codes apply only to later answers | The codebook changed midstream | Version the rules and revisit every response |
Full chapter index
Define the coding job
Code and revise transparently
Recall and transfer
CHAPTER 1 OF 8 · Define the coding job
1. Start with the research question and the data boundary
Suppose a student group asks, “What makes it difficult to attend our optional revision session?” Their immediate aim is to describe barriers mentioned in the responses. They are not yet testing which barrier causes absence or estimating its prevalence across the whole university.
Keep the original wording, response identifiers and collection context. Work on a separate coding copy. A code is a short label attached to relevant text; a codebook explains when that label should and should not apply.
Our exercise assumes five complete fictional answers, one from each of five people, with multiple barriers allowed. In a real project, identify missing answers, repeated submissions and the sampling method before reporting a denominator.
If you have not yet collected responses, the earlier guide to university questionnaire design addresses question wording and piloting. Better coding cannot recover a question respondents interpreted in incompatible ways.
Read every answer once without counting. Note repeated ideas, unexpected points and statements that resist your first impression. A response may mention a barrier only to say it is not a problem.
For this exercise, use three initial codes. TIME means an explicitly stated clash with another commitment. TRAVEL means an explicitly stated difficulty reaching or returning from the venue. ACCESS means an explicitly stated problem obtaining the materials needed for the session.
Write an exclusion beside each definition. TIME excludes a time mentioned without a clash. TRAVEL excludes a statement that travel is easy. ACCESS excludes a positive comment saying materials were available. Leave an uncertainty note when the meaning is unclear.
These are modest descriptive categories chosen for this question. They do not represent all dimensions of access, and they are not a validated instrument. A different research aim might require different distinctions.
CHAPTER 3 OF 8 · Code and revise transparently
3. Worked case: five answers, four initial assignments
The responses below are the complete fictional dataset. The labels are our initial coding decisions under the definitions just given.
| ID | Full response | Initial codes |
|---|---|---|
| R1 | The session overlaps with my part-time shift. | TIME |
| R2 | I can attend the time, but the last bus makes the journey home difficult. | TRAVEL |
| R3 | My shift clashes with the session and I cannot open the required worksheet link. | TIME; ACCESS |
| R4 | Travel is easy for me and the worksheet opens. I would like more worked examples. | None of the three barrier codes |
| R5 | I can attend, but I do not know which room to go to. | Uncertainty note: room information |
R3 receives two codes because it explicitly describes two barriers. R4 receives none: the words “travel” and “worksheet” occur, but they do not describe the defined problems. R5 is not forced into ACCESS because our initial definition concerns materials, not location information.
Checking answer: TIME applies to two people, TRAVEL to one and ACCESS to one. There are four code assignments across three people with at least one of these codes. “No initial code” applies to two responses; it does not mean those responses contain nothing worth reporting.
Keep the passage supporting each assignment. Even in a spreadsheet, a column containing the relevant excerpt makes it easier to audit the link between label and meaning.
CHAPTER 4 OF 8 · Code and revise transparently
4. Improve the codebook without rewriting the past silently
R5 reveals a useful distinction: uncertainty about the location differs from difficulty opening learning materials. Add an INFORMATION code defined as explicitly missing practical information needed to attend, such as the room. Exclude requests for additional subject explanations.
Call this codebook version two. Record the date, author and reason for the addition. Then revisit all five responses under the new definition, not only R5.
R5 receives INFORMATION. R4's request for more worked examples still does not fit that code: the respondent is requesting learning content, not missing attendance information. It can remain a separate memo or become another category if the research question expands and the dataset warrants it.
Checking answer: after recoding, the four barrier codes have five assignments across four people. R3 still contributes two assignments. R4 remains outside these barrier codes. This distinction prevents the false claim that “five coded barriers” means “all five people reported a barrier.”
A small log preserves your reasoning: version one used three codes; version two added INFORMATION after identifying a location-information issue; every response was rechecked. Do not claim another researcher checked the coding unless that happened.
If two readers disagree, begin with the disputed text and the inclusion rule. Perhaps one reader treated any mention of travel as a problem, while the other required an explicit difficulty. That is a definition problem you can inspect.
If a code covers almost everything, it may be too broad for the question. If every response gets its own code, compare whether some labels describe the same idea at different levels of detail. Neither merging nor splitting is automatically better.
If important statements disappear from the summary, return to uncoded responses and memos. A minority view can be important even when it is not frequent. The number of mentions does not determine the seriousness of a barrier.
Software can store, retrieve and count assignments, but it does not decide whether your interpretation is justified. Follow the institution's approved arrangements for identifiable or confidential data, including any use of external tools. The fictional exercise requires no upload of real responses.
CHAPTER 6 OF 8 · Repair and report
6. Practise coding and report with the right denominator
For guided practice, consider a sixth fictional response: “The bus journey is fine, but the session clashes with my shift and the worksheet link will not open.” Apply version two without relying on isolated keywords.
Checking answer: assign TIME and ACCESS. Do not assign TRAVEL because travel is explicitly described as fine. There is no missing practical information, so INFORMATION does not apply.
For independent practice, return to the original five responses only. Write a two-sentence findings statement that gives the count for TIME, mentions the response with two codes and states one limit on generalisation.
Checking answer: “Two of the five fictional respondents mentioned a clash with another commitment. One response described both a clash and a materials-access problem; these five answers do not establish how common the barriers are across the university.” Equivalent wording is fine.
If you display percentages, name the denominator. TIME is two out of five people, or 40%, in this tiny exercise. Summing the version-two person percentages gives 100% here only by coincidence: overlapping codes mean the sum need not equal 100%.
CHAPTER 7 OF 8 · Recall and transfer
7. Recall the rules and transfer to another project
Tomorrow, explain the difference between a code assignment and a person without looking at the table. Then explain why R4 did not receive TRAVEL. Reopen the data to check the explanation.
Transfer to feedback about a community workshop for older learners. A person might mention transport, timing and helpful staff in one answer. Your university-session codebook is a starting hypothesis, not a ready-made fit. Read the new question and data before adapting it, and preserve positive as well as negative accounts when relevant.
There is a local-to-global methods connection. The IPSA-NUS Methods School's introduction to qualitative data analysis discusses research design, coding and transparent analysis in a specific course. It does not make that course's approach a requirement for every Singapore degree.
Belgium's KU Leuven guidance on documenting qualitative research emphasises recording what codes mean and who applied them. That documentation makes a project understandable beyond its original classroom or research team.
Is a code the same as a theme?
A code labels a relevant piece of data. A theme is an interpretive pattern developed through analysis in methods that use that concept. Renaming a frequent code “a theme” does not by itself establish an explanatory pattern.
Must every answer get a code?
Not every answer must receive one of your current substantive codes. Track relevant uncoded material and explain missing, unclear or out-of-scope responses. Do not confuse an empty answer with an answer that challenges the codebook.
Must every qualitative project use agreement scores?
No. Different methodologies make different assumptions about interpretation and collaboration. Follow the justified method for your project. This beginner exercise uses explicit descriptive definitions so you can examine decisions; it does not prescribe a reliability statistic.
Can I report a striking quotation instead of counts?
A quotation can illustrate a finding, but it does not establish how common the experience is. Preserve context, protect participants as required and explain why the example supports the claim.
Sources and next reading
Reviewed 23 October 2026. All respondents, quotations, categories, calculations and revision decisions in the worked case are fictional and original.
IPSA-NUS Methods School: Introduction to Qualitative Data Analysis provides the institution-specific Singapore connection. KU Leuven: Documentation and metadata for qualitative research supports the importance of clear code documentation.
Continue through the How Singapore Education Works hub for research, school and adult-learning guides.
Learning at other stages
Continue with a different learning task in this series. Each guide has its own worked cases, practice and limits.
Comparing collections in kindergarten