10th Biennial ACSPRI Social Science Methodology Conference in conjunction with the Australian Social Network Analysis Conference (ASNAC)

Open-text questions can add qualitative depth to large-scale surveys, but their value depends on how the resulting text is analysed and interpreted. This paper examines what short qualitative survey data can contribute to the study of complex social behaviour. It uses free-text accounts of how people responded to workplace discrimination and harassment in the 2025 Australian Survey of Social Attitudes (AuSSA) as a case study.

Participants who reported experiencing or witnessing workplace discrimination or harassment in the previous five years were asked, in their own words, what actions they took or who they spoke with. We analysed data from 301 participants, producing 215 accounts from targets and 243 from bystanders.
We used directed content analysis, drawing the initial coding structure from existing frameworks of target and bystander behaviour. These frameworks supported comparison with previous research and anchored interpretation in the existing evidence base. They also provided a useful discipline in a sensitive area where assumptions about how targets or bystanders “should” respond can reproduce misleading or harmful stereotypes. Using established frameworks helped make our own interpretive assumptions more visible and reduced the risk that those stereotypes would inadvertently shape coding categories. Additional categories were developed where accounts could not be accommodated within the a priori schemes.

The analysis used a coding-reliability design. Two coders applied a fixed codebook independently, agreement was assessed statistically, and disagreements were resolved by consensus. Both coders also kept analytic notes throughout the process to record observations not captured by the coding structure.

The qualitative data revealed complexity that would have been difficult to capture through closed-response categories showing how context could change the interpretation of behaviour. Leaving a workplace, for example, sometimes followed unsuccessful attempts to seek help or resolve the problem through formal channels, complicating its interpretation as a passive response. Accounts of taking no action also distinguished between seeing no need to act and perceiving available pathways as ineffective, unsafe or blocked. These distinctions illustrate the risks of interpreting behavioural categories without considering the circumstances participants describe. The analysis also highlights the limits of short qualitative survey data. Accounts were often too brief to establish chronology, intent or outcome with confidence.

This case study demonstrates the value of combining systematic coding with attention to the context and limitations of individual accounts. Open-text questions can complement closed-response measures by preserving participants’ framing and revealing distinctions that predefined categories may overlook.