Mixed Methods Research in ICT4D: When and How to Combine Qualitative and Quantitative Approaches

The dominant research methods in ICT4D are qualitative — case studies, ethnographic work, interpretive inquiry. This is appropriate: the field’s central questions are often about processes, mechanisms, and context, and qualitative methods are well-suited to these. But qualitative research alone cannot answer questions about scale (how widespread is this pattern?), representativeness (is this case typical?), or comparison (are outcomes better in Program A than Program B?).

Mixed methods research — combining qualitative and quantitative approaches within a single study — has grown significantly in the ICT4D literature as researchers have recognized that neither tradition alone is sufficient for the field’s full range of questions.

Why Mixed Methods?

The core argument for mixed methods is simple: different research questions require different research methods. A study that wants to understand both what is happening (qualitative) and how much it is happening (quantitative) requires both approaches. A study that uses quantitative data to identify patterns and qualitative data to explain those patterns — or vice versa — is doing work that neither alone can do.

The weakness of purely qualitative ICT4D research is that its findings are hard to generalize — findings from one or a few cases may not apply elsewhere, and there is often no systematic evidence for how widespread a phenomenon is. The weakness of purely quantitative ICT4D research is that it can identify correlations without explaining mechanisms — knowing that mobile money adoption is associated with higher consumption does not tell you why, or what conditions produce this relationship.

Mixed methods can address both weaknesses.

Designs for Mixed Methods Research

Sequential Explanatory Design

Quantitative data is collected and analyzed first, then qualitative research explains the quantitative findings.

Example in ICT4D: A survey of 500 smallholder farmers finds that farmers who used an agricultural advisory SMS service had 15% higher crop yields than non-users. This finding raises the question: why? Qualitative follow-up interviews with a subsample of farmers (users and non-users) explore what advice they received, how they used it, and what other factors influenced their yields. The qualitative phase explains the quantitative pattern.

Appropriate when: You have quantitative data (or can collect it relatively efficiently) and need qualitative insight to explain what the numbers show. Also useful when quantitative results are surprising or counterintuitive.

Sequential Exploratory Design

Qualitative research is conducted first to develop understanding, then quantitative research tests or extends the findings.

Example in ICT4D: In-depth interviews with community health workers in three sites identify that the most important barrier to consistent use of a mobile data collection tool is not technical literacy (as program designers assumed) but concern about data privacy — workers are worried that patient data could be accessed by unauthorized people. A subsequent survey of 200 CHWs across 12 sites tests whether privacy concern is a widespread barrier, and which aspects of the tool design are associated with greater or lesser concern.

Appropriate when: The research domain is understudied and the right survey questions are unknown until qualitative work has been done. This design avoids the common mistake of large surveys that measure the wrong things because the survey developer did not understand the domain well enough.

Concurrent Triangulation Design

Qualitative and quantitative data are collected simultaneously and integrated in the analysis.

Example in ICT4D: A study of a digital health program simultaneously conducts health facility surveys (quantitative: service delivery statistics, technology use rates) and observation and interviews at a subset of facilities (qualitative: how the technology is actually used, what staff think of it, how it affects clinical workflow). The two data streams are analyzed separately and then compared for convergence or divergence.

Appropriate when: Neither method alone would provide sufficient understanding, and the research timeline does not permit sequential collection. Concurrent designs are more complex to manage but can be more efficient.

Embedded Design

One method is embedded within a primarily single-method design to answer subsidiary questions.

Example in ICT4D: A large randomized evaluation of a mobile literacy program (quantitative) embeds a qualitative process evaluation within it — observational work with a subset of program sites to understand how the program is being implemented and what the implementation looks like in practice. The qualitative work does not change the quantitative evaluation design but provides essential interpretive context.

Integration: The Core Challenge

Mixed methods research is only mixed — rather than just two separate studies — if the methods are genuinely integrated. Integration can happen at multiple points:

At the design stage: Designing the qualitative and quantitative components so they address complementary questions and the data sources can be meaningfully combined.

At the data collection stage: Sampling qualitative respondents from the quantitative sample (so the same individuals contribute to both) or using quantitative findings to guide qualitative sampling (selecting cases for in-depth study based on quantitative patterns).

At the analysis stage: Using quantitative findings to select interpretive frameworks for qualitative analysis; using qualitative themes to code quantitative open-ended responses; building explanatory models that combine quantitative relationships with qualitative mechanisms.

At the interpretation stage: Presenting findings in integrated form — not “the survey found X, and separately the interviews found Y” but “the survey found X, and the interviews explain why through the mechanism Z.”

Poor integration — presenting qualitative and quantitative findings in separate sections that never connect — is common in nominally mixed methods research and represents a failure to realize the method’s potential.

Challenges in Mixed Methods ICT4D Research

Researcher capacity: Rigorous mixed methods requires fluency in both qualitative and quantitative traditions — skills that most researchers develop in one tradition during training. Teams that include both qualitatively and quantitatively trained researchers need effective collaboration and communication across methodological cultures.

Resource requirements: Well-designed mixed methods studies are more expensive than single-method studies. Both collection phases require investment in design, data collection, and analysis. This is a real constraint in ICT4D research contexts with limited funding.

Time: Sequential designs take more time than single-method designs. Funders who want results quickly may push toward one or the other method rather than a sequential approach.

Publication conventions: Academic journals have historically been organized by method — qualitative journals and quantitative journals. Publishing well-integrated mixed methods research is sometimes complicated by journals that are more comfortable evaluating studies within a single methodological tradition. This is changing as mixed methods has become more established.

Quality Criteria for Mixed Methods Research

Evaluating the quality of mixed methods research requires applying appropriate criteria from both traditions:

From the quantitative tradition: Validity and reliability of measures, appropriate sample size and sampling strategy, appropriate statistical analysis for the data and research question.

From the qualitative tradition: Credibility of interpretations (supported by data, subjected to member checking), transferability (sufficient contextual description for readers to assess applicability), dependability (transparent methods), confirmability (findings grounded in data rather than researcher projection).

Mixed methods-specific: Adequacy of integration (are the methods genuinely combined or merely adjacent?), adequacy of design fit to the research question, coherence of the rationale for using multiple methods.

Frequently Asked Questions

Is mixed methods research more credible than single-method research? Not inherently. A rigorous single-method study is more credible than a poorly designed mixed methods study. Mixed methods is more powerful when the research question genuinely requires it — when understanding the phenomenon requires both the depth of qualitative and the breadth of quantitative approaches. Using both methods to appear comprehensive, without genuine integration, adds cost without adding value.

How large should the quantitative sample be in mixed methods research? It depends on the specific quantitative research question. If the quantitative component is descriptive (how widespread is this pattern?), sample size depends on desired precision. If it is comparative (are outcomes better in Group A than Group B?), it depends on expected effect size and statistical power requirements. There is no fixed rule — consult a statistician or quantitative methods reference for the specific analysis planned.

Can I use mixed methods in a single researcher dissertation? Yes, though it is challenging. Sequential designs are more manageable for a single researcher than concurrent designs. The key is that both components be genuinely rigorous — not a perfunctory quantitative survey to satisfy a requirement for mixed methods.

What software is used for mixed methods analysis? Qualitative analysis is typically done with NVivo, Atlas.ti, or MAXQDA. Quantitative analysis with SPSS, Stata, or R. For analysis that bridges methods — such as quantitizing qualitative data (turning themes into categorical codes for frequency analysis) or visualizing mixed data — different software may be used at different stages.

How do ICT4D researchers typically report mixed methods findings? Most journal papers present mixed methods findings in integrated results and discussion sections, with the qualitative and quantitative threads woven together. Some papers present findings in sequential sections and integrate only in the discussion. The integrated approach is more rigorous but requires more careful writing.

Development Informatics