2.5 How can we delve deeper into these processes?

The last activity on quantitative data revealed some big trends at the national level, but we concluded that it is difficult to establish whether and how migration is linked to IG.

We need qualitative data at the micro-level to try to understand these possible connections. As noted in the last section, we used semi-structured interviews with some key participants. Conducting these interviews usually followed some well-worn paths, but equally, we needed to tailor the approach to the specific project rather than use a one-size-fits-all approach.

For MIAG, we had to:

  • identify possible participants
  • discuss the project and get their consent to participate
  • conduct the interview
  • transcribe the interview
  • analyse the transcription through a process of coding.

In the next activity, you are going to get involved in the latter stages of this process by coding some of MIAG’s interview transcripts.

Before that, read this short description of what a semi-structured interview involves. As you read, consider why might it be a good technique to unpack some of the processes linking immigration to inclusive growth.

What are semi-structured interviews?

The semi-structured interview is a qualitative data-collection strategy, where the researcher asks informants a series of predetermined but open-ended questions. The researcher has more control over the topics of the interview than in unstructured interviews, but in contrast to structured interviews or questionnaires that use closed questions, there is no fixed range of responses to each question.

Researchers who use semi-structured interviewing develop a written interview guide in advance. The interview guide may be very specific, with carefully worded questions, or it may be a list of topics to be covered. The interviewer may follow the guide to the letter, asking the questions in the order they are given, or the researcher may move back and forth through the topic list based on the informant's responses. In either case, the topics of the interview guide are based on the research question and the tentative conceptual model of the phenomenon that underlies the research.

(Given, 2008)

The key is that while we had what Given describes as a ‘tentative conceptual model’ – that migration is linked to IG – we did not want to assume what the mechanisms were that linked the two. We wanted the participants to tell us about their experiences as immigrant business owners, and having a range of topics to be discussed seemed the best way. It also left room for them to go off in directions that we hadn’t necessarily predicted, but which gave us useful insights into the questions we were seeking to answer.

Activity 2.7: Qualitative data analysis exercise

Timing: Allow approximately 45 minutes

Step 1

When we code interview transcripts, we need to have some themes or questions in mind. These were largely inclusive growth processes as set out in the MIAG framework around economic and non-economic factors. For this activity, we are going to focus on one question: to what extent, and in what ways, do migrant-owned businesses create jobs in the local economy?

Step 2

We have selected two extracts (both around 500 words in length) from our semi-structured interview transcripts that were originally conducted in Kenya in September 2020:

These transcripts have been anonymised so you cannot tell who might be speaking. Read the transcripts and use the highlighter function to identify any phrases, sentences or longer quotes that you think answer the question about job creation. Remember to save your documents.

What can you say in response to the question about job creation? Do immigrant businesses create jobs? Who gets these jobs? Are these good jobs?

By signing in and enrolling on this course you can view and complete all activities within the course, track your progress in My OpenLearn Create. and when you have completed a course, you can download and print a free Statement of Participation - which you can use to demonstrate your learning.

Activity 2.8: Coding transcripts

Timing: Allow approximately 30 minutes

Now watch Videos 2.10–2.13, where Craig Walker and Ben Lampert, both of The Open University, discuss how the MIAG project coded and interpreted these transcripts in relation to the IG concepts.

As you watch the four videos, use the space below to make notes on the following questions:

  • Why did MIAG use semi-structured interviews?
  • How were the interviews designed?
  • How was the data analysed?
  • How did MIAG make sense of this data?
Show transcript|Hide transcript

Transcript: Video 2.10 Part 1: Asking structured questions.

Ben Lampert:
Hi. I’m Ben Lampert. I’m a member of the MIAG research team, and I’m here today with Craig Walker, a fellow colleague on the MIAG team who’s been heavily involved in leading the research on the project. So Craig, what I wanted to talk to you about today was your involvement in the semi-structured interviews. You’ve used a number of different methods in the project: why did you decide to add semi-structured interviews into the mix?
Craig Walker:
Yeah. Thanks, Ben. I think inclusive growth is such a complex concept. But not just that theoretical understanding, but actually to analyse it empirically. We’ve seen this from all those frameworks that are out there.
That multitude of multilateral agencies from the World Bank, United Nations, African Development Bank – they all have these complex sets of indicators, because it is such a slippery slope. It’s such a slippery idea. And I think to really develop MIAG’s understanding of inclusive growth, we needed to reflect that complexity in the data that we were collecting. So we had the quantitative data – so we were gathering existing data from the World Bank, International Labour Organisation to build up our framework of indicators. But that quantitative data is very descriptive. So it will tell you what is happening, but it’s very difficult – not impossible, but difficult – to really get to grips with the cause and effect. So while we might see things like GDP going up, increases in employment, we’re not necessarily getting behind it to understand why that is happening.
Now, moving to the survey, that allowed us to start to do this. But again, a survey is largely quantitative data. We introduced some open-ended questions – and bearing in mind, the survey was very much focused on the migrant business owner themselves, so we were mining down behind that macro data at the national level to start to unpack and understand the migrant business owner, but it was still largely quantitative data.
The semi-structured interviews, being qualitative, provide a richness and depth that you can’t get through that quantitative data. So they really allowed us to start to get behind those descriptive accounts, which they did in two ways.
First of all, that individual migrant journey: they provide us with detailed interpretations of the individual migrants’ stories to really interrogate and explore the intricacies of their lives and their business operations.
But secondly, what the qualitative data allows us to do through those interviews is move from what is happening to why. So it really allows us to start to make those causal links. So for example, looking at that survey data, what we noticed quite early on is that large numbers of migrant business owners were employing nationals – which we hadn’t necessarily expected, because in the literature, there is perhaps a tendency for migrant business owners to employ members of their own family, or members of their own nationality. So for example, Indians in Kenya, there is a literature on that, and we had, perhaps, expected, for reasons of familiarity, that they would employ more Indians. That wasn’t what we were seeing. But what that survey data didn’t allow us to do is to start to unpack it and understand why that was happening.
So for example, then, through the interviewing, it allowed us to mine and ask specific questions around that. So we started to find out some of those business owners were employing nationals because it was cheaper labour than if they were employing other Indians, or that, actually, the nationals held knowledge which helped their business operations. So for example, they knew the landscape of petty politics at the local level. So they were helping navigate the bureaucracy with filing applications for particular work permits, business permits. Also, for example – again, we saw this in Kenya quite a bit – is you will have nationals fronting a business. Just for ease of setting up and operating businesses, you will have a foreign migrant who is working with a national – the national is actually fronting that business for them.
So while MIAG focused on migrant business owners, these interviews weren’t exclusive. The survey was; we did focus just on migrant business owners. But actually, we broadened out who we were interviewing.
Ben Lampert:
OK. Thanks, Craig, that really makes a lot of sense, and it’s clear that the interviews really add something that you couldn’t quite get from the survey and from the background quantitative data.
End transcript: Video 2.10 Part 1: Asking structured questions.
Video 2.10 Part 1: Asking structured questions.
Interactive feature not available in single page view (see it in standard view).
Show transcript|Hide transcript

Transcript: Video 2.11 Part 2: Designing an interview.

Ben Lampert:
It’s clear that the interviews really add something that you couldn’t quite get from the survey, and from the background quantitative data, but how do you go from realising that they’re going to add something to actually designing interviews that can get at these things, and working out what to ask all the different types of respondents that you’re engaging with in this project?
Craig Walker:
Yeah, good question. As I’ve just said, we broadened out who we were interviewing, because we needed to build up a broader picture than just the experience of the migrant business owners themselves. You should bear in mind that MIAG had a very practical focus of wanting to inform migration policy as a result of this project.
Now, we broke our sample of interviews down into five key groups, the logic being that, as you’ve just said, each of them adds something different to complement the data and our understanding that we’re building up for that empirical analysis through the interviews.
So those five groups, the first are what we termed MDAs. So these were ministries, departments, agencies – so effectively, government – any formal body that had some kind of connection to migration and migration policy.
The second were CSOs, so civil society organisations. And we targeted, there, specifically organisations who will be working with or supporting migrant populations in those countries. We engaged with a couple of charities, for example, that support micro start-ups for migrant business owners.
Third category were what we termed representative – so these could be, loosely, community leaders. So for example, in Nigeria, there is the Nigerian British Business Forum. So their role is to broker business relationships between Britain and Nigeria, so you will find British migrant business owners there. And the logic of targeting that group is that the interviews with the migrant business owners themselves allow us to understand that individual journey. But what the representatives do is they give us that broader picture to look at the wider trends of, say, the British migrating to Nigeria over the last 30 years – any changes in levels of migration, types of sectors, people might be investing in, et cetera.
And then the fourth were the migrant business owners themselves, which were the largest sample group. And then the fifth and final were nationals. So we wanted to interview and speak with, say, Mozambicans who were working for migrants, because it just introduces a very different perspective, because I think with interviews, the actual technique itself, there is a risk that people are telling you what they think you want to hear – so actually speaking to a range of people who are working for these migrant business owners just introduced that extra layer of experiences.
So the entire sample size, the interviews, were 115 in each of the countries. So we ended up with 460 interviews! And the structure of those questions – so while each of them, each of those five groups would have had questions specific to them: so for example, with the MDAs, so the government interviews, there were very specific questions there around the types of data that government may be collecting on migration, how that data is stored, how it is shared across different agencies, and how it might be used to inform policies. Very specific questions for that group that aren’t necessary to ask the others.
But we had a consistent line of questioning across all of the five, which was designed specifically to try and understand the different ways that migrants were contributing to growth. And the way that we did this is the line of questioning reflected that framework that we developed. So those indicators, it allowed us to be consistent in terms of our line of inquiry across that quantitative secondary data, to the survey, and then into our interviews.
So for example, we had a section looking at inclusive growth through quality of employment – so how many jobs were being created, what the gender balance was in employment, the ratio of nationals being employed to migrants. Another section was looking at the innovations in the market – so types of new products migrants might be introducing, the skills, the knowledges that they bring, and whether they’re creating competition with local companies.
Also, we had sections on wider social and political implications. So here we were looking at issues, as I’ve said, with those representatives. Are migrant groups organising – self-organising – through associations? Are they joining unions? But also, because of us looking at inclusive growth, the ways in which migrants are actually giving back to their community.
So there were questions around what you could broadly term CSR. So are they giving back to the communities? How do they do that? But then also, how migrants might be engaging politically – so do they have a voice? Are they able to engage with government? Are they being consulted?
Ben Lampert:
That’s really interesting. I think what strikes me, out of all of that, is just how much data you’ve gathered there. That’s over 400 interviews you mentioned, each covering quite an array of different factors related to migrants’ lives, and to inclusive growth. So how do you begin to start analysing all of that once you’ve had it all transcribed?
End transcript: Video 2.11 Part 2: Designing an interview.
Video 2.11 Part 2: Designing an interview.
Interactive feature not available in single page view (see it in standard view).
Show transcript|Hide transcript

Transcript: Video 2.12 Part 3: Refining the data.

Ben Lampert:
What strikes me out of all of that is just how much data you’ve gathered there. That’s over 400 interviews you mentioned, each covering quite an array of different factors related to migrants lives and to inclusive growth. So how do you begin to start analysing all of that once you’ve had it all transcribed? Where do you start?
Craig Walker:
Yeah, 460 interviews. It’s a big job when some of those transcripts are 20 or 30 pages long! So we started by developing the code book, and to do this, we revisited that original framework because we needed that consistency across all those data sets.
So we had those key, inclusive growth variables based on that overall conception. But there is a risk, as we know, that inclusive growth is very complex. So we had to be quite systematic; we didn’t want to be too rigid with the coding. But we needed that deductive start, so using that conceptualisation to keep that consistency of framing.
So what we had, was we had three key codes that we started with in that code book. So ‘migrant’, ‘inclusive growth’ and ‘governance’ were the overarching codes. Then what we did was we broke each of those down into key dimensions, or aspects, of each. So for example, when it came to the migrants, we had subcategories, which were ‘migrant motivations’, ‘migration history’ and also ‘socioeconomic’ dimensions of their the lives. And so each of these categories allows us to start to mine down a little bit further into the detail.
We went down to a third level with each of the codes. For example, if we look at inclusive growth, we had a sub-code within that around ‘employment-generation’ and ‘skills development’. But it required more fine-tuning. So within that, we had a code specifically, for ‘jobs host’. So we would code data here, which referred to any passage in the transcript where a migrant business was creating a job or a job opportunity for the hosting community. So in Ghana, for example, you had a migrant business owner who was employing a local Ghanaian.
So the codebook provided us with this structure for interrogating the interview. So we would go through all of those individual transcripts and be highlighting and coding particular passages against each of those individual codes. That: very much a deductive approach to analysis and coding of the data, but actually, there was such a richness to some of these interviews, that as we started doing that – as we started that coding process – we very quickly started to notice findings emerging that we hadn’t anticipated. So we introduced some flexibility into the coding process. So here we moved more from that deductive, having a framework to work with, to more inductive, allowing some of those findings to inform new codes that we hadn’t anticipated.
So for example, as I have a particular interest in the political dimension of this project, and while we had, within that governance coding structure, formal engagement and political participation, so for example, we had a code, ‘Does government actually widely consult migrant communities when they’re informing policy?’ Likewise, ‘Do migrant communities organise to advocate or lobby against government?’ We had these formal ways in which we were coding for political participation engagement. But I was seeing in some of the interviews that I was doing, particularly in Nigeria, interesting patterns emerging where migrant business owners, they were telling me that, actually, we see government as a bit of a problem. So if we can keep them at arm’s length, it’s better for our business operations. And there were different strategies that they were employing to basically avoid, or circumnavigate, local government, because they saw it as hampering their business operations.
This is a form of informal politics that we hadn’t anticipated. So we introduced a code around that.
Ben Lampert:
OK Craig, I get that about the categories developed from looking at the inclusive growth framework and using that to deductively guide your coding, but how about things that emerged from the interview? Surely there was some stuff that you just weren’t expecting or didn’t necessarily fit neatly under the code you’ve already generated from the inclusive growth framework? Was there any examples of that happening?
Craig Walker:
Definitely. And another one that we saw – that we added – was local capacity, and how that impacts on business operations. There is the obvious here, in terms of hard infrastructure: so are there roads, secure power supplies to allow the businesses to operate?
But we noticed, coming through some of those interviews, that migrant businesses struggled with local capacity in other ways. And we’re talking, here, about sourcing raw materials having to be imported. But also in terms of, maybe, levels of education or training, that nationals had that they needed to operate their businesses.
One example that sticks out, actually, is we interviewed an Indian business owner in Kenya. And his business was events management. And to run his business, he said he cannot source any of his equipment within Kenya – the quality doesn’t exist at the moment, actually. So a barrier for him, and his business operations, was him having to import everything that he needs, technically, to run those events, from Europe.
Ben Lampert:
I really get a sense of a kind of two-pronged approach happening there, where you’ve got this deductive coding based on the inclusive growth framework and the kinds of indicators you’re expecting to identify in the data using that framework, but also letting stuff emerge from the interviews that you maybe didn’t expect or come through more strongly than you expect. Or in some ways isn’t captured by the inclusive growth framework –` you then inductively created new codes to capture those unexpected and new insights. So I really get how those two processes work together.
End transcript: Video 2.12 Part 3: Refining the data.
Video 2.12 Part 3: Refining the data.
Interactive feature not available in single page view (see it in standard view).
Show transcript|Hide transcript

Transcript: Video 2.13 Part 4: Findings and insights.

Ben Lampert:
What I’d like you to tell us a bit more about is just moving from that process of coding through to actually making sense of the data in a broader sense, developing findings from that. You’ve got all these codes, but how do you how do you actually then determine what your key findings are and the key insights that have come out of the interviews as a whole?
Craig Walker:
It’s actually quite a difficult process to manage when you have such a large data set like this: 460 interviews. So we had to work very closely and collaboratively as a team. The first step was actually that we couldn’t rely on just individually [reading] those transcripts. We had to use some form of qualitative analysis software, which in our case was NVivo.
And we worked as a small team within the wider team by dividing those transcripts up and everybody individually coding their transcripts against that code book. But coding can be very subjective: what I’m coding against a particular code, how I understand it – can’t guarantee that you are in the same way. So we had, over a number of weeks, actually meeting as a team, running through that code book, testing it against a couple of transcripts. How are you understanding migration history? How are you understanding innovation? How are you understanding knowledge transfer? To try and, as close as we can, introduce that reliability that we are actually coding across our own sets comparatively, basically – so that we can try and ensure as close as we can we’re all coding in the same way.
And actually, it was really interesting. Through that process and through those meetings, we started to share insights and some emergent findings. One of the interesting ones very early on was actually, when it comes to gender, and some of the ways in which migrant business owners were talking about female employment and why they would perhaps choose to employ females over males. So through that very collaborative process, we were able to start to inductively see some of these findings emerging.
The second was that once we’d actually got an entire coded set on NVivo. Everybody could individually then interrogate that data for their own particular interests, because we are a large, multidisciplinary, multi-country team. We’ve got economists, qualitative researchers, sociologists, geographers. So everybody could then take away that data set and, through the software, you can mine that data by running reports against particular codes that you are interested in. For me, it’s the politics and the political participation; but for others, it’s something completely different.
And then the third was that once we got that data, you almost go back to the beginning of the project. So it was revisiting the literatures that are already out there on inclusive growth – those gaps that we’ve been identifying at the beginning of the project, those big questions that we were asking: how is that interview data speaking to what we were originally posing in this project?
Ben Lampert:
OK. Thanks, Craig. I guess there’s just conversations going on there at different levels. There’s the conversations between the team members and building up between what they’ve coded and how they see the bigger pictures forming from that coding, and then a conversation also between what’s emerged from the data and what we set out to do in the original research questions at the start of the project.
And that’s where we start to get the genesis of those findings that it will be written up as the project’s outcome, basically. So thanks, Craig. I think that’s a really rounded overview of what we’ve tried to do and how doing these semi-structured interviews has really added so much to the project, and really enabled us to link the patterns that we see in the quantitative data to causal mechanisms and social relationships that make these changes happen on the ground.
End transcript: Video 2.13 Part 4: Findings and insights.
Video 2.13 Part 4: Findings and insights.
Interactive feature not available in single page view (see it in standard view).
To use this interactive functionality a free OU account is required. Sign in or register.
Interactive feature not available in single page view (see it in standard view).

Discussion

Compare your answers to the following notes:

  • Why did MIAG use semi-structured interviews?
    • IG is complex to theorise and analyse empirically
    • MIAG had quantitative data, but this is descriptive
    • Need to understand why things are happening
    • The survey started to address the ‘Why?’ questions, but is still quite quantitative
    • They provide richness and depth of the individual migrant journey and to understand causal links; for example, focused on employment practices
  • How were the interviews designed?
    • Widened the scope of who was interviewed
    • Five groups were interviewed: government, civil society organisations, representatives such as community leaders, migrant business owners and nationals who were connected to migrant businesses
    • A range of people allow us to get different perspectives
    • 115 interviews per country, 460 in total
    • Questions were tailored to each group of respondents
  • How was the data analysed?
    • Some interviews were quite long
    • Developed a code book, so MIAG went back to original IG variables (this is a deductive approach)
    • Had to be open to contextual issues
    • Started with three broad codes (migrants, IG and governance) and then broke these down further into different levels
    • Code transcripts using these codes
    • As the project started MIAG found lots of richness, so they moved to a more inductive approach and added these as new codes (for example, around politics)
    • Some findings fall outside our expected IG concepts, such as local capacities
    • In summary: a two-pronged approach
  • How did MIAG make sense of this data?
    • It was not easy, because the data set was so big
    • Worked closely and collaboratively
    • Used qualitative analysis software
    • Had to meet as a team to generate a shared sense of how to interpret the transcripts; this led to greater consistency across the team and allowed people to share thematic findings
    • They then had to revisit the IG literature and see how these emergent themes spoke to these concepts and theories

The discussion between Ben and Craig around the semi-structured interviews reveals MIAG’s attempts to capture the complexity of the links between migration and development. The key here was around causality: that is, we can use other data to discern some patterns, such as discovery of oil leading to more skilled inward migration. But these patterns don’t tell us how all this happens, so qualitative methods are one way to explore the causal mechanisms linking migration and economic opportunities.

The question of causality linking migration and development, and vice versa, is highly complex and – as the experts discussed in Activity 2.4 – is also context-specific. By focusing on immigrant business, MIAG homed in on a critical area of causality: how businesses operate and how they create growth opportunities. Week 3 will analyse this in more detail and you will also learn more about other methods used by MIAG.

2.4 Analysing migration and inclusive growth data: quantitative

Conclusion