HomeDondena

Dondena Interviews

Giovanna Merli, who visited Dondena this semester from Duke University, is shedding light on Chinese emigration to Africa, and using new network-based sampling approaches to do it.

 

 

If somebody misses your talk, what are they going to miss. Give me your two-minute elevator speech about Chinese immigrants in Africa.

First I’ll tell you why we decided to focus on that population.

 

That was going to be my next question, but we’ll start with that!

I have studied China and Chinese populations for the last 25 years, so China provides the context for most of my research to date. Emigration flows from China have been rising very rapidly over the past two decades and migration today is an increasingly important demographic and social stratification variable.   But it is not easy to recruit population-representative samples of migrants. Migrants are often a rare and hidden population, that is, in addition to often representing a small proportion of the general population, they are also hidden in the sense that there is no conventional sampling frame of migrants that would enable researchers to draw a sample with known probability of selection for each sampling unit. There is no complete list of Chinese residents in Tanzania. Also, were we to go the conventional route and draw a population-based sample of the Tanzania population, the large-scale screening required to ensure the inclusion of a minimum number of Chinese residents for analytical samples would be prohibitively expensive.

Over the past few years, I have become very interested in methods to recruit samples of hidden populations that rely on respondents’ networks to recruit other respondents. I thought that the difficulty of reaching Chinese migrants in Africa for sampling would offer a good testing ground for these new methods, these new network-based sampling approaches. The study of Chinese migration to Africa would also be of significant substantive interest because Africa is a recent, and to some extent unexpected, destination for Chinese international migration. So, with some colleagues, we decided to pilot a survey to test these new methods in Tanzania, a country that has seen a rapid increase in Chinese immigration.  

China is a sending country of emigration, historically. Chinese have migrated everywhere, but then there was a halt to international migration after the founding of the People’s Republic in 1949. For about 50 years China was virtually closed. At least until the end of the 1980s it was really hard to get passports. I mean, only people who went on official trips abroad would get passports, or students who were part of exchange programs, so it was very hard to emigrate internationally. There has been, of course, a lot of internal rural-to-urban migration in China, especially since 1979, but the current flows of international Chinese migration are really a new phenomenon, and emigration from China has taken off for different reasons. One is that the Chinese government encouraged large state-owned enterprises to compete and invest overseas. Large, state-owned Chinese enterprises are moving up the value chain and shifting jobs to Africa, closer to the natural resources they need, with the result being that growing numbers of Chinese laborers, skilled and unskilled, are sent to Africa and a growing number of spontaneous migrants are attracted to increasing economic opportunities in Africa. The issuing of passports increased tremendously around 2000, and that is probably connected with China’s entry into the WHO. So it is easier today to leave China, there are more people who are leaving and more destinations to reach in Africa, Latin America, and Europe.  I already alluded to two types of Chinese migrants in Africa. One is state-led, so those who are sent to Africa to work on a Chinese large-scale, state-owned investment project; and the other one is spontaneous, comprising entrepreneurs, small traders, and their families who go to Africa to open stalls, restaurants, engage in trade, much like the Chinese here in Milan.

 

How many people are doing this?

The numbers are not known. That’s what everyone is asking, how many Chinese are in Africa?  Different African countries have had different histories of immigration. Except for South Africa and Mauritius, where there are already 3rd or 4th generation Chinese (many of them from Taiwan), Chinese immigration to Africa is recent, it started in earnest around the late 1990s. The figure of 1 million Chinese across Africa is the figure we see most often cited in the press, but it is not clear where that figure comes from. Really nobody has an accurate estimate.

So the main rationale for my study, besides its methodological appeal, was the quantitative description of the demography and social organization of a Chinese community in Africa. We observed significant differences between state-led and spontaneous migrants in terms of socio-demographic characteristics, and I think we have been quite successful at describing them. In addition, this study allowed us to test a new network-based sampling approach.

 

And what did you think of the methodology, how did it perform?

You know, it was a pilot. We did not have a huge amount of resources, this was not an NIH-funded study. It was funded internally at Duke by the Duke Global Health Institute and the Provost’s office, and we are very grateful to them. Logistically it was a hard study to carry through. Trained Chinese-speaking interviewers are nowhere to be found in Dar es Salaam, so we recruited and trained Chinese-speaking interviewers in the U.S. who then travelled to Tanzania to implement the survey.  Before data collection started, we made numerous trips to Dar es Salaam, and worked very hard to establish a good rapport with the Chinese community, so when data collection started, it proceeded quite smoothly.

Then it took a really long time to obtain approval from the relevant research unit in Tanzania and so, because of our academic schedule here in the U.S., we only had about three months to organize and carry out the data collection. So the study was relatively small—that is, small in terms of the size of the sample we recruited. To recruit the sample, we used a network-based sampling approach called Network sampling with memory (NSM). This was developed by my colleague Ted Mouw at Chapel Hill and his student Ashton Verdery, who is now an assistant professor at Penn State. There is a nice methodological introduction to NSM in a 2012 issue of Sociological Methodology. NSM is an improvement over Respondent Driven Sampling (RDS). RDS is widely known and used in the public health field. It is a peer recruitment link-tracing sampling design where referrals from a set of respondents are used to recruit other respondents so as to generate samples of hidden populations. But RDS relies on multiple important assumptions about the referral process and the structure of the underlying population network, which are often not met in practice. Like RDS, NSM is a link-tracing sampling design. Different from RDS, NSM collects detailed network data from each respondent—that is, it asks respondents about their friends and acquaintances who are members of their social networks. This information is then used to reconstruct the underlying social network of the target population. Relative to RDS, this is more labor-intensive because it requires interviewers to collect a network roster from respondents. But we have shown that in the context of Chinese migrants in Africa, that it is completely feasible to collect this type of information.

 

Like a family tree?

Right! We ask multiple waves of respondents minimally identifying information—the last four digits of their cell phone number, and their last name—on members of their social networks. So this information then allows us to link all the nodes of the network and to identify the individuals who have been nominated by multiple respondents. The method also comes with a sampling algorithm, which is usually run during the interview to direct the sampling process over the network and allows the sampling algorithm to direct the sampling process towards as-yet-unexplored areas of the network. So researchers have a high level of control on the sampling process. This improves the accuracy and precision of the resulting estimates.

Instead, with RDS, it is respondents who guide the sampling process.  You instruct respondents to recruit in a way consistent with assumptions but you don’t have as much control over the sampling process, and these assumptions about respondents’ referral are often not met. In other work, we have shown with data collected among female sex workers in China that these assumptions do not hold, leading to biased estimates of population characteristics.

 

Tell me about in your field, and define it as broadly as you like, what do you like about what you see going on right now and what do you wish you could see more of?

There is a narrow and broad definition of my field. I am a demographer, I was trained in demography and my field is population sciences. For some demography is narrow, encompassing a set of techniques by which data collected in censuses, surveys and vital registration systems about age, sex, births, deaths, migrations, marriages are used to describe population structures and compute vital rates, and demographic models that infer regularities in the relationship between population structures and vital rates. This is often referred to as “formal demography.” For others, demography is much broader and encompasses the study of fertility, mortality, and migration from a variety of viewpoints including sociology, economics, anthropology, biology, etc... As such, demography is inherently multidisciplinary and it is often referred to as “population sciences,” a term which encompasses the treatment of relation between aggregate demographic structures and rates and their individual-level social, economic, biological, and political determinants. 

What do I like? I like the increasing interdisciplinarity of the population field, the need for the contributions from other disciplines to provide increasingly more complete and satisfying explanations of the relationship between social, economic, cognitive, biological factors, and demographic outcomes. But I also like the application to demographic studies of methods, like the application of simulations, which allow one to understand the linkages and feedbacks between aggregate population structures to individual-level factors. The social networks toolkit is finding a comfortable place among methodological innovations in demography that help the understanding of the relationship between micro and macro processes.  

Also what I like is the growing amount of rich data sources that enable population scientists to carry out their research. Today, we rely on increasingly complex, often longitudinal, study designs fielded across a variety of settings to collect data that enable us to explain the causal mechanisms of demographic behaviors, conduct new empirical tests, and elaborate new theories of behavior. This was unthinkable only 20 or 30 years ago.

 

To learn more about Giovanna Merli, click here. To learn more about the Dondena Seminar series, click here.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Last updated 08 October 2015 - 14:09:10