Lisa Madlberger, Òscar Coromina, Mylynn Felt, Jeroen de Vos, Emilio Fernández, Laura E. Cañuelas, Alexei Tsinovoi, Ana Ranitovic, Asaf Nissenbaum, Gisele Costa, and Mariola Pagan
In 2013, the Rana Plaza factory collapse in Bangladesh killed over 1,100 people (The Economist, 24 April 2014) and created a surge in western public awareness to the topic of unsafe and unethical factory conditions. This led to movements such as the #CleanClothes and #WhoMadeMyClothes campaign. The Rana Plaza disaster was not the first to generate discussion of factory conditions; however, it heightened global attention to the issue. The western public discourse of factories is dominated by two frames: unjust working conditions versus responsible company representations (Deegan, 2008).
This opens a question about how marginalized groups would represent themselves since mainstream media channels primarily use experts to speak for them. Schneider, Chamberlain, and Hodges (2010) conducted a content analysis of newspaper representations of homelessness. They conclude that the discourse is dominated by the voice of experts speaking for the homeless with only a few quotes by transient people speaking in their own behalf (p. 161). In another study, Radley (2010) gave cameras to homeless individuals and asked them to take whatever pictures they wanted and then explain them in an interview. Results show that the photos are similar to those taken by other groups, such as hospital patients. This includes photos of nature and friends posing and smiling -- not typical images portraying homelessness in newspapers. Chang (2009) sought to share the stories of migrant workers in Chinese factories. She notes an absence of factory worker self-representations in literature on factories.
This notion makes us curious about finding the voice of those working for and living in the communities of the textile manufacturing companies. How do locals characterize the factory? There are a few examples of factory workers reaching out to those who will eventually purchase the product they create. For example, one factory line worker left images of herself in the factory on a new phone (Sorrel, 2008). Others have left messages on the tag labels of clothing they sew (Rustin, 2014).
Recent studies note an increasing saturation of smartphones used across the global market. In fact, there are now more smartphones than people globally (CISCO, 2015). This is placing the internet in the hands of many who cannot afford a computer but who own smartphones. Much has been said of the enabling power of social media to give voice to grassroots civic movements (see, for example, Bakardjieva, Svensson, & Skoric, (2012); Khamis and Vaughn, 2011). With some skepticism that we would find a large populace of textile workers with regular access to social media, we, nevertheless, sought to explore the regional public discourse of textile factories associated with Nike, H&M, and Adidas.
The goal of our project was to answer the following questions:
Can we find the voices of the workers online?
What are regional differences in worker representations?
What is the difference between NGO, Brand, and Worker Representations?
Where are the most leaky (i.e. highly visible on social media) factories to be found?
Do different platforms host different voices?
We started off trying to focus on one of the most urgent places for investigation. The collapsing garment factory Rana Plaza in Bangladesh in 2013 brought the international news to focus on the labor conditions in the Bangladesh' textile industry. Therefore, we thought this event would be a good starting point to limit down our research question. However, just after starting enquiring on the topic we ran into language problems, for the target population only speak Bengali and we did not. It is also a language that uses a non-Latin alphabet. Instead, we decided to focus on places with more accessible languages, considering the language capacities present in our group was quite extensive: Russian, Spanish, Portuguese, English and Dutch. Using the list of names we obtained from three big clothing brands: Nike, H&M and Adidas, we were able to identify a set of almost a thousand factories spread over our five countries of choice: Indonesia, India, Mexico, Brazil and Turkey. In a sense, we used the lists of the brands as a form of expert list to get insights in the field of research. These extensive lists were only to be meant as a starting point, knowing that each of these factories are not producing exclusively for these brands, but can work for various brands from day to day.
The countries were presented with the following amounts of reported factories:
Country | Number of Factories |
India | 300 |
Mexico | 43 |
Indonesia | 189 |
Turkey | 323 |
Brazil | 102 |
The list of factory names was retrieved from the company websites:
Adidas | http://www.adidas-group.com/media/filer_public/d5/8d/d58dec52-283a-4b5a-91a8-9079d9e4898d/jan_2015_-_subcontractor_factory_list.pdf |
Nike | http://manufacturingmap.nikeinc.com/ |
H&M | http://sustainability.hm.com/en/sustainability/downloads-resources/resources/supplier-list.html |
We used the names of the factories as a starting point to discover content on four different platforms: (1) Facebook (2) YouTube (3) Twitter (4) Instagram.
Since factories sometimes have lengthy names, with parts of the name denoting the legal form, we realized that the keyword or hashtag searches querying for exact matches provided by the APIs of these services are too restrictive. Therefore, we developed a work-around using google scraper to search the sites of these platforms using the search functionality of Google.
To identify the Facebook pages related to the factories we used queries of the following format:
site:Facebook.com "Factory name" inurl:pages - inurl:directory
From the result tab-file returned by the google scraper, we extracted the likes for each Facebook page in order to identify the most active pages for each country. In a next step, we used Netvizz to scrape the pictures from these pages.
Taking the factory names as a starting point for enquiry in Facebook was not very fruitful for both Brazil and Mexico; therefore, we experimented using other techniques. Key in these different approaches was local knowledge: fortunately enough we had Oscar with knowledge of Mexico and Gisele coming from Brazil in our research team. Oscar new the local term for textile workers to be Maquiladoras ('machinists'), and we found out with Giselle that the local (mostly illegal) textile workers positioned in Brazil were called Trabalho Escravo ('working slaves'), terms which will be further elaborated on in the findings. We used these local, more political-oriented keywords for search query in the YouTube video channel scraper, to get insights into the online video space of these topics.
With the entire group we tried to find related video content of the respective factories chosen from the list. In order to do so we took the list of factories and entered it into the google scraper to search on www.youtube.com. This brought us a list of related youtube videos which would, in turn, be used within the Netvizz tool to quantify the engagement with each of these individuals videos on Facebook. Based on the engagement level we were able to distinguish the more important results from the less interesting ones, which was not based on YouTube's internal search query, but instead uses Google's indexing power.
To retrieve relevant data from Instagram we applied two strategies:
We used three procedures in order to geolocate the factories, all of which yielded limited results. The first procedure consisted of extracting the GPS coordinates for each factory through their respective addresses, first using the DMI Geolocator tool, then using the Yahoo Maps source code in Google Sheets. However, both of these provided very broad results and insufficiently precise GPS coordinates due to the lack of uniformity and standardisation of the addresses. The second procedure entailed extracting the Facebook Place ID for each factory through the Place pages, which gave more precise results. In the third procedure, we manually searched for GPS coordinates of the factories on Google Maps. Having done all three searches, we compiled a list of coordinates which we then input into the Instagram scraper. We retrieved numerous results; however, a smaller number of these were factory-relevant, while most consisted of pictures that appeared to have been taken outside of the premises of the workplace, leading us to conclude that the users may be posting material taken outside of the factory whilst at the factory due to the internet access it provides.
We also conducted a search on Instagram using factory names and the keyword textile factory as seeds, which yielded more factory-relevant results, though in many cases, similar material as that posted on Facebook.
We applied the same technique to identify factory-related Tweets. We used the following query to find Tweets using the Google Scraper.
site:Twitter.com "factory name" inurl:status
We identified the central organization connected to the issue of labor conditions in textile factories, being the Clean Clothes Campaign and ran a co-like analysis in Netvizz to identify the central actors, and created a Gephi network. We also ran the Issue Crawler based on these results, but since the retrieval process took a long time results were not ready for the presentation. We used the name of the central actors produced by co-like analysis to scrape Instagram based on user names and Facebook public pages.
We didnt have the ability to create Twitter database, and YouTube tools were not relevant for the type of visual analysis we aimed for.
Like_Network_cleanclothes_on_facebook.pdf
We coded the visual representations in the NGO images based on the criteria used for the factory and employee images as well. Finally, we montaged the NGO images onto a single visual surface.
We scraped pictures from the websites of H&M, Nike, and Adidas aiming to find how the factory workers were visually represented from each of the companys perspective. To do this, we used Google Image Scraper and filtered the three queries with the same keywords: workers, factory, manufacturing, manufactured, manufacture, manufacturer. Down Them All was used to download the results, and they were saved in different folders, according to the company. Afterwards, we did a montage of the three sets of pictures using Image Plot.
To better understand the wide patterns of of the resulting corpus, we performed a content analysis on the sampled images. The analysis was limited to images showing people, in order to avoid language barriers and remain focused on the workers themselves. We used the following coding scheme to assign a topic and gender representation to each of the images
Topic
Work (work situations, trainings, work-related activities)
Play (religious activities, after-work life, sports, fun)
Protest (protest activities, accidents, emergencies)
Gender
Men (only men visible)
Women (only women visible)
Mixed (men & women visible)
Yes, we can find the voices of the workers online. In our project we found content expressing the perspective of textile factory workers on facebook, youtube and instagram.
Using our data collection methodology (factory name search) we could retrieve content for Turkey, India and Indonesia, while for Mexico and Brazil we mostly found official factory sites providing marketing content rather than insights from factory communities.
Having examined the platforms, our initial finding is that the voice of factory workers is indeed present in social media, in the form of their shared pictures depicting daily life from their perspective. However, this is evidently more true in certain factories and regions than in others.
Our content analysis of Facebook and Instagram images shows that while there are more images depicting individuals in work scenarios, there is also some representation of play, e.g. work parties and after-work life. There is even a certain amount of critical content. The images represent men more than women; however, there are also many images of men and women working together.
We performed a cross-country analysis to examine regional differences in the content being displayed online.
It is worth mentioning that we did not find any factory-related content in Brazil. We hypothesize this might be due to the predominance of immigrant workers in Brazilian factories, who would have less access and higher risk while using social media. Breaking the data down, it is clear that the different countries depict similar patterns of topics in images. However, Turkey tends to post more work-related content, while India displays more non-work-related imagery. Mexico has the highest rate of critical content, but it is also noteworthy that Turkey had none. We also looked at images appearing on NGO social media, which display a clear emphasis on critical content and very little play activities, unlike the samples from the various factories.
For gender representation, the cross-country differences are much clearer. In Turkey and India, men are represented alone much more often, while in Mexico women have a small majority; there is a greater female majority in Indonesia. NGOs are again different than the factories, showing more representation of women and mixed images, and very few men alone. Together with the data on image topic, it appears that content from NGOs tends to treinforce preconceived notions of factory life.
Comparing the images retrieved from the factories, from the NGOs and the brand websites, we found that the representations differ:
While brands tend to show more women and generally picture workers in work scenarios and NGOs showing more representations of women and mixed images, and very few men alone,
the pictures retrieved from factories show more men than women or mixed images.
If we look at images appearing on NGO social media picturing factory workers, we can see a clear emphasis on protest content.
We used Image Plot to organize pictures by color and saturation, which allowed us to go deeper into identifying which company has given workers a place in their webpage and which has not. While H&M has a balanced distribution of products, nature and workers, Adidas includes no workers at all. It is mainly focused on branding with products and people with certain lifestyles. However, Nike was positioned in between this polarization since it includes a few (no more than 5) pictures of factory workers.
Although using different strategies, we found that it was difficult to locate the content posted by factory workers in Latin America; In contrast, in India and Indonesia we found multiple factory communities represented on facebook and factory related content on Youtube and Instagram, by using the factory names or locations as search criteria. Based on these findings, we suggest the concept of leaky factories - factories from developing countries that have ample social media presence, as expressed in active profiles on social networks and multiple related items on web-based content platforms. We cannot discern if this is due to a purposeful policy or lack of opposition by the factory, the profile of workers and their social media habits, or something else entirely. What is clear is that this pattern typifies certain factories and not others. Furthermore,find our sample shows that these leaky factories can be found more often in Asia than in Latin America.
Shahi Export - An example of a leaky factory
On a Shahi Exports Facebook page, we found pictures of small children holding signs protesting child labor, as well as a YouTube video recording a company meeting intended to address sexual harassment in the work place. One of the most shocking contents we found was certainly a YouTube amateur video from three years ago that has very little explanation but clearly shows a body being carried out of a building. Along with an abundance of content from Facebook pages and some tagged images from Instagram, this factory displays a greater tendency to appear in social media than the vast majority of its peers.
As a last point of imagery analysis, we tried to compare the two most informative platforms Instagram and Facebook. On Instagram we mainly find non-work images which might be because of a geographical based scrape gathering pictures of the informal sphere. Moreover, Instagram had no images containing criticism and did show a majority of women (which is probably country-related as well). Facebook, on the other hand, had mostly men alone or mixed images, while offering some critical positions on the topic. It should be mentioned that this analysis is based on an unbalanced sample, containing 58 items from Instagram and 463 from Facebook.
We also observed a tendency for workers at different levels to utilize different platforms. For example, in Indonesia we saw more line workers using Facebook while office staff preferred Instagram. Corporate-sponsored Facebook pages and YouTube videos also reflect more of the executive voice. However, since we did not explicitly code this attribute, this is an hypothesis to be tested in future research.
The factory name search was not as effective in Mexico as using the local term maquiladora, which we identified due to the Spanish speakers in our research team. This refers to the practice of outsourcing manufacturing to sweatshops in order to bypass government control.
We used the word maquiladoras as a keyword for getting the top 100 YouTube videos related to the issue in Mexico and Central America which then were used as a seed for accessing to related videos with the YouTube Data Tool. As a result, we obtained a network of 3666 connected videos that most of them were not related at all with the issue. However, filtering our sample and zooming into the non-profit and activism category on YouTube, we can see the efforts of different NGOs to report abuse on the textile factories through documentaries and awareness campaigns. Interestingly, we can clearly ground issue spaces by looking at clusters that correspond to the different regions in which maquiladoras are spreading (Maquiladoras began in Mexico and is gaining momentum in other Latin American countries, as this image allows us to see).
The tag cloud of the description field of the Video content in the NGO shows the pictures of maquiladoras as a place in which exploiting children and women is common. Also, we can see the cities where factories employing these practices exist. While no factory names appear, we do see one brand, Nike, has a small mention.
Although the term Maquiladora is in Spanish, and was born in Mexico, this type of factory was quickly to be found throughout entire parts of Central America. However, looking for maquiladoras in Brazil was not as fruitful as we hoped it to be. However, one of our team members is Brazilian, and shortly elaborated on possible local terms for these workers, which brought us to the name "trabalho escravo na moda", which means 'clothing working slave'. However, doing a Facebook and YouTube (video network) scrape brought us mainly into a journalistic and NGO filled sphere, whereas the voice of the workers seemed to be absent at all. Most documentaries found on YouTube are journalistic accounts of the factories conditions. These documentaries are part of a campaign to give a serious alarm to society (about contemporary slavery in Brazil).
We found ourselves lost in our search for the voice of the workers, and asked another expert for help grasping the local conditions. Carlos, a Brazilian journalist who works with public communication and health press office, was able to tell us that in the factories the workers are mostly illegal migrants from Cuba and Bolivia trying to make a living. This is an inherent part of the Maquiladora's nature, producing in the grey area just outside the law; this is not very surprising. However, as we hypothesized later, potentially these workers are completely absent in online self-representations due to their illegal status. Unfortunately, this finding also represents the clear limitation of doing online based research; for us it would be very hard to either verify or falsify this hypothesis without any local empirical research or the use of different expert resources.
When comparing the results retrieved using the google scraper with the results retrieved using DMI TCAT to query the Twitter Search API, we observed that we were able to draw tweets up to five years ago, but TCAT provided more content, just only recent tweets.
Since Scraping google results is a time-intesive process (due to the CAPTCHA requests) this technique is only suitable for small numbers of Tweets; therefore, it is not applicable for trending topics.
For non-trending topics it might, however, be useful to discover historic tweets that can lead to relevant userprofiles, which can be used as a seed for retrieving user-timelines.
In Indonesia, Instagram allowed us to geo-locate the pictures which were published in a 30 metres radius from the factories. This function is available through the Instagram Scrapper, where one can scrape images based on their GPS coordinates and a specified radius around it, or a specific Facebook location ID. Such query is limited because these parameters are not available for every image. Nevertheless, the images that did contain them portrayed leisure moments or breaks in the daily work activity. They are joyful images, individual selfies or group ones in which we observed mainly young women sometimes with their factory as a background. Instagram in Indonesia is fundamentally a feminine activity.
Textile factory workers do have a presence on social media platforms. Through their images, posts, and videos, they portray factory work as well as evidence of work-based communities participating in after-work events and work parties. Of the five countries we examined, Indian factory workers conveyed the most play versus work imagery. Turkey portrayed the reverse situation, with a strong emphasis on work. It is unclear whether this difference reflects factory work practices in each location or if the difference is due to variant expectations about use of social media related to work life; however, there are differences in the types of work photos typically posted by textile workers in each region. While images tied to factory pages or containing captions with factory names portray work scenarios most frequently, followed by social gatherings tied to work relationships, some also portray a voice critical of factory practices. These portrayals, however, are largely drowned by the more prevalent images of happy or engaged workers.
Discourse tied to factory-life is not currently accessible through common hashtags. We found that searches using keywords, such as maquiladora and trabalho escravo, and factory names allowed us access to social media content focused on textile factory life. Google queries directed at each platform proved more directed and fruitful than searches within the platforms, themselves. Using platform-relevant digital tools such as Netvizz, we were able to analyze factories according to each region and identify the ones that seemed to leak more factory-worker voice. These leaky factories are more common in India than in Mexico or Brazil.
The majority of the social media discourse related to factories in the Latin Americas was related to working conditions in textile factories. However, this content comes, not from the factory workers, but from NGOs and activists speaking in their behalf. It is speculated that even if workers in these factories utilize social media, their precarious status as largely immigrant workers limits their digital voice on their working situation. Indian companies such as Shahi Exports, by contrast, foster enough social media presence related to work life that a space opens up for critical content as well as praise for factory life. Facebook and Instagram images from this company convey many work parties, work-sponsored athletic teams, and after-work events. However, there are also images of children holding signs protesting child labour as well as YouTube videos portraying a death at the factory as well as a meeting to address sexual harassment. It is speculated that the greater the social media presence of factory workers discussing work-in-general, the more opening is created for the possibility of dissent.
Preliminary results indicate differences between which tier of factory employees utilize different social media platforms. This is discussed further in the section on Future Research. However, one limited finding on this question is that in Indonesia, the textile factory workers who use Instagram are predominantly female.
Locals mediate their everyday life in the factory throughout social media in a largely playful way. Individual and group selfies are common among factory workers, as they are for social media users in all professions. This is unsurprising since the mediated portrayal of individuals through social media typically represents a carefully constructed image of the best self. This finding supports what Radley (2010) noted when homeless people captured pictures of nature and of their friends smiling rather than the kind of images that would evoke sympathy for individuals in dire need of support. Social media posts are one element more of the simulacra. Individuals project throughout various platforms the pleasant part of their lives. The devastating reality of their work conditions are more often portrayed by the NGOs and activists which denounce the inhuman labour conditions of many of the workers. Only in the leaky factories where a largely accepted social media voice is present do textile workers appear to voice criticism of working conditions.
Bakardjieva, M., Svensson, J., & Skoric, M. (2012). Digital citizenship and activism: Questions of power and participation online. JeDEM -eJournal of eDemocracy and Open Government, 4(1), i-iv.
Chang, L.T. (2009). Factory girls: From the village to city in a changing China. New York, NY: Spiegel & Grau. ISBN 9780385520188.
CISCO (3 Feb. 2015). Cisco Visual Networking Index: Global Mobile Data Traffic Forecast Update 20142019 White Paper. Visual Networking Index. Online: http://www.cisco.com/c/en/us/solutions/collateral/service-provider/visual-networking-index-vni/white_paper_c11-520862.html
Deegan, M.A.I.C., (2008) Motivations for an organisation within a developing country to report social responsibility information, Accounting, Auditing & Accountability Journal, 21(6) 850-879.
Khamis, S. and Vaughn, K. (2011). Cyberactivism in the Egyptian revolution: How civic engagement and citizen journalism tilted the balance. Arab Media & Society 14(3) 1-25.
Radley, A. (2010) What people do with pictures, Visual Studies, 25(3), 268-279. DOI 10.1080/1472586X.2010.523279.
Rustin, S. (25 June 2014). This cry for help on a Primark label can't be ignored. The Guardian. Online:
http://www.theguardian.com/commentisfree/2014/jun/25/primark-label-swansea-textile-industry-rana-plaza
Schneider, B. Chamberlain, K., and Hodges, D. (2010). Representations of homelessness in four Canadian newspapers: Regulation, control, and social order. Journal of Sociology and Social Welfare 37(4) 147-172.
Sorrel, C. (21 Aug. 2008). IPhone ships with image of factory worker who tested it. Wired, Online: http://www.wired.com/2008/08/iphone-ships-wi/
I | Attachment | Action | Size | Date | Who | Comment |
---|---|---|---|---|---|---|
JPG | Co_Like_Network_FB_Thumbnail.JPG | manage | 165 K | 14 Jul 2015 - 11:14 | Main.micania | |
JPG | Company_ImagePlot.JPG | manage | 88 K | 14 Jul 2015 - 11:09 | Main.micania | |
JPG | FactoryMap.JPG | manage | 90 K | 14 Jul 2015 - 10:11 | Main.micania | |
JPG | FocusAreas.JPG | manage | 81 K | 14 Jul 2015 - 10:12 | Main.micania | |
png | India_-_body_carried.png | manage | 129 K | 14 Jul 2015 - 12:41 | Main.micania | |
png | India_-_sexual_harassment_meeting.png | manage | 1 MB | 14 Jul 2015 - 12:41 | Main.micania | |
Like_Network_cleanclothes_on_facebook.pdf | manage | 102 K | 14 Jul 2015 - 11:12 | Main.micania | ||
jpg | NGO_image_collection.jpg | manage | 594 K | 14 Jul 2015 - 11:15 | Main.micania | |
png | image00.png | manage | 98 K | 14 Jul 2015 - 12:20 | Main.micania | |
png | image01.png | manage | 54 K | 14 Jul 2015 - 11:20 | Main.micania | |
png | image02.png | manage | 54 K | 14 Jul 2015 - 09:30 | Main.micania | |
png | image03.png | manage | 1 MB | 14 Jul 2015 - 09:34 | Main.micania | |
png | image04.png | manage | 49 K | 14 Jul 2015 - 12:02 | Main.micania | |
png | image05.png | manage | 357 K | 14 Jul 2015 - 12:08 | Main.micania | |
png | image06.png | manage | 17 K | 14 Jul 2015 - 09:31 | Main.micania | |
png | image07.png | manage | 31 K | 14 Jul 2015 - 09:50 | Main.micania | |
png | image08.png | manage | 47 K | 14 Jul 2015 - 09:32 | Main.micania | |
png | image09.png | manage | 914 K | 14 Jul 2015 - 09:32 | Main.micania | |
png | image10.png | manage | 109 K | 14 Jul 2015 - 12:20 | Main.micania | |
png | image11.png | manage | 57 K | 14 Jul 2015 - 10:09 | Main.micania | |
png | image12.png | manage | 94 K | 14 Jul 2015 - 12:11 | Main.micania | |
png | image13.png | manage | 459 K | 14 Jul 2015 - 08:54 | Main.micania | |
JPG | leaky.JPG | manage | 65 K | 14 Jul 2015 - 12:40 | Main.micania |