Dataset creation
Part of my research was a mapping exercise that was dedicated to capturing non-industrial strikes around the globe in the conjuncture of crisis. I have discussed the results and their implications for my research question in Chapter 4. In this Appendix, I explain how I proceeded when I conducted my actual research, which was influenced methodologically by an approach to mapping used by colleagues in a project that was about transitions to renewable energy in 34 countries in Africa (Müller et al, 2020). This section is dedicated to how I created my dataset; in the next one, I discuss how I coded the data. Since my approach to data analysis is contained in Chapter 4, I do not repeat the different steps in detail here.
Inspired by Silver’s Forces of Labor (2003), my starting point was a search for the coverage of non-industrial strikes in online archives of well-known quality papers, that is, The Mail & Guardian (Johannesburg), The Times of India (Mumbai), The New York Times, Página12 (Buenos Aires), The Guardian (London), Die Tageszeitung (taz, Berlin) and Público (Madrid). I chose those papers to ensure that there was geographical balance; that they were written in languages that I understood; that their archives were accessible online; that they had reputation for good journalism with a sound factual basis; and that they were interested in labour issues.
I conducted online searches for each paper. The search terms were ‘public sector’ or ‘service sector’ and ‘strike’ (or the respective terms in German and Spanish).1 They ensured that I was able to pick up a wide range of stoppages, but the number of hits was always significantly higher for ‘public sector’ and
With the help of the articles found through my online searches, I created a simple database in the form of a spreadsheet. It contained detailed information on 244 strikes in the public and service sectors from January 2007 to October 2020 from 33 countries and autonomous territories. To order the information at this initial stage, I used a range of simple, descriptive categories: country; type of country (home country, neighbouring country or other country in relation to the place of publication of the newspapers selected); city or region; sector; duration; date; number of participants; participating unions; as well as reasons, demands and outcomes.
Inspired by the ‘snowball’ approach often used for selecting interviewees (see Auerbach and Silverstein, 2003: 18; Ritchie et al, 2003: 94), I also used the articles found through online searches to click on linked additional coverage from the same newspaper. This helped fill in my categories and check the accuracy of the data that I included. In several cases, I could not fill the ‘duration’, ‘numbers of participants’ and ‘unions’ categories; if so, I left them blank. I excluded opinion pieces because they are, by definition, not about presenting factual information but making a political point. Furthermore, I left out most articles that referred to strike announcements because stoppages are often called off at the last minute. I only included them if the strike in question was to be held on the same day, and if the newspaper in question contained other articles that allowed me to verify that it indeed had happened, or if I found such articles through an online search. In other words, I complemented the dataset through targeted searches. Finally, I also attempted to draw different manifestations of the same strike together where coverage focused on local actions, for example, picketing or demonstrations. Importantly, some disputes are long-winded and sometimes run over several years. Consequently, I decided to list strikes individually that formed part of one and the same dispute but were clearly separated in time. Moreover, I included general strikes if there was proof that service or public sector workers were involved. As it turned out, publicly employed workers often played a leading role in them.
Due to the fact that various G20 countries were missing in the initial dataset, I expanded the sample by also using two online platforms for my searches, labourstart.org and labournet.de. Furthermore, I added news media coverage of strikes that I found through researching my three cases covered in Part II. This way, I created my final sample of 387 strikes, which were located, all in all, in 56 countries and autonomous territories (see Figure 4.1).
Coding
As my research is based on strong theoretical assumptions, I started the coding process with a deductive move. I identified a range of provisional categories – and mutually exclusive codes falling under those categories – before I started analysing my first sample. I derived the first category from my research question and complemented this with four categories that came out of earlier work in the field of strike research that my colleague Jörg Nowak and I had carried out over the past decade. In this work, we identified a number of key characteristics of mass strikes under capitalist conditions (Nowak and Gallas, 2014: 311–12; Gallas, 2018: 239–40; 2020: 184; Nowak, 2019: 49–50).
I used the categories and codes in question both to record and arrange the information in the articles that I had collected for my first sample. My aim in this first round of coding was to check whether the codes were applicable in the sense that they could indeed be used to classify the existing information. My guiding assumption was that codes worked well if there were no (or only a few) doubtful cases that were hard to categorize.
The first category, ‘class effects’, directly spoke to my research problem. It referred to the question of whether the unifying effects of a strike in class terms outweigh its divisive consequences, or whether the contrary is the case. Correspondingly, my codes were ‘exclusive solidarity’ and ‘inclusive solidarity’, with the former referring to strikes whose divisive effects outweigh their unifying consequences, and the latter to stoppages where the opposite applies. I imagined that there would be two scenarios of how I would identify class effects: for one, through establishing how workers inside or outside the constituency of the strike in question saw it,
In line with the literature, I also employed four other categories and the corresponding codes that were less clearly connected with the research problem. My expectation was that systematizing strikes this way would allow me to find patterns that spoke to my research question in an indirect fashion, or that it would at least generate observations that enhanced my understanding of the role of strikes in contemporary capitalism.
Accordingly, my second category was ‘mode of confrontation’, which was about contextualizing the demands of the workers and deciding whether they attempted to fend off worsening conditions of work (broadly understood) and threats to existing arrangements, or whether they fought for improvements. Accordingly, the two codes I used were ‘offensive’ and ‘defensive’. These codes proved to be straightforward in their application. There was only one strike in my sample, the general strike in India in February 2012, which did not quite fit because there was a mix of offensive and defensive demands: The striking workers campaigned for an increase in the minimum wage and efforts to contain inflation, which can be seen as offensive demands, as well an end to privatization plans of the government, which is a defensive demand (Krishnan, 2012; Shah Singh, 2012; TOI, 2012). Consequently, I created a third code (‘offensive and defensive’) to leave room for such contradictory scenarios, but left the basic distinction intact because it appeared to be working in terms of classifying the existing information.
The third category was ‘character of aims’, which was referring primarily to the constituency of a strike and the instance addressed through its demands. The codes were ‘mostly economic’, ‘economic and politicized’,
The fourth category was ‘mode of intervention’ and referred to the question of whether a strike was ‘symbolic’ – the first code – in the sense that it was about demonstrating the potential power of workers through creating limited disruption, or whether it was about ‘enforcement’ – the second code – that is, about stopping work until the instance addressed in the demands chooses to negotiate or settle. Whereas the general strikes mentioned are typical examples of symbolic strikes because they usually end after one or two days, the latter can be quite drawn-out affairs. What is noteworthy in this context is that the disruption created through symbolic strikes varies considerably. In the cases where there were recurring (and sometimes rolling) strikes, their effects were felt in the everyday lives of people fairly frequently over a certain time span. If they were a one-off event, this was of course not the case. But since even short symbolic mass strikes had quite disruptive effects for a day or two and discernible consequences for the economy and everyday routines, it did not make sense to further break down symbolic strikes.
The fifth category, geographical extension, referred to the question of whether strikes were ‘local’, ‘regional’, ‘national’ or ‘transnational’. It is noteworthy in this context that there are indeed some transnational strikes. There was a general strike against austerity in Portugal and Spain in November 2012, plus various strikes of people employed with online retailer Amazon as well as stoppages of pilots and flight attendants working for the Irish budget airline Ryanair. The application of the codes for this category was unproblematic because the articles usually contained the information in question.
Finally, I added a category called ‘type of country’. This was not directly or indirectly connected with the research question; the aim was to gauge the breadth of the sample. It referred to the countries where strikes took place. The codes ‘home country’, ‘neighbouring country’ and ‘other country’ highlighted where the nearest of the seven newspapers discussed was located in relation to the country in question. These codes proved easy to apply; the numbers on the geographical spread of items in the sample were produced with their help.
The five categories emerged out of my theorization of class relations and strikes. But sticking exclusively to a deductive approach creates the risk of falling into the trap of subsumption. I drew inspiration from an approach to qualitative empirical research developed by Claes Axel Belfrage and Felix Hauf called ‘Critical Grounded Theory’ (Belfrage and Hauf, 2015). This proved to be fitting insofar as it was based on critical realist ontological assumptions, and I found the way in which this approach connects conceptualization with empirical analysis in the research process particularly relevant. According to
The first new category was ‘economic extension’, with the codes ‘single branch’ and ‘multiple branches’ capturing the issue of whether strikes remained within the confines of a specific branch of the economy or not. It is important to note that not just the private sector comprises multiple branches but also the public sector: the work of teachers is fundamentally different from that of civil servants or sanitation workers. My hunch was that information on the economic extension of strikes could be relevant for my research question because strikes involving multiple branches bring together different groups of workers, which could, potentially, be seen as an expansive move. The codes proved straightforward in their application. It was possible to apply them to every single strike in the sample.
The second category referred to the ‘ownership of means of production’ in order to get an understanding of whether the strikes concerned publicly or privately owned entities. This was of relevance for my research question because, first, class relations in the public sector are somewhat different from those in the private sector – the employer is a state body, and not a direct representative of capital – and, second, because it was also important to leave room for different strike dynamics playing out in the different sectors. Accordingly, there were three codes: ‘public’, ‘private’ and ‘public and private’. I introduced the latter code to do justice to the fact that broad mobilizations in particular, for example general strikes, cross the public/private divide.
And finally, I added an open category called ‘branch of the economy’, which allowed me to classify strikes according to where in the topography of the economy they are located. By going through the sample, I created 17 codes including ‘sanitation’, ‘media’, ‘postal services’ and ‘healthcare’, and complemented them with three codes for strikes involving multiple branches (‘public’, ‘private’, ‘general’).
To enhance the reliability of my codes, I started coding from a ‘clean’ version of my second sample and left a few weeks between the two coding exercises. Operating this way, I checked the reliability of my coding.
My last step, an inductive move, was to identify different strike patterns, again with the help of three rounds of coding. I found five patterns – the collective bargaining strike, the extension strike, the expansive-politicized strike, the class-based strike, the strike for political aims – and then related them to the ‘character of aims’ category (see Figure 4.2). This allowed me to zoom in on the class question, as is explained in Chapter 4.
The German terms were ‘Öffentlicher Dienst’ (which is more common than ‘öffentlicher Sektor’), ‘Dienstleistungssektor’ and ‘Streik’; the Spanish words were ‘sector público’, ‘sector servicios’ and ‘huelga’. I decided against using ‘paro’, which is synonymous with ‘huelga’ but also means ‘unemployment’ and thus creates a lot of false hits.
The numbers do not add up to 387, the overall figure for items in the sample, because it contains five transnational strikes, all of which count as strikes in several countries.
Besides, there are doubts over whether the image of South Africa as being a country with a very high number of strikes is justified. With reference to ILO data, Haroon Bhorat and David Tseng (2014) contest it. They argue that strike incidence is not vastly different if the country is compared to other middle- and high-income countries. But even so, strike incidence should not necessarily be less than a fourth of those in Germany, which is why it is plausible that media bias also plays a role.
By definition, general strikes aim to mobilize workers both inside and outside manufacturing. They have been included in the sample if there was significant participation of non-industrial workers so that they could not be seen as industrial strikes.