Strikes and lockouts: The need to separate labour conflicts

Abstract

Quantitative research on labour conflicts has offered innumerous insights into the workings of labour markets. With few exceptions this research is about conflicts, that is, strikes plus lockouts. The current situation is the result of practical difficulties separating strikes and lockouts, unwillingness by statistical bureaus to make the distinction, and the recommendation by leading scholars to refrain from distinguishing between the two types of conflicts. This article demonstrates that strikes and lockouts are not only theoretically distinct, their empirical manifestations differ. Employers and employees struggle over influence and division of income. Occasionally the two parties use, or threaten to use, their respective tools: the lockout and the strike. As long as the scientific community treats employers and employees as separate categories, we should also do the same with lockouts and strikes.

Introduction

Quantitative research on labour conflicts ¹ is almost exclusively based on statistics that amalgamate strikes and lockouts (Hamark, 2014; van der Velden, 2007). ² In other words, no distinction is made between tools used by two different social categories, namely employees and employers. This state of affairs is the combined result of practical difficulties separating strikes and lockouts, unwillingness by national statistical bureaus to make the distinction, and the argument made by leading scholars that researchers should not distinguish between the two types of conflicts.

Quantitatively oriented research on labour conflicts has a long history and spans several academic disciplines: economics, history, economic history, work science and sociology. It has provided insights on the importance of labour market institutions for conflict levels and the movement of conflicts over the business cycle. Certain occupations and branches are affected more often than others, and conflicts in the last couple of decades have been feminised and tertiarised. Major outbreaks of conflicts cluster around war and appear to move in long waves (Bordogna and Cella, 2002; Franzosi, 1995; Hansen, 1921; Kelly, 2012; Korpi and Shalev, 1979; Silver, 2003; Vandaele, 2016).

Yet, the amalgamation of strikes and lockouts into conflicts narrows the boundaries of research. Collapsing the two types of conflicts into a single category inflicts a bias toward strikes. Since lockouts are relatively few, they tend to drown in an ocean of strikes. Indeed, many quantitative studies on conflicts are presented as strike studies, by the authors themselves and others. Effectively, the practice of bundling strikes and lockouts together makes one of employers’ main weapons invisible.

This article demonstrates that strikes and lockouts are not only theoretically distinct, their empirical manifestations diverge regarding frequency, duration, involvement and volume. Furthermore, it seems like strikes and lockouts behave differently during the business cycle, and a micro-level perspective suggests that the two conflict types have dissimilar effects on the psychological well-being of the participants involved. It is shown that lockouts differ across countries with regard to magnitude, tactics and aims. Also, lockout activity in relation to total conflict activity is far from negligible. At times lockouts even account for the bulk of workers involved or conflict days.

The evidence or indications rest on an ongoing project on Scandinavian strikes and lockouts (discussed in the penultimate section) and a limited amount of published studies.

In this context, the objective of this article is three-fold. First, to explain the lack of quantitative research on strikes and lockouts proper – with a special focus on the claim that research should be about conflicts. Second, to present theoretical and empirical arguments in favour of studying strikes and lockouts separately. Third, to point out specific areas in Scandinavian history where disaggregated conflict statistics would advance research. The last objective is motivated by the fact that the article belongs to an empirical project on Scandinavian strikes and lockouts.

The remainder of the article is arranged as follows: The next section reviews quantitative studies that distinguish between strikes and lockouts, and the following then considers why quantitative research is focused on conflicts. The consecutive two sections contemplate arguments for separating strikes and lockouts, the latter section with special reference to the three Scandinavian countries before the Second World War. Finally, a conclusion ends the article.

Strikes and lockouts in earlier research

It would take too much space to give justice to the massive amount of quantitative studies devoted to labour conflicts. Besides, others have already done the job excellently (Franzosi, 1989; for more recent discussions about cross-country comparability, see Dribbusch and Vandaele, 2016; Lyddon, 2007).

As an illustration of the overall problem addressed in this article, consider the following trademark of the literature: classical, well-cited cross-country strike studies – such as Clark Kerr and Abraham Siegel (1954), Arthur Ross and Paul Hartman (1960), Edward Shorter and Charles Tilly (1974), and Walter Korpi and Michael Shalev (1979) – are in fact not strike studies, although presented so by the authors themselves and others (e.g. Franzosi, 1989). Based on statistics merging strikes and lockouts, they are about conflicts (Hamark, 2014).

What follows is a review of efforts to distinguish between strikes and lockouts. Inevitably, this means focusing on lockouts since the main, if not exclusive, motive of researchers to separate the two conflict types has been to accentuate the action of employers.

Michael LeRoy (1996) demonstrated that US replacement lockouts – during which operation continues with workers from the outside – increased in the 1980s, following Ronald Reagan’s handling of the 1981 PATCO strike, and lockout permissive court decisions. Since the Bureau of Labor Statistics records only conflicts in aggregation, LeRoy had to create his own database, based on documentation from the National Labor Relations Board. Like its US counterpart, the Australian Bureau of Statistics does not distinguish between strikes and lockouts. Nonetheless, by constructing a database of his own, Chris Briggs was able to show that the lockout, almost unheard of since the Great Depression, returned to Australia in the 1990s and was increasingly used around the turn of the millennium. Manufacturing employers were especially eager to make use of the resurrected weapon: in 1999–2003 one-quarter of total conflict volume in manufacturing was due to lockouts (Briggs, 2004). Institutional changes paved the way, notably a law from 1996 that allowed employers to use lockouts to coerce employees into signing individual and non-union agreements (Cooper et al., 2009: 344). Studying the period 1961–2001, KR Shyam Sundar (2004) showed how the relative composition of Indian lockout volume – commonly referred to as ‘shape’ – changed over time. In addition, the study presented descriptive statistics regarding the distribution of lockouts by sector (private or public), region and industrial branch, as well as employers’ motives for locking out – the most common being ‘indiscipline and violence’. Linda Briskin (2016), relying on Canadian official statistics 1976–2012, noted an increasing share of lockout activity over time. Using complementary qualitative sources Briskin also suggested a transformation from defensive to offensive lockouts; instead of merely responding to worker militancy, employers went proactive, often with an anti-union agenda. To the best of knowledge, lockouts proper have been studied quantitatively in only two countries for the period up to the Second World War: the Netherlands (van der Velden, 2000) and Sweden (Hamark, 2018), though Flemming Mikkelsen (1992) also presents lockout data for the three Scandinavian countries.

Keeping in mind that the studies above refer to countries where lockouts were legal during the time of investigation, they all reveal non-negligible lockout activity in quantitative terms. For instance, the share of lockouts of total conflict volume in India was 30% in the 1960s, climbing to 60% in the 1990s (Shyam Sundar, 2004: 4378). Another finding that holds true for several of the studies is the concentration of lockouts to certain regions and industries.

Methodologically, the studies have made a strong case for measuring lockout activity across several dimensions (since the ‘shape’ of conflicts varies in time and space).

In comparison to strikes, lockouts are fewer across all the studies. The finding is consistent with the inherent logic of the employment relationship, in which the employer usually can alter conditions of work without taking industrial action whereas employees cannot (Hyman, 1972: 23–24). The average lockout is, however, more voluminous than the average strike – technically because of the combined effect of longer duration and higher involvement (Canada is an exception regarding involvement). The longer duration is partially an effect of employers trying to weaken unions or deunionise entire workplaces (Cooper et al., 2009). Such battles tend to be drawn-out since labour has every reason to resist. The higher involvement is largely driven by multi-employer lockouts aimed at draining union funds (Hamark, 2018; see also Swenson, 2002).

Considering the limited amount of studies, it is understandable that cross-country comparisons are absent. As far as is known the only exception is Briggs’s (2005b) study on Australia and New Zealand, which due to limitations of official statistics could compare only lockout frequency (comparative conflict research has established that frequency is the indicator with the greatest reliability problems, see e.g. Lyddon, 2007: 28–29).

However, comparative research on the legality and institutional practices of strikes and lockouts does exist (Briggs, 2005a), to the benefit of future cross-country studies using disaggregated conflict statistics.

Briggs (2005a: 482–483) offered a literature review on lockout typologies. Typically, scholars have distinguished between different types of lockouts along three criteria: scope, purpose, and whether they are offensive or defensive.

A labour conflict constitutes a radical break with normal working life, and a few studies have investigated the effect of strikes on mental health. In a recent, innovative approach researchers compared the mental health status of strikers and locked out workers (Turner et al., 2020). The results suggested that strikers did better.

Why conflicts?

The disinterest in strike and lockout research proper is based partly on empirical obstacles, partly on what may be called political and philosophical considerations.

The most obvious reason for the merger is that researchers rely mainly on statistics from the UN agency International Labour Organization (ILO), in charge of collecting national statistics on labour conflicts. The ILO does not distinguish between strikes and lockouts, which reflects the lack of ability or willingness by member states to report other than aggregated conflict statistics (van der Velden, 2006: 342).

In practice it can be hard to unambiguously classify a conflict as either a strike or a lockout. For instance, imagine that an employer decides to dismiss unionised employees (arguably, a hindrance to ‘the normal work activities of employees’ ³ ), and that a group of employees reacts by putting down their tools. In such cases employers are likely to report a strike, employees a lockout. ⁴ Furthermore, as pointed out by Florence Peterson (1938: 3), ‘a strike might develop into a lock-out or vice versa’.

Some statistical bureaus indeed have seen practical problems. The United Kingdom abandoned distinguishing lockouts from strikes in 1895, as did the USA in 1921 (Lyddon, 2007: 25). ⁵ On the other hand, in the Scandinavian countries strikes and lockouts have always been separated in the official statistics (except for a few shorter periods). Even so, there remain practical difficulties in Scandinavia also.

Complicated, yes. Social science is riven by similar complications. For instance, unemployment is often decomposed into structural, cyclical and frictional unemployment. Whilst analytically useful it is still a challenge to establish even approximately how many people belong to a particular group. Yet few scholars demand the abolishment of decomposing.

Furthermore, available data suggest that practical difficulties to distinguish strikes and lockouts might be less than suggested. Defining the total number of lockouts as the sum of pure lockouts and so-called mixed conflicts (basically when a lockout follows a strike or vice versa), only one-third of total lockouts in the Netherlands, 1890–1940, were mixed conflicts (van der Velden, 2006: 347); about the same relation was found for India, 1966–1999 (Shyam Sundar, 2004: 4378). When comprehensive statistics on labour conflicts were compiled in Denmark and Sweden around 1900, the statistical bureaus made the unfortunate decision to include conflicts with unknown character in the mixed conflict category. Three decades of individual conflict reports issued by Statistics Denmark make it possible, however, to remove conflicts with unknown character. It turns out that mixed conflicts constituted 40% of total lockout frequency, 1897–1929. ⁶ Thus, in these three examples most lockouts were pure.

The third reason for the lack of strike and lockout studies proper is mainly non-empirical. It can be traced back to a piece by the late nineteenth century scholar Fred Hall (1898). ⁷ Hall’s text has both etymological and historiographical value. He showed, for instance, that the word ‘lockout’ originally was used exclusively to denote what we today would call sympathy lockout (Hall, 1898: 17–18). ⁸ Also, he demonstrated that the sympathy lockout antedated the sympathy strike (Hall, 1898: 16, 44). Crucial for the present context, Hall claimed that strikes and lockouts essentially were the same thing. The only difference was which side took the initiative. Hall thought the distinction was not merely pointless but also harmful – to labour. Most conflicts are strikes. And if this fact was to show up in official statistics, Hall believed workers would be blamed for industrial disorder (Hall, 1898: 24–26). ⁹

Later on, in 1926, the ILO followed in Hall’s footsteps, stating that the discrimination between strikes and lockouts is ‘based on the notion of responsibility, which is an inadequate basis for statistical definitions’ (International Labour Office, 1926: 13–14). Since then a number of scholars have repeated the basic argument regarding the uselessness or even the harmfulness of separating strikes and lockouts, albeit with varied nuances.

Peterson argued that the distinction between a strike and a lockout rests on ‘the notion of responsibility or moral obligation, . . . concepts which are impossible to define for statistical purposes’ (Peterson, 1938: 3). KGJC Knowles noted that strikes may be provoked by employers and lockouts by workers, and concluded (by quoting the ILO declaration from 1926) that ‘the question from which side the declaration is finally made is rather immaterial’ (Knowles, 1952: 299). The distinction between strikes and lockouts, he added, provides ‘no indication of where the “war guilt” lies’ (Knowles, 1952: 300). Paul Edwards reminded us that we cannot use conflict statistics to judge which side is to blame, concluding the distinction was of ‘little analytical significance’ (Edwards, 1981: 288). A more recent example of an IR scholar questioning the value of making a distinction between the two types of conflicts is Dave Lyddon (2007: 25), leaning against Hall and the ILO of 1926. ¹⁰

In summary: even if strikes and lockouts were always unambiguously distinguishable and disaggregated statistics available, we should study amalgamated conflicts nonetheless. Otherwise we step into a swamp of unscientific moralising. Which also, Hall claimed, is a detour likely to hurt labour.

The distinction between strikes and lockouts, Knowles asserted, provides no indication of where the ‘war guilt’ lies, and rightly so: strike and lockout data do not tell who’s to blame, neither in the nineteenth century nor today.

Knowles was also correct when pointing out that a lockout may be provoked by workers, a point which nicely illustrates that statistics have to be put in context. To understand the social relations at hand, we have to study actual concrete conflicts, as well as their messy backgrounds.

Yet, the who’s to blame argument by Knowles and other scholars is odd from a scientific standpoint.

Iraq invaded Iran in 1980. Is this fact irrelevant? Or worse, is it harmful – since someone may interpret the statement as if Iraq was the only part to be blamed for the terrible war fought between 1980 and 1988? An adequate account of the Iran–Iraq conflict may include the UK/US overthrow of Prime Minister Mosaddegh in 1953 and the subsequent installation of the Shah, the Iran revolution of 1979, the oil reserves of the Middle East, the sectarian split in Iraq and so on. Nothing in the statement ‘Iraq invaded Iran’ prohibits us from digging deeper into the causes of the war.

As mentioned, Peterson argued against the separation because it would have misguided (moralistic) implications. Thereafter, she immediately retreated from her principled position, claiming that:

Were strikes and lock-outs of approximately like frequency, the theoretical distinction between them might necessitate an attempt to distinguish them in tabulation despite all the difficulties involved. But because of the relatively strong position which the employer usually has in the bargaining relationship, he very seldom needs to resort to a lock-out. (Peterson, 1938: 4)

If anything, this is an argument in favour of making the distinction between strikes and lockouts. The quote says that (1) there is a theoretical difference, but (2) since lockouts are so few, the difference does not matter. While the logic of (2) is not entirely clear, indeed lockouts are few in relation to strikes for exactly the reason Peterson mentioned: the strong position of employers in bargaining relationships. But, as shown in the previous section, while lockouts are infrequent, their magnitude in terms of involvement or volume is far from marginal. Since there is no a priori reason to favour frequency over involvement or volume, Peterson’s pragmatism is uncalled for.

It is unclear which is the least constructive: to interpret the distinction between strikes and lockouts as who’s to blame, or to avoid making the distinction because of how others might interpret it.

The case for strikes and lockouts

Even though the ILO is still unable to present anything but amalgamated conflicts, the organisation has changed its mind since the 1920s and for a long time emphasised the importance of distinguishing between strikes and lockouts.

Empirical investigations, theoretical reasoning and the view of actors involved support the attitude of the ILO.

A recent Canadian study compared the effects of a strike and a lockout on union members’ psychological distress, six months after industrial action was resolved (Turner et al., 2020). Both conflicts involved teachers, occurred at the same time and lasted four weeks. The study found that locked out workers experienced less psychological well-being than strikers, also after controlling for age, gender and experience. The tentative explanation was that workers to some extent are directing during a strike, whereas control is virtually absent during a lockout. The result aligns with earlier studies on strikes and mental health, which have suggested that strikers highly involved in strike activity do better than those with low level of engagement. Hence, even if one accepts the view that the only difference between a strike and lockout is which side takes the initiative, the distinction is still useful since the initiative matters for the individuals involved.

Moving from micro to macro, the influence of the business cycle on strikes has been discussed for over a century (Ashenfelter and Jonson, 1969; Franzosi, 1995; Hansen, 1921; Rees, 1952; Rist, 1907). With tight labour markets and high capacity utilisation, employees’ chances of winning are better than with slack markets and low utilisation. Therefore, strikes are usually more frequent during economic expansions. ¹¹

‘Strike’ studies normally include lockouts. This fact does not, however, invalidate the conclusion regarding the cyclical behaviour of strikes. The literature commonly uses frequency of conflicts as independent variable and since the number of lockouts is generally small compared to strikes, from a practical viewpoint it does not matter too much if conflicts or strikes proper are studied. ¹²

Still, the merger of strikes and lockouts restricts the business cycle literature. In periods of economic contraction, employers possibly have a greater chance to fight off strikes and to launch successful lockouts. The lockouts preceding the 1909 Swedish general strike fit the picture, as does the 1931 Norwegian lockout (Hamark and Thörnqvist, 2013; Mikkelsen, 1992: 233–238; Schiller, 1967), although single events like these obviously cannot settle the matter. A quantitative study on Canadian conflicts concluded that duration was countercyclical, i.e. conflicts were longer in downturns (Harrison and Stewart, 1989). Since lockouts, as discussed above, on average are longer than strikes, the study implies that lockouts are countercyclical. Analysing Dutch data, Sjaak van der Velden (2006: 357–358) showed that the correlation between unemployment and frequency of strikes differed substantially from the one between unemployment and frequency of lockouts – indicating that the two types of conflicts behaved differently over the business cycle.

To sum up. There is some evidence to suggest that lockout activity is higher in recessions. While the results make theoretical sense, they are far from conclusive. More research is needed to support or refute the idea of countercyclical behaviour. It requires, however, that lockouts are not mixed with strikes.

Non-empirical reasoning suggests there are important asymmetries between strikes and lockouts (see e.g. Lundh, 2010: 128). Generally, a strike aims to hurt the employer financially. Ideally only a few people in key positions are called to strike, as a cost-efficient way of causing losses to the firm. The lockout on the other hand is most effective large-scale (at least when employers try to alter collective agreements rather than abolish them). By locking out union members en masse, ideally employers achieve a quick bloodletting of union strike funds, thereby forcing a favourable deal at the bargaining table. The largest (in relative terms) European labour conflict in the nineteenth century followed this line. In 1899 the Danish Employers’ Association answered the unions’ decentralised strike strategy (omgangskruen) with massive lockouts – leading to the famous September Agreement (Schiller, 1967: 11–12).

Another asymmetry stems from the fact that in the sphere of production, managerial prerogative prevails (with varying degrees of retrenchment). Employees wishing to change current work conditions have little formal influence. Going on strike is one option to put pressure on the employer. By contrast, employers can lean on managerial prerogatives and seldom need to resort to industrial action (Peterson, 1938: 4).

Strikes have gendered dimensions (Thörnqvist and Fransson, 2015). It seems plausible that lockouts do as well. Before the Second World War men worked in industries with much higher union density than women. This points in the direction that men were targeted by lockouts more often than women (also when taking into account that the majority of wage earners were men). On the other hand, men were seen as breadwinners. Given this ideological notion it was probably conceived as a less damaging act – and hence easier – to lock out women. It indicates that lockouts were disproportionally directed towards female-dominated industries. Theoretically, then, it is not clear what the relationship between lockouts and gender looks like; it is a matter of empirical investigation.

Strikes are used by employees, lockouts are launched by employers. Whatever the empirical outcome of conflicts, on a theoretical level this simple fact is enough to distinguish between them. The major actors involved in conflicts – the employees, the employers and the state – value the distinction.

An overwhelming majority of OECD countries either outlaws lockouts or puts restrictions on them – stricter than on strikes. The right to strike is seen as necessary to counterbalance the bigger resources of employers. If, however, lockouts could be used freely, they would take the sting out of strikes and cement an unequal distribution of power between labour and capital (Briggs, 2005a: 487–489). In this context it is irrelevant whether or not these states make accurate assessments of the power balance, or if restraining or outlawing lockouts is a clever policy (indeed, a few OECD countries are instead guided by the principle of ‘parity of arms’). The point is that lawmakers and courts appreciate the difference between strikes and lockouts.

In the Scandinavian countries the right to take sympathy or secondary action is extensive (Fahlbeck, 2006). The right applies to both employers and employees. As past and current debates on secondary action reveal: the question at heart is balance of power. Since the 1990s, Swedish employers’ associations have been arguing that secondary action taken by unions regularly violates the principle of proportionality (Almega, 2016; Svenskt Näringsliv, 2016; see also Börjesson and Gustafsson, 2016). To stop misuse the labour law should be sharpened. The resistance to secondary action is not based on moral principles but on the calculation that nowadays the unions benefit more. Times are changing. At the beginning of the last century the chairman of the Swedish Employers’ Association (SAF), Hjalmar von Sydow, explained that without the legal right to take secondary action, the employer associations might as well cease to exist (Casparsson, 1951: 230). At that time massive sympathy lockouts were used by employers to enforce managerial prerogatives.

The way forward when looking backwards: Disaggregated statistics in pre-Second World War Scandinavia

This article belongs to an empirical project called Database on Scandinavian strikes and lockouts, c.1900–1938. A forthcoming paper (Hamark, 2022) will present aggregated year-by-year data on strikes and lockouts respectively and for each country – Denmark, Norway and Sweden – in which strike and lockout activity is captured using four different measurements: frequency, duration, involvement and volume. In addition, a database of individual lockouts in each country includes variables such as the number of employees involved, start and end dates, location and whether or not the parties belong to a union/association. The database can be used to study lockout activity over the business cycle for which yearly data are not appropriate, and to investigate lockout activity according to industrial branch or even sub-branch. ¹³

The following discussion is connected with Scandinavia but implications are broader in scope. Focus is on lockouts, the type of conflict often made invisible by the amalgamation.

In 1938 the Swedish Saltsjöbaden Agreement was signed. Even though the content of the agreement was of limited substance, it was the first main accord between the Swedish Trade Union Confederation (LO) and the SAF, and it has been heralded as paving the way for decades of labour market peace, of a consensus culture and of the Swedish model. Unsurprisingly, researchers have put a lot of effort in investigating the background of the agreement. Considered from a power-relational perspective, was it negotiated based on the strength of workers or employers?

Sociologists Walter Korpi and Michael Shalev (1979, 1980; see also Korpi, 1983, 2006) represent the internationally dominant view in stressing the relative strength of labour. According to them, there was a sharp drop in conflicts in the mid-1930s; labour market peace thus preceded Saltsjöbaden. The lower level of conflicts was the combined result of two factors: the seizure of governmental power by the social democrats in the 1930s and high union density. The so-called power resources hypothesis states that stable left-wing governments and strong unions shift the focus of class struggle from the sphere of production to the political arena. The shift is beneficial to workers since capital’s inherent advantage in production is countered by democratic power (Korpi and Shalev, 1979: 179–184; 1980: 320–321).

Union density, of course, is conventionally assumed to correlate with labour strength. Indeed, no other country could match Sweden’s unionisation growth in the 1920s, and in the mid-1930s only Denmark was on par with the Swedish participation rate (Kjellberg, 1983: 220, 272ff.).

Yet Korpi and Shalev’s interpretation of causality has been challenged by political scientist Peter Swenson (2002, 2009) on precisely this point. In the 1920s, when Swedish employers recurrently used the lockout – also targeting non-organised labour – employees ran to the unions for cover. Since statutes obliged unions to pay locked out members, union membership was the equivalent of buying a lockout insurance. In Swenson’s interpretation, therefore, the increasing union density was a token of employer strength. But, as Swenson (2002: 340) himself acknowledges, only comparative research could ‘substantiate or refute this argument’. Such comparative research should preferably be conducted along two dimensions, cross-industry and cross-country, allowing two hypotheses to be tested: (1) union density increased most in industries with high lockout activity, and (2) union density increased most in countries with high lockout activity.

Swenson’s Capitalists against Markets (2002) is a qualitative comparison between US and Swedish industrial relations systems. Swedish and American employers had a choice when facing growing labour organisation and militancy in the form of strikes: to replace strikers with other people, or to enlarge conflicts using massive lockouts. And employers chose differently, Swenson (2002: 73) tells us: ‘Lockouts were to Swedish developments what strikebreaking and blacklisting was to American ones.’

But without quantitative data on lockouts – and in this particular case also strikebreaking – conclusions could be tentative at best. In the Swedish–US comparison Swenson is probably correct since various studies show the extensive use of strikebreakers, spies, armed guards and paramilitary groups in the US labour market (Gitelman, 1973; Norwood, 2002; Smith, 2003; Tuttle, 1969). Yet it would be comforting even in the Swedish–US case to have also quantitative, comparative data.

A yet unpublished database on Scandinavian strikes and lockouts, 1900–1938, produced by the present author, has been merged with the corresponding series for the Netherlands (van der Velden, 2016). A few results are worth mentioning, all of which take into account demographical differences. ¹⁴ First, Norway and Sweden had relatively high involvement in conflicts, Denmark and the Netherlands relatively low. This is in line with earlier research (cf. Korpi and Shalev, 1979; Shorter and Tilly, 1974). Second, in Denmark in the 1920s an astonishing 65% of all workers involved in conflicts were locked out. And despite being a low-conflict country, during the four decades Denmark had the highest rate of locked out workers of all four countries (and by inference very few strikers). The ratio of locked out workers in Denmark to the Netherlands was substantial, 6:1.

Apparently employer strategy differed widely between the Netherlands and Denmark. The choice of strategy is however obscured if we study only amalgamated conflicts.

Even if Dutch employers usually refrained from using the lockout, there were exceptions – and one in particular. Out of all lockout activity in the Netherlands during 160 years, 1840–2000, four years at the beginning of the 1920s accounted for 40% (van der Velden, 2006: 352–353). Also, the Dutch lockouts were unequally distributed across industries, and massively so. From the 1890s until the Second World War the textile industry and in particular the textile region Twente was dominating to such an extent that the procedure of locking out was called the Twente system (van der Velden, 2006: 346). Again, without statistics that disaggregate conflicts we have small possibility to discern events like these.

How could we explain the different choices made by employers in the USA and Sweden, or by employers in Denmark and the Netherlands? Swenson’s (2002) US–Swedish comparison explicitly or implicitly generated a handful of valuable hypotheses, such as: lockout activity was higher where (1) the supply of labour was relatively scarce and (2) the state tended to be employer non-accommodating. Strikebreaking presupposes a reserve army of labour and state support in physically protecting strikebreakers or removing picket lines. Sympathy lockout activity was higher where (3) employer associations were strong in relation to individual companies and (4) employers faced unions with centralised strike funds. Sympathy lockouts rely on powerful employer associations and that companies act in the (perceived) general interest of employers, either willingly or by force. Moreover, sympathy lockouts are a means to drain union funds. For lockouts to be effective, employers therefore need a counterpart who controls a centralised fund and who has authority to put pressure on its local branches.

Needless to say, none of the hypotheses can be tested without lockout data.

Equally interesting is the impact of employer choice. Apart from the argument discussed above that lockouts made unions ‘strong’, there is the idea that lockouts fostered centralisation of the entire industrial relations system (Swenson, 2002). If every small strike potentially can be met by a massive lockout, union leadership must discipline its members; without discipline union funds would rapidly be depleted. This puts pressure to centralise the right to call a strike (for instance the abolishment of union ballots), as well as to increase the national board’s authority over local branches. In turn, union centralisation gives impetus to further centralisation of collective bargaining and conflict resolving institutions.

The argument remains to be weighed against comparative lockout statistics. Not only historical accuracy is at stake. It touches upon a theoretical issue: the optimal level at which the interests of employers and employees should be negotiated. As argued by Frans Traxler (1995), employers are superior in individual exchange relations. Consequently employees seek to avoid individual exchange relations by means of unionisation. In more general terms: employers benefit from bargaining at a level as decentralised as possible, whereas employees prefer centralised bargaining. Yet, paradoxically, during our period of interest employers in some countries and industries choose centralised bargaining (Crouch, 1993). A part of the solution to this contradiction might be employers selecting or rejecting the sympathy lockout. John Bowman has suggested that employers embracing the lockout explains the rise of centralised wage bargaining in Norway. ‘Through the use of the lockout, employers could define the scope of labor conflict and, in turn, the scope of the eventual settlement’ (Bowman, 2002: 1002). Centralisation served employers strategically, amongst other things by curbing whipsawing.

Any lockout that extends beyond the individual workplace or company requires coordination. This raises the issue of collective action in industrial relations (Bowman, 1998; Kelly, 2012; Offe and Wiesenthal, 1980). Imagine an industry-wide lockout targeting high wages. If the lockout is successful, all employers in the industry benefit. At the same time, individual companies have strong incentives not to cooperate but rather to take a free ride, by continuing to produce and let the other companies carry the burden. Free riders, ‘lockout breakers’ in this case, jeopardise the collective effort of employers. Under what circumstances were employers and their associations able to overcome the problem of free riding? In Sweden lockouts were generally indiscriminate: all workers within the affected branch were targeted, including ‘innocent’, non-unionised workers. The reason spells solidarity. It was considered unfair that firms with non-unionised labour should be allowed to continue production; all employers should carry the burden of the lockout (Swenson, 2002: 75–76). Interestingly, in Denmark lockouts mainly seem to have targeted unionised workers only (Jensen and Olsen, 1901: 120; Statistiske Meddelelser, 1926: 20–21). Is Denmark an example of a country where free riding existed but did not constitute a problem? Only future research could tell.

Finally, the aforementioned ‘strike’ studies by Ross and Hartman and other scholars shape our collective understanding of history. For instance, historian Klas Åmark (1989: 62) concludes that Sweden was the ‘land of strikes’ (strejklandet framför andra) from the 1890s to the mid-1930s (see also Johansson, 2011; Korpi, 2006: 188; Moene and Wallerstein, 2006: 150). But such an inference can be validly drawn from empirical studies on ‘strikes’ only if lockout activity is negligible – which it was not. A future cross-country study that distinguishes between employers and employees may show that Sweden was not the land of strikes: during the period Åmark refers to, lockouts and mixed conflicts accounted for nearly half of total conflict volume. ¹⁵

Conclusion

Quantitative studies on labour conflicts are in most cases precisely that: studies on conflicts, that is, an amalgamation of strikes and lockouts. Strikes and lockouts proper are still under-researched.

A salient reason is that the vast majority of member states reports only conflicts to the ILO. In order to advance research scholars have to look elsewhere. Occasionally they can elaborate on existing national statistics (Briskin, 2016; Hamark, 2022), at times they have to build databases of their own by scanning newspapers, journals, legal documents, etc., supplemented by surveys and interviews (Briggs, 2004; LeRoy, 1996; van der Velden, 2000). Importantly, with improved software reading digitised material, the process of collecting and structuring microdata is facilitated (van der Velden, 2017: 12).

In practice it can be hard to unambiguously classify a conflict as either a strike or a lockout. It might be, as Briskin (2016: 198) suggests, that the ability to distinguish between strikes and lockouts has improved recently, alongside employers’ increased usage of offensive lockouts (rather than defensive lockouts in response to strikes). Still, the problem would not go away.

No quick fix exists but van der Velden (2007: 15) ought to be listened to: while admitting ‘practical difficulties in distinguishing some lockouts and strikes’, he nevertheless concludes that ‘rescuing the lockout from aggregate figures of industrial disputes is still an important scientific duty’.

For well over a century several distinguished scholars have argued we should not bother to discriminate between strikes and lockouts; then we risk ending up in a moralistic who’s to blame-debate which is meaningless at best, harmful at worst.

True, defining a particular conflict as a strike (lockout) does not establish that employees (employers) are the ‘accountable’ side. Certainly, figures on strikes and lockouts may be interpreted in such terms. But the fear of statistics being misused is a poor argument against the separation of conflicts. After all, statistics could always be misused.

The arguments in favour of studying strikes and lockouts separately range from theoretical (employers and employees have different interests; managerial prerogative prevails at work; strikes and lockouts are asymmetrical; strikes have gender dimensions, lockouts probably do too), over ‘practical’ (lawmakers, employers and unions appreciate the difference between strikes and lockouts) to empirical (strikes and lockouts have different ‘shapes’, appear to behave differently over the business cycle, and seem to have different effects on the psychological well-being of workers involved; lockout activity varies across countries and industries and over time, often without close correlation to overall conflict activity).

Commenting on the findings of LeRoy, Briskin (2016: 195–196) concludes that ‘the main weapon of [US] employers in the 1990s [was] masked by the fact that the US government collects no specific data on lockouts’. The conclusion does not contradict that the study of strikes would benefit from separation, too. While aggregated statistics tend to make only lockouts invisible, numerous ‘strike’ studies are nonetheless contaminated by the presence of lockouts. Thus, a careful distinction between the tools of employees and employers has the potential of revitalising the study of strikes.

Going outside the boundaries of ILO-reported conflicts is advantageous also for other reasons. Cross-country comparability is often limited by partiality and inconsistency of national statistics. For instance, the German statistical bureau does not provide the frequency of conflicts, the US bureau reports only conflicts with a minimum of 1000 workers involved, and several national bureaus do not report conflicts at all (Dribbusch and Vandaele, 2016; van der Velden, 2017). In the compilation of worldwide statistics and the creation of new datasets, the Global Hub Labour Conflicts at the International Institute of Social History in Amsterdam stands out (van der Velden, 2017).

Unpublished Scandinavian lockout statistics reveal that employer militancy, in terms of involvement and volume, contributed substantially to overall militancy before the Second World War. As a case in point, in Denmark in the 1920s 65% of all workers involved in conflicts were locked out.

There was clear within-Scandinavia variation in lockout activity. Yet including also the Netherlands in the comparison (the only country outside Scandinavia for which data are available) makes it even more evident that employers choose differently how to combat wage increases or to defend managerial prerogatives.

The Swedish labour market went through a remarkable transition in the first half of the last century: from wide-ranging militancy to quiescence. The most influential interpretation of the transition is the power resources hypothesis, stating that labour – through strong unions and control over the executive – was able to shift the focus of class struggle from the sphere of production to the political arena (Korpi and Shalev, 1979). Swenson’s (2002) heretical position is that massive lockouts forced workers into unions. Hence, unions only appeared strong.

Cross-country and cross-industry statistics on lockouts and union density rates have the potential of shedding new light on a discussion of importance not only for the interpretation of Sweden at a crossroads. As Korpi and Shalev (1979: 171) themselves stated, the power resources hypothesis as such ‘is largely based upon an interpretation of the long-term evolution of industrial conflict in Sweden’.

What next?

Lockout typologies already exist, and further research is hardly hampered by lack of theories or hypotheses regarding lockouts, at least not in the short run. By and large, the missing piece is disaggregated conflict statistics. Such statistics would, as Briggs (2004: 101–102) notes, ‘lay the foundations for subsequent analysis of the causal forces behind . . . lockouts and the factors which shape their use by employers’.

At present there are a limited number of quantitative studies covering single countries, but still no published cross-country comparison of lockouts based on the most solid measurements, namely involvement and volume. While the few existing studies have offered several valuable insights, comparative research is better suited to test hypotheses such as ‘lockout activity is positively associated with scarce supply of labour’ and ‘sympathy lockout activity is positively associated with strong employer associations’. Testing these and other propositions would enhance the understanding of why employer strategies differ across countries.

Yearly data are not always sufficient. For instance, to reach conclusive evidence on the relationship between lockouts and the business cycle, monthly data are needed (Rees, 1952: 372). Correspondingly, national averages are good enough for many purposes but hide the clustering of lockouts to certain industrial branches (see e.g. Shyam Sundar, 2004; van der Velden, 2006). Furthermore, quantitative studies should acknowledge that lockouts do not constitute a monolith. In the context of modern-day Australia, Rae Cooper and colleagues (2009: 344–346) distinguished between ‘big bang lockouts’ and ‘bargaining lockouts’. The former are prolonged conflicts, often with the aim of forcing workers to sign non-union agreements, while the latter are typically short and occur in the context of collective bargaining. Cooper et al. further suggested that the lockout choice by firms was affected by market structure characteristics, such as the degree of global competition.

Combining qualitative and quantitative analysis seems like a promising point of departure. What types of lockouts are there, how common (or long or large) are they and how to account for their variation? The questions are relevant not only in the context of single countries. Before the Second World War, Scandinavia probably had the highest lockout activity in the world. The lockouts were often huge, aiming to secure managerial control but not to crush the unions. On the contrary, their success depended on the presence of a reasonably strong union confederation with centralised strike funds. Lockouts still occur in Scandinavia but the magnitude does not resemble the interwar period; the power repertoire of employers worldwide has broadened, not least through increased capital mobility. While the lockout has not vanished, neither globally nor in Scandinavia, it seems to have transformed. If the old lockout in Scandinavia was ‘orderly’, within the realms of collective bargaining, the modern lockout is often anti-union, as evidence from Australia, New Zealand, Canada and the USA suggests.

Sympathy lockouts as a collective action problem are insufficiently researched, despite the burgeoning literature on collective action following Mancur Olson (1965).

The difference between Danish and Swedish lockouts with respect to non-unionised workers was pointed out above, but the sympathy lockout constitutes a challenge for future research also at a more fundamental level. In one of few attempts to analyse lockouts as a collective action problem, Bowman (1998) argued that in the labour market (though not in the product market) capitalists basically face the same challenges of cooperation as workers (cf. Offe and Wiesenthal, 1980). Bowman, examining Norway in the twentieth century, showed how employers were reluctant to participate in sympathy lockouts, especially ‘in sectors with high fixed costs, in sectors experiencing strong demand, and in sectors in which low levels of employer organization expose cooperators to competitive incursions by free-riding competitors’ (Bowman, 1998: 322). Furthermore, since sympathy lockouts are built on the idea of reciprocity, in some cases even companies that benefited from a particular lockout were reluctant, knowing they had to pay back at a later stage.

While Bowman clearly pointed out the impediments, a troubling fact remains: however serious the obstacles to Norwegian capitalist collective action were, massive sympathy lockouts did occur (Mikkelsen, 1992: 214–237) and their occurrence needs to be explained.

It requires amongst other things statistics that discriminate between single employer lockouts and multi-employer, sympathy lockouts.

There are several reasons for distinguishing between lockouts and strikes. The most prosaic is arguably also the most essential: the two types of conflicts are tools in the hands of people with partially opposing interests. Employers and employees struggle over influence and division of income. Occasionally the two parties use, or threaten to use, their respective tools, the lockout and the strike. As long as the scientific community treats employers and employees as separate categories, we should also do the same with lockouts and strikes.