The Global Diffusion of Work-Injury Insurance: The Role of Spatial Networks and Nation Building

Abstract

Work-injury law often marks the beginning of the modern welfare state and as we argue this type of insurance is particularly important to nation-state building: it placates myriad social groups needs and demands while binding them symbiotically with the state. We analyze any first laws and the first instance of social insurance as outcomes in 151 states (1880–2010). Our network diffusion and event history models reveal that spatial proximity and democratization are key predictors of first laws. However, nation-state formation and trade density are additional predictors of social insurance in particular. We conclude with implications of these findings for understanding nation-building within global networks.

Introduction

This chapter is a product of the research conducted in the Collaborative Research Center “Global Dynamics of Social Policy” at the University of Bremen. The center is funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation)—project number 374666841—SFB 1342.

Work-injury laws, also historically known as workmen’s compensationand accident insurance, are among the oldest welfare state laws and often marked by scholars as the beginning of the welfare state (Abbott and DeViney 1992; Flora and Alber 2009). They extended the basic principles of responsibility for damages done to persons or property found in civil and common law traditions. Their implementation coincided with new, dangerous forms of employment in factories, mechanized extraction, and construction that came with industrialization. The workers, families, and communities affected by these new industrial risks demanded protection, putting pressure on policymakers (Wilensky 1975; Kangas 2010).

Usually, the first work-injury law introduced in a state was employer liability. Although these laws often standardized compensation levels and explicitly covered work-related damages, they did little to reduce worker subversions and revolts. They mostly did not insure against risk because the burden of proof of harm fell on the worker and compensations were small and limited in duration (Flora and Alber 2009). More advanced and effective laws involved the creation of social insurance. These placed all employees into an insurance scheme, with automatic entitlements to compensation unless otherwise proven by the employer or insurance agency in court. Social insurance offered financial predictability, less workplace violence, and protection of both employers and employees from legal battles (Tripp 1976; Berkowitz and Berkowitz 1984; Pavalko 1989).

Employer liability laws tended to pave the way for social insurance laws. However, history suggests no standardized trajectory (Haggard and Kaufman 2008). Types of work-injury laws are only weakly correlated with industrialization and democratization, leaving functional theories limited in their capacity to explain cross-country patterns of welfare state development (Alber 1982). For example, Britain was the earliest industrializing society and quite democratic, but a latecomer to social insurance in 1934. Germany was less industrialized and less democratic, but a pioneer in 1884. Meanwhile, France was similarly developed as Germany and slightly more democratic, but only introduced social insurance 10 years later in 1894 (Breznau and Lanver 2020).

Factors of political economy are likely responsible for such variation. Work-injury social insurance, and the general implementation of social insurance across policy domains, is an important strategy in political rule. Social protection laws shift the burden of individual risk to the collective level, meaning that workers and rulers become bound together with a common welfare interest. The extent of these provisions often depends on the success of left/labor parties and special interests. In general, this strategy is an adaptation of military conscription. Dating back to the Greeks and Romans, ruling powers used the promise of veteran benefits to motivate and compensate soldiers engaged in potentially deadly work to defend the territory and its people (Ierley 1984). Soldier provisions often included forms of health care, pensions, survivor benefits, and compensation for injuries obtained on duty. The utility of welfare provisions for expanding and defending a nation’s borders simultaneously aided in developing and expanding allegiance of soldiers to that nation. This process can be transferred to workers and their support of their organizations, political parties, and the national government (Obinger et al. 2018).

Theoretical Framework: Nation-State Institution and Codification

The idea of a nation-state and democratic rule of law was quite new in the late 1800s when states first started making national work-injury policies. It was precisely the expansion of rights by a state that led to workers having new forms of power to exert in the state (voting, the right to organize). At the same time, the architects and policymakers developing nation-states were able to build their own bases of power by offering these rights (Marshall 1950), and with these rights came institutional developments like social insurance and the stable bureaucratic structures to implement it that strengthened the worker-state relationship. The allegiance of workers in all their forms was crucial to nation-building because they represented the largest segments of society demographically and were increasingly exposed to resistance tactics that threatened the state through urban concentration and the spread of Communist doctrine.

Theories of power resources and institutions argue that politics and the ability of special interests to organize and exert power influenced the timing and scope of work-injury laws. As argued by Korpi (1983), the agents of the working class—especially unions, socialist movements, and “left”-parties—competed with employer organizations and special interest groups for influence in development of social, economic, and political institutions via the state. The mobilization of the working class, motivated in particular by the spread of new socialist ideas, imposed pressure on the old and new power elites, forcing them to take strategic action. It is thus a combination of power resources and ruling strategies which fostered the stabilizing role of social insurance in nation state-building. The Communist movements were just one example of the threat of revolution that lurked among populations in early European state-building after the French Revolution (Breznau 2020).

In Bismarckian Germany, for example, the introduction of social insurance went hand-in-hand with the Sozialistengesetze, which were various laws prohibiting a range of socialist, social-democratic, and communist associations. These laws were dismantled in 1890 due to the continued success of the socialist party and its allies; however, by this point Germany had introduced the world’s first national, and relatively comprehensive work-injury social insurance scheme. Arguably, this contributed to the failure of a socialist revolution despite strong socialist organization, because the state became strong through public legitimacy, not only the use of force. In a similar vein, Béland and Koreh (2019) suggest that in both Canada and Israel social insurance policies played a prominent role in state-building, despite extremely different historical and institutional trajectories. Just as with taxes, the authors stress the importance of revenue raising in contributory schemes, emphasizing the possibility to use those programs to expand state fiscal capacities and legitimacy. As states legislate in more areas and control or regulate more forms of insurance, the state itself becomes indispensable to the economy and social welfare.

Whether giving or regulating or simply reacting to worker revolts, state-building was certainly not a linear, “friendly” process, but one of contradictions and conflict. Nation-building entails attempts to vest the highest level of power and authority into nation-state governance, a power to which all other power resource groups (parties, unions, employer organizations, etc.) are subjugated. From a Gramscian perspective, this would be construction of a hegemonic state through both coercion (strong laws against revolutionary organizations) and consent (providing citizenship, new rights, and social insurance).

State leaders and policymakers did not operate in isolation. They constantly engaged in policy learning from other states as they competed not only for strengthening local, regional, and global power and stability. This was facilitated by international organizations, political treaties, and trade. Liu and Leisering (2017) argue that the adoption of Bismarckian systems was the product of international norm-setting by the International Labor Organization (ILO). For example, Japan directly imported German civil law and later Bismarckian social insurance principles in its (strong) state development, and the latter was specifically intended to head off labor problems witnessed across Europe (Gordon 1985). The Communist party was certainly an international organization by the late 1800s and had great influence on worker movements even before the ILO and in countries across the globe from the USA to Japan.

It is therefore our main contention that work-injury and especially social insurance were particularly important in the process of nation-building. Given that work-injury protection in the form of social insurance tended to preclude or coincide with other forms of social security (pensions and unemployment protection, for example), it should be a key event in the construction of successful nations. As it requires legitimacy and a clear nation-state within which to enact the policy, it should come in the early stages of nation-state development. Therefore, we hypothesize that social insuranceis more likely in the years immediately afternation state formation.

At the same time, employer liability policies were often enacted to placate rather than empower workers. They paid lip service to worker movements and calls from elites or humanistic organizations to help the lot of the suffering workers. In many cases they were simply laws to achieve recognition and clout internationally, as with appearing ‘in line’ with the powerful ILO; workmen’s compensation being among the earliest conventions (C012 1921 and C017 1925). Given that employer liability laws did not specifically link workers with the state in a solidaristic way because they left the burden of proof (thus risk) on the workers, we expect that enactment of employer liabilitylaws is not more likely in the years immediately afternation state formation.

We incorporate our unique hypotheses within the larger theoretical perspective motivating the collective research in this book. Namely, we are also investigating the development of social policy as a process of globalization, so we test our hypotheses while simultaneously attempting to account for diffusion and historical trajectories. Diffusion processes related to colonialism, trade, migration, and culture are often theorized and tested in the development of social spending or the introduction of welfare state laws (Collier and Messick 1975; Schmitt 2015; Egger et al. 2017); but we are aware of no study that simultaneously accounts for both density of adoption among network ties and event history trajectories to estimate the likelihood of adopting work-injury insurance.

Data and Methods

All data and replication materials are available in our Project Repository.¹ We use the Global Work-Injury Policy Dataset (GWIP v1.0) (Breznau and Lanver 2020) to measure the year of introduction for our dependent variables first law and first social insurance. We recode the latter to the year when all formal blue-collar workers were de jure covered. Some countries introduced a first social insurance law that covered only certain segments and we do not count this. To be consistent with our nation-building theory, the law should apply to the industrial, blue-collar workforce whose allegiance is arguably necessary for national success, and certainly strong development in a global capitalist economy. Full information on all covariates is available for 151 countries from 1880 to 2010.

Our primary test variable is the year that a country became an independent state and the four years following. We also include the year prior to state formation to account for any measurement error associated with chopping time data into yearly points; although in only two cases, a law appeared the year before state formation. We would argue this is not coincidental but a part of the state formation which starts many months or years prior to a formal declaration of independence. To identify state formation, we take the institution of a state government that had autonomous or semi-autonomous (usually under the purview of a monarch) domain over policymaking most often established via a constitution and covering consistent borders or people as our criteria. For example, Germany in 1871 and Japan in 1890.

Our other independent variables in the analysis account for modernization and political regime by measuring Gross Domestic Product (GDP) per capita in ten thousand 2010$US and the degree of democratization using the regime scale from the Varieties of Democracy project. The latter is a continuous measure from most authoritarian to most democratic. Next, we incorporate four network variables measuring colonialism, culture, trade, and spatial proximity. These variables are described in detail in Chapter 1. Briefly, culture refers to similarities in time-varying indicators of political liberties, rule of law, gender roles, dominant religion, language group, government ideology, and Huntington’s classification of civilizations. Trade refers to dyadic trade densities provided from the Correlates of War data and spatial proximity refers to capital distances. As network variables, they measure both dyadic network linkages and the rate of diffusion among closer ties. Thus, their effects are weighted, a process also described in more detail in Chapter 1. See also the methods of Valente (1995).

We tested two network diffusion effects of colonialism, one normalized where the effect of colonialism is constant after colonialism ends and one where former colonies are still treated as part of the colonial network after independence. This matters little empirically, as in both cases the point estimates are close to zero and/or show huge confidence intervals (Models 4 and 5 in both Appendix Tables). Consistent with modernization arguments, we tested the robustness of our findings using percent of the labor force in agriculture for a restricted 83 country sample using the Banks CNTS data, but it was not any more informative than our GDP variable and thus not reported here (see Project Repository for these results).

Results

We start with a descriptive portrayal of the event of nation-state formation in Fig. 2.1. Each row of the figure on the y-axis is a single country plotted across the years 1880–2010. We centered the year measure so that year of state formation is zero. Thus, the black vertical line is when each state became independent. The dark grey segments represent time periods in each nation’s history where there was no work-injury coverage. The brown segments indicate a country has employer liability, and the green segments indicate a work-injury social insurance scheme.

Fig. 2.1

Nation-state formation as a “treatment” effect for the introduction of work-injury law

Figure 2.1 shows that social insurance is something ubiquitously introduced after nation-state formation. This means that establishing a constitution, declaring independence, or various other events that established a politically self-ruling state entity are changes in state status that occur prior to social insurance introduction. Of 150 countries that enacted any law, 133 (89%) introduced social insurance. Of the countries that introduced social insurance by 2010, 25 (19%) introduced it during state formation. Moreover, many countries that have not introduced social insurance are much younger (indicated by the missing segments in the lower right of Fig. 2.1). Whether this is a causal relationship, we cannot conclude; however, the fact that many states establish social insurance at the moment of statehood or within the first 4 years thereafter suggests that social insurance is often part of state and nation-building, given of course that many other factors are at play, and nation building is a process that is continual rather than discrete. Further evidence comes from our regression models whose main coefficients are displayed in Fig. 2.2 (full results found in Tables 2.1 and 2.2 in the Appendix).

Fig. 2.2

Discrete-time hazard diffusion models of work-injury law in 150 countries, 1880–2010, Log-Odds

In Fig. 2.2, the blue lines refer to Model 2 and the red lines to Model 3 (see Tables 2.1 and 2.2 in Appendix), these models are identical except for the inclusion of network exposure by spatial proximity. Moreover, Models 2 and 3 are otherwise identical for the dependent variables first law (left panel, Models “A” in Table 2.1) and first social insurance (right panel, Models “B”). The results reveal that spatial proximity, state-founded, and democratization all exhibit positive statistical associations with the yearly likelihood of adopting a first law. The effects are considered important because they reflect a plausible statistical range that is significantly greater than zero with a 95% confidence interval (CI) after robust clustering standard errors by country. The “OR ~” labels added to the plot for the significant coefficients are exponentiated coefficients, “odds ratios,” representing the increase in likelihood of adoption in any given year with a one-point increase in that variable, all else equal.

The spatial proximity variable is weighted by both the distance between the states and the percentage of members adopting a first law, thus it is difficult to interpret. However, we can say that the difference between having few other states nearby adopting and most states nearby adopting is a 31-fold increase in the likelihood of adoption in a given year. The CI is very wide, so this is not a highly reliable value and might range from just a few times more likely to 50 times more likely.

The state-founded variable suggests a 2.2-fold increase in likelihood of adopting a first law, and the democratization variable has a 1.2-fold increase. However, we have to keep in mind that odds ratios are always relative to the actual baseline likelihood of adoption in a given year. This is difficult to pinpoint because of the time period dummies necessary in this type of model. In Table 2.1 we see that an average likelihood of adoption in any given period of a first law (Model 3A) is 0.003 or roughly 0.3% all else equal. The state-founded variable represents a (2.2*0.3% =) 0.66% likelihood of adoption. The window of five years of state formation is our test variable, thus the likelihood of adopting at some point in the entire five-year period is 3.3%.² The democratization variable is measured on a scale from 0 to 9, so we can say that when a state goes from a score of 2 (or “rather authoritarian”) to a score of 7 (“rather democratic”) the odds actually increase by (5*1.2 =) sixfold leading to a likelihood of (6*0.3% =) 1.8%; a perceptibly large increase.

The results are slightly different for adoption of a first social insurance law. Here, network exposure to trading partners matters as well as the other significant effects for any first law. We can say that the state-founded variable is even more influential. Taking again a rough average of 0.3% baseline likelihood (Model 3B) a country is (3.6*5 =) 18-fold more likely to adopt during state formation, which is an overall likelihood of (0.3*18 =) 5.4%, a considerably sized likelihood in a five-year-period.

Note that our selection of Models 3A and 3B was supported by a better fit to the data than Models 1 or 2 or any robustness checks in both dependent variables’ cases.

Discussion

A political economy perspective on institutions suggests that democratic regimes are more likely to enact universal policies because the citizens are relatively equal, at least in comparison to authoritarian regimes where certain groups are often targeted to maintain power (Grünewald 2021). The measure is fraught with uncertainty, however, as who qualified as a voting citizen changed dramatically over time. Women were excluded early in democratic development and often racial/ethnic/religious groups as well. Moreover, workers were not citizens in early democracies and the franchise extended only slowly from high-status landowners or nobility to petty bourgeoisie to the working masses who were also at first mostly working in informal or subsistence labor.

What is ostensibly striking is the lack of association of GDP with likelihood of adoption. We are careful here because democracy could be seen as a process of modernization, or at least as a form of institutional isomorphism along a modernizing trajectory. Therefore, we do not interpret the GDP coefficient as evidence against the modernization thesis. It is more likely an artifact of event history modeling in many countries. The ranking of countries in 1900 is roughly the same in 2000 in terms of GDP, and this means that most countries follow more or less a yearly, linear trajectory in GDP growth. Therefore, GDP tracks time in the hazard model and fails to introduce unique explanatory variance. To demonstrate why modernization probably matters despite the insignificant GDP coefficient, we divide our countries into a mean split between “high” and “low” GDP in 1900. Then we plot the cumulative hazard rates of adoption in Fig. 2.3 by group. The darker blue hazard lines show that those with higher GDP in 1900 are more likely to adopt sooner than those with lower GDP in 1900. This is evidence of the role of GDP, but we could of course divide the sample into more and less democratized and possibly “Western and non-Western cultural spheres” and plot a similar hazard rate trajectory. As these are all interrelated (Breznau et al. 2011, footnote 5), we do not argue for a direct effect of GDP per se but are careful not to rule out development and modernization as playing an important role, also indicated by a large effect of democratization in our models.

Fig. 2.3

Cumulative hazard rates of adoption by low versus high GDP in 1900

What is not so often explored in the social security and welfare state development literature is the concept of spatial proximity. The classic study by Collier and Messick (1975) and work by Castles and colleagues (Castles and Mitchell 1992) suggests that there are families of nations when it comes to social policy adoption timing. Thus, the early adopters were European, then came the British influence sphere (New World and Japan), and then Latin America and so forth. Our study reveals more about the mechanisms behind this process than the previous works because it includes multivariate event history modeling with network diffusion effects. This gives a direct statistical link of dyadic ties and network weighting as a model of the theoretical reality of social exchange and transmission of ideas through the convenience of spatial contact.

In a way, our study opens a new area for researchers to attempt to disentangle trade—what is passed through actions related to economic exchange—from the transmission of ideas based on both convenience and political networks. We assume that ideas passing through international organizations such as the ILO, religious and charity organizations, the Communist party, and university exchanges are endogenous to our measure of trade partnerships. States that trade more with each other also have exchanges in these other areas. This is often facilitated by shared culture, language, or geopolitical agendas. As our reproducible code and data are freely available, we look forward to researchers expanding on our ideas and models. Nonetheless, the trade effect is striking, and we are not aware of a strong theory of policy diffusion based on trade in the welfare state literature.

Conclusion

Using an event history time-series model of 150 countries from 1880 to 2010 with the inclusion of network diffusion variables, we show that state formation is an important statistical predictor of the introduction of a first work-injury law in a given nation’s history. It is an even more important predictor of a first social insurance law, increasing the likelihood of adoption during the five-year early state formation phase by 18-fold. This points to the important role of social insurance in state-building. Of course, stronger states with more wealth or geopolitical power were more likely to adopt social insurance, especially earlier in their histories, as with the colonial empires of Western Europe and later the Russian Communist Party ‘empire’ constructed via the Soviet Union where countries adopted social insurance as part of the system of socialism. Both western and eastern Europe saw social insurance arrive after the start of industrialization, but the former did it without workers overthrowing their governments. In both cases, workers attaining social protections, in addition to expanded rights in the transition out of serfdom, represented a means of constructing a nation in addition to a state (western Europe) or a nation in opposition to the existing state (Russian and Communist revolutions). The cohesiveness of the nation, or what many refer to as solidarity, was built on this enfranchisement and protection of workers, who in exchange offered more efficient or effective labor to improve the state.

We argue that state formation and strengthening are compatible with power resource theories. Once established, the state is vying for complete control over the individuals and organizations within it. Moreover, without a productive workforce that has a sense of national unity, the state is less likely to be as economically competitive or as defensible against outside invasions. We expect a similar mechanism is at play between states and industrial workers as with states strategically using social provisions to improve the number of soldiers and their willingness to fight for the nation. Thus, the introduction of a social insurance scheme that covers all formal industrial workers is a monumental state achievement and institutionalizes the state in a way not present beforehand. Essentially, the welfare of the state and worker become coterminous, where each has an interest in the other via social insurance.

There are limitations to this study; in particular, the timing of nation-state formation is difficult to measure. Germany became an independent state in 1871 but before this there was a German people that arguably constituted a nation, albeit within various forms of empire. After that, the geopolitical form of Germany changed during and after the World Wars, and then again with reunification. This is nothing particularly unique to Germany. Poland had shifting borders throughout the nineteenth and early twentieth centuries as did China and countries that once comprised the Soviet Union. Austria is another challenging measurement case because it was part of the Austro-Hungarian Empire but also operated as a semi-autonomous state introducing social insurance for miners in 1854 independently of the empire. Therefore, we should proceed with caution in interpreting results, as state formation is not a discrete event in many cases but an ongoing contested process. Finally, even though we found no clear association, colonialism may play a role in work-injury policy because the effect is confounded by culture and spatial proximity, which many former colonies have in common to some extent.

We did not expect an effect of state formation on the introduction of a first work-injury law, as these were largely not social insurance laws but employer liability laws that were ineffectual at staving work-related risks. The effect of state formationon social insurancefor blue-collar workers was much larger than on first work-injury laws statistically speaking, but it is not as clear cut as we expected. It is possible that employer liability had a symbolic role in appeasing workers in the very short term and was thus a step in the nation-state construction process, if only as a gesture. As we measured introduction of policies, we do not capture further developments of the state whereby policies are expanded in benefits and coverage and create even more solidarity, i.e. nation building. Given the new global data now available for investigating such phenomena, we see great potential for further investigation of this and other classic research on the welfare state that has been characterized by a focus on the Global North.

Appendix

See Tables 2.1 and 2.2.

Table 2.1 First law regression results

Table 2.2 First social insurance regression results