Steam democracy up! Industrialization-led opposition in Napoleonic plebiscites

Abstract

Which dimension of economic development spurred support to democracy? This study focuses on industrialization as the dimension triggering the process of political “modernization”. It uses a new dataset on Napoleonic plebiscites under the second French Empire (1852–1870). The results in those plebiscites provide a detailed cross-départements (French main administrative units) measure of opposition to autocracy. This study uses the variations in the thriving French modernization to disentangle the effect of industrialization on the vote from the one of other dimensions of economic development. Doubling industrial employment in the Puy-de-Dôme département (median of the distribution) would have decreased support to autocracy by 2.5–5.0 percentage points. An IV strategy using distance to the first city having adopted steam engines, access to coal and waterpower as instruments suggests causality. The baseline results are robust to controlling for other explanations of the vote and to using alternative specifications and estimation methods.

1. Introduction

Autocrats are suspicious of economic development, because they face a trade-off between fostering and hindering economic development (Olson 1993). Economic development indeed increases future economic rents but jeopardizes the political status quo. Some autocrats are famous for having cautiously hindered development. Mobutu, for instance, boasted about never having built a road in Zaïre.¹ Is this behavior rational? Does economic development actually unleash a political opposition in autocracies? Lipset (1959) suggests that it does. He describes a broad process of “modernization” in which economic development fuels the demand for democracy and institutional change. What may ignite the process is unclear (Squicciarini and Voigtländer 2016), and no facet of modernization has so far emerged as the trigger driving the process leading to an opposition to autocracy (Acemoglu et al. 2014). For example, the level of GDP per capita, the emergence of a middle-class (Barro 1999), a split in the economic interest of the elite (Lizzeri and Persico 2004) and human capital accumulation (Papaioannou and Siourounis 2008) have been considered as drivers of modernization.

One of the drivers of modernization mentioned by Lipset (1959) has however been neglected: industrialization. Yet, it is often perceived as the main cause of the first wave of democratic transitions (Hirschman 1994; Fukuyama 2014). This dimension is moreover topical today because of concerns that the current deindustrialization could weaken democracy (Rodrik 2016).

Yet, previous studies of democratic changes did not focus on industrialization per se. Narratives described a broad process encompassing urbanization, industrialization, accumulation of human capital and cultural changes (Hirschman 1994; Fukuyama 2014). Empirical studies have not focused on industrialization but rather on urbanization (Glaeser and Millett-Steinberg 2016), revenues and human capital (Murtin and Wacziarg 2014) or did not distinguish between industrialization and other dimensions of modernization (Aidt and Franck 2015). Hence, the effects of the different dimensions of modernization on the opposition to autocracy have not been disentangled from the influence they have on each-other. Consequently, the literature failed to identify the impact of industrialization per se. Hence, it did not detail a causal chain explaining the process of modernization and the way modernization crystallizes into an actual opposition to autocracy remains unknown.

By using new data on the results of Napoleonic plebiscites in nineteenth century France, this article can identify the causal impact of industrialization per se on the opposition to autocracy. The Second French Empire (1852–1870) provides a remarkable framework to disentangle the various aspects of modernization.

Firstly, Napoléon III used male universal suffrage twice to signal popular support for the Empire. The first Napoleonic plebiscite (1852) saw a massive support to the Empire across-départements.² This support shrank in the 1870 plebiscite and cross-départements differences appeared.³ The results of those plebiscites thus provide a detailed and original measure of the emerging opposition to Napoléon III’s autocratic regime across-départements.

Secondly, under the Empire, France experienced massive modernization. It urbanized, modernized its infrastructure, and raised the stock of human capital.⁴ Nevertheless, the timing of industrialization, urbanization, growth and human capital accumulation differed across départements.⁵ The variation in industrial employment provides a direct measure of industrialization. It allows identifying and disentangling the impacts of industrialization, urbanization, wealth, and human capital accumulation. The identification is particularly clean as France was already a highly centralized country, and départements faced a homogeneous institutional framework, limiting unobserved heterogeneity.

Thirdly, the French case provides natural instruments to establish causality. In mid-nineteenth century France, varying access to power sources and technology conditioned cross-départements differences in industrialization. French industrialization mainly relied on the use of both watermills and steam engines. The adoption of the steam engine took off in Northern France and unequally diffused from there (Franck and Galor 2015, 2017). Accounting for both these dimensions, this paper uses the access to power sources (both availability of coal and presence of adequate water flows) and distance to the first city having adopted the steam engine to predict industrialization.

Taking advantage of these features, this study addresses two interrelated questions: Does industrialization influence the emergence of a political opposition in autocracies? Does this effect run through on another dimension of modernization?

The results suggest that industrialization fueled a political opposition to Napoléon III’s regime. Increasing industrial employment by 10 percent decreased the share of “Yes” ballots in the 8 May, 1870 plebiscite by 0.35–0.71 percentage point. Industrialization explains a difference of 2.26–5.51 percentage points in support to autocracy between a mildly-industrialized départements (Puy-de-dôme, the median of the distribution) and a top-industrialized département (Marne, the 75th percentile of the distribution). This effect remains stable when controlling for other dimensions of modernization, alternative explanations of the vote and using different specifications.

2. Industrialization and political opposition in autocracy

The current literature extensively documents the impact institutions have on economic development (Acemoglu et al., forthcoming). Yet, the existence of a reverse relation remains unsettled. Section 2 returns to the vast work on institutional responses to economic development to stress the theoretical underpinnings of this study. It first discusses how industrialization bred structural changes spurring opposition to autocracy. Second, it emphasizes how industrialization intrinsically prompts opposition to autocracy.

2.1 Indirect effects of industrialization on the opposition to autocracy: intertwined dimensions of modernization

First, by increasing productivity, industrialization raises revenues that in turn seed the support for new political regimes. Second, it prompts the formation of an opposition to autocracy by boosting human capital accumulation. Third, industrialization often interlaces with urbanization. This agglomeration effect may as well foster opposition to autocracy.

2.1.1 Industrialization, income, and opposition to autocratic rule.

Industrialization, through technological improvement, increases income. According to Lipset (1959), economic development in a country increases the demand for a democratic regime among its citizens. Barro (1999) empirically confirms this intuition of a positive impact of wealth on democracy using a panel of 100 countries. Empirical results from Barro (2015) confirm the importance of lagged GDP per capita on indexes measuring law and order and democracy using various estimation methods.

Further works document the impact of the level of GDP per capita on the likelihood of an actual democratic transition—the most likely outcome of a genuine opposition to autocracy. Boix and Stokes (2003) observe that the level of development in a country increases the probability of transition to democracy: above a GDP per capita of $12.000, a country surely undergoes a transition to democracy in the short run. They stress that this dynamic led the first wave of democratic transitions in Western Europe. Epstein et al. (2006) confirm these findings using a trichotomous measure of democracy in a sample of 169 countries from 1960 to 2000. Similarly, Treisman (2015) refines the modernization hypothesis and presents empirical evidence showing that GDP per capita in developmental dictatorships increases the probability of democratization after the incumbent dictator’s exit. By increasing productivity and generating wealth, industrialization encourages the transition to democracy. This transition is however expected to materialize only in the medium run, after a certain level of income is achieved.

2.1.2 Industrialization, increasing human capital and opposition to autocratic rule.

In nineteenth century France, industrialization may have boosted another factor fueling the demand for democracy: human capital. Because the complexity of the production process and mechanization increased the need for skills, industrialization contributed to increasing human capital. Franck and Galor (2017) investigate the impact of exogenous across-départements differences in steam engines adoption at early phase of industrialization on literacy and education achievement in France. They confirm causality from industrialization to literacy and education attainment. De Pleijt et al. (2016) find that English counties that adopted steam engines at early phase of industrialization later had a higher share of skilled workers. Overall, the literature establishes a positive impact of industrialization on human capital; either on educational outcomes or on skills.

Beyond the association between wealth and democracy, the accumulation of human capital plays a critical role in the process of modernization (Papaioannou and Siourounis 2008; Murtin and Warcziag 2014). Schooling indeed incites people to interact and collaborate. It therefore contributes to the formation of civic capital (Glaeser et al. 2007). Higher civic capital fosters political participation. Since a higher portion of society takes part in political activities, the support for an inclusive political regime increases. It moreover fuels political grievance if it is not rewarded accordingly (see Campante and Chor 2012; using the Arab Spring as an illustration). Accordingly, industrialization may have spurred opposition to Napoléon III via this channel.

2.1.3 Industrialization, urbanization, and opposition to autocratic rule.

Industrialization is often accompanied by urbanization.⁶ It spurs the migration of the workforce to urban centers (Rosenberg and Trajtenberg 2004). Hence, it generates agglomeration effects prompting the organization of a political opposition and eases the access to “revolutionary technologies”.⁷ Urbanization is expected to foster the returns from these technologies as it simplifies their implementation and their diffusion. It also eases the diffusion of information and increases civic capital (Glaeser and Millett-Steinberg 2016). The diffusion of information and ideas matters for the organization of a political opposition (Aidt and Jensen 2014). Therefore, urbanization encourages the formation of a political opposition by reducing the cost of protesting and the expected cost of repression (DiPasquale and Glaeser 1998).

Yet, the capacity of dictatorships to control protests spreads the support to autocratic regimes (Djankov et al. 2003). If the autocrat fails to control them, then the drawbacks of autocracy may outweigh its benefits. As a consequence, people favor democratic institutions over autocratic ones to guarantee the implementation of benevolent public policies (Glaeser and Millett-Steinberg 2016). By encouraging protests, urbanization stresses the inability of the regime to control protests, which then fosters opposition to autocracy.

2.2 Intrinsic effect of industrialization on opposition to autocratic rule

Numerous structural changes stem from industrialization. However, industrialization per se may also directly encourage the emergence of an opposition to autocracy.

2.2.1 Industrialization and working-class.

A new poverty emerged in newly industrialized France. Chamborant (1842, p.186) observes: “the situation of the ploughman […] is miserable; but the one of workers in manufacturing, attached to machines, is hundred times worse”. Indeed, the transition from a mostly agricultural to an industrialized mode of production widens inequality (Kuznets 1955), in accordance with empirical evidence from nineteenth century industrialization in France (Morrisson and Snyder 2000; describe an increase in inequality during the Second French Empire).

This rise in inequality increased demand for inclusive policies. Justman and Gradstein (1999) develop this argument using industrialization in Great Britain as an illustration. Increasing revenues due to industrialization widened the income distribution. The median voter decided to implement inclusive policies as the disenfranchised control resources to ensure output maximization.

2.2.2 Industrialization and new interest groups.

By increasing inequality, industrialization may have raised concerns over social issues, thus encouraging inclusive policies. Beside the new working class, it also promotes a new interest group: industrialists. The utility of this group directly rests on the benefits of the industrial activity. A pre-existing group (the landed elite) does not benefit from the industrial activity. A conflict between the two elites then arises over the type of public policies to implement: “clientelism” and particularistic policies reducing growth or public policies increasing public goods and productivity (Lizzeri and Persico 2004; Llavador and Oxoby 2005; Seim and Parente 2013). This split reduces the political hegemony of the landed elites. Congleton (2004) in particular shows that by empowering industrialists industrialization increases the representativeness of parliaments. Industrialization in that sense vivified the political participation of certain groups.

2.2.3 Reaction of interest groups to redistribution.

Industrialization increases political grievance (e.g., because of widen inequality) and empowers new interest groups (e.g., working class and industrialists). Since industrialization enlivens new political groups, its effect on the demand for democracy hinges on collective actions due to redistribution. As a consequence, discontent may emerge from both new groups protecting the emerging sector or groups from declining traditional sectors.⁸Ansell and Samuels (2010) show that in a case of rising income inequality, comparable to the consequences of industrialization; new economic groups will have a direct advantage in changing the type of regime in order to be protected from expropriation. Industrialists, as they grow, may engage in a contract with the autocratic leader prompting a smooth transition to inclusive institutions. This logic applies to the 1870 plebiscite. The plebiscite was precisely on the consistence of inclusive reforms and its campaign mainly leant on the need for further reforms.

New interest groups may also demand democracy by another channel. Caprettini and Voth (2017) observe that the adoption of the threshing machine in Great-Britain encouraged the formation of the Swing Riots (which prompted the adoption of the Great Reform—Aidt and Franck 2015). Industrialization-led income redistribution fuels social unrest. This social unrest generates a “Threat of revolution” (Acemoglu and Robinson 2001). In response to this threat of revolution, elites implement more inclusive institutions to avoid a wasteful revolution.

3. Industrialization and the 1870 plebiscite

3.1 Peculiar French industrialization and political opposition

In the mid-XIX century, the French industrial sector was mostly composed of small units of production located in rural valleys (Lévy-Leboyer 1996, p183). These small units developed along waterways to benefit from both market access and waterpower. Meanwhile, craftsmen worked in small-scale workshops located in cities. In France, industrialization consequently barely correlated with other dimensions of modernization.

This rural industrialization generated political turmoil. In the Loire basin, workers sabotaged coal mines. In order to avoid a total lockout of the region, the army intervened, sometimes resulting in casualties. For example, at La Ricamarie in June 1869, an army’s gunfire killed fourteen civilians who were trying to stop a military convoy leading forty strikers to jail. On 8 October, 1869, the army directly shot strikers, killing seventeen persons in Aubin. No coordination existed at the national level; however Republicans used these strikes as a catch line to show the inability of the Empire to cope with the structural changes caused by industrialization.

3.2 The 1870 plebiscite

To contend with the opposition, Napoléon III promised that ministers would have to defend their bills before the legislative chamber in 1867 and also implemented liberal reforms (abolition of pre-required authorizations to publish a journal and to convene non-political meetings). In the 1870 plebiscite, voters were asked to approve the following statement: “The people approve the liberal reforms conducted by the Emperor with the help of the State services and ratify the senatus-consulte of 20 April, 1870”.⁹ The plebiscite directly referred to liberal reforms but its primary objective was to maintain the regime. The vote virtually had no impact on the implementation of the reforms but aimed at legitimating the autocratic nature of the Second Empire. As stated by Napoléon III on 24 April, 1870, “By voting yes, you will stave off the threats of the revolution, you will establish freedom and order on solid grounds. Finally you will help me to transmit the crown to my son”.¹⁰

For these reasons, the need for political liberalization was the main theme of the campaign. Gambetta, one of the most prominent Republican leaders, used that rhetoric several times.¹¹ Hence, voting “No” in the plebiscite mainly meant rejecting despotism and autocracy. A leaflet from the opposition listed the autocratic nature of the Empire as the main reason to vote “No” in the 1870 plebiscite.¹² Even partisans of Napoléon III used this interpretation of the vote. In an explanatory note, a citizen of Chalonnes supporting the Emperor questioned the motivations of “No” voters.¹³ He wrote “incorrigible revolutionaries want to vote against the plebiscite because, according to them, it does not grant enough liberties: Which ones? Should we abolish the Empire and establish a Third Republic?” As stated by both parts, the opposition constructed its discourse on the need for a Republic and more inclusive institutions.

4. Data

4.1 Support to autocracy

Figure 1 shows that the opposition principally emerged around Paris. The Rhone valley was another center of opposition, as well as the départements bordering the Mediterranean Sea. On top of these main regions, several départements clearly opposed Napoléon III. Northern-East France also showed little support to Napoléon III.

Interestingly, heterogeneity existed within regions. For example, in Southern-West France, Gironde and Haute-Garonne clearly opposed more than other départements in the region. The same applies to Brittany with the opposition appearing in the Finistère département.

4.2 Industrialization

Figure 2 shows the diffusion of industrial employment. Two regions are clearly at the forefront of industrialization in 1866: Northern France and the Rhone valley. Within-regions heterogeneity existed, among which Gironde in Southern-West France and Finistère in Brittany.

The two maps overlap. The Rhône valley, départements neighboring Paris were well-industrialized and opposed more in the 1870 plebiscite. Interestingly, the maps do not only exhibit similar regional patterns but the same within-regions heterogeneity: the Gironde département and the Finistère département, being industrial leaders in their regions, showed less support to Napoléon III in 1870.

5. Method

5.1 Empirical models

Discrete variables are expressed in logarithm to ensure that extreme observations do not drive the observed correlation.¹⁴ Regression coefficients are estimated using heteroscedasticity-robust standard errors. As in Franck and Galor (2017), the baseline controls include geographic covariates prompting modernization through agricultural development: latitude, yearly average temperature, yearly average rainfall, digitalized from a 5′×5′ grid map by Hijmans et al. 2005; and soil quality from Ramankutty et al. (2002). A dummy variable for départements neighboring Paris and the aerial distance to Paris account for the possible dual diffusion of political opposition and economic development from Paris. Dummy variables both for maritime départements and border départements account for a possible influence of foreign institutions and trade on the political preferences and economic development in these départements. Adding these control variables also disentangles pre-industrial development due to favorable natural conditions from actual industrial development.

The effect of trade is also directly controlled for using market integration in the 1790s as an additional control variable.¹⁵ The more integrated to trade a département is the less it is expected to support autocracy as trade encourages the formation of a middle-class (see Acemoglu et al. 2005) or because of a possible backlash following the Cobden–Chevalier treaty.¹⁶ Controlling for market integration in the 1790s also allows for distinguishing industrialization from pre-industrial market potential and development. In the more conservative specification, the baseline specification throughout the robustness checks section; the share of Yes ballots over votes cast in the 1852 plebiscite is added as an extra control variable to capture long-term determinants of opposition to autocracy.

5.2 Identification strategy

OLS estimates would be biased if political preferences determined the adoption of technology. For instance, Napoléon III may have used targeted economic development in area of high opposition to contend political opposition. As stated in his proclamation of 2 December, 1851, his goal was “to end a revolutionary era by fulfilling people’s legitimate claims”. In that case, the observed correlation would suffer from reverse causality and the coefficient for the main independent variable would be artificially inflated. Likewise, vote-buying using industrial programs in barely industrialized regions would generate a downward bias in the estimate. Omitted variables would have the same consequences (i.e., unobserved département characteristics influencing the 1870 plebiscite and correlating with industrialization).

An IV strategy tackles these possible issues. Exogenous determinants of industrialization, such as access to power sources and the geographic distance to the first city having adopted steam engines (Fresnes-sur-Escaut), serve as instruments. Franck and Galor (2015, 2017) show that the adoption of steam engines in nineteenth century France varied with the distance to Fresnes-sur-Escaut. The aerial distance of each département’s centroid to Fresnes-sur-Escaut is then added as a first instrument. In addition, Crafts and Wolf (2014) present empirical evidence of the importance of power sources (water flows and coal) in the location of the cotton industry in nineteenth century Great Britain; providing two additional instruments for industrialization. The share of a département’s surface covered by a carboniferous ( = coal bearing) geological strata measures access to coal (as in de Pleijt et al. 2016). Access to suitable water flows is computed using data from gauging stations in modern France.¹⁷ The measure is the fifth percentile of the historical distribution of the flows in cubic meter per second in each gauging station averaged at the département level (as in Crafts and Wolf 2014). It assesses the availability of suitable water flows for the industrial activity. The interaction between these two instruments stresses the possible complementarities between these two power sources, as steam engines have been implemented in some places as a support to water wheels (Crafts and Wolf 2014). These instruments theoretically respect the exclusion restriction. They are exogenous geographical features. They directly relate to industrialization and do not operate via another channel (e.g., urbanization or pre-industrial development).

6. The effect of industrialization on the opposition to autocracy

Table 1 reports the results of the baseline model focusing on the effect of industrialization on the opposition to the Empire. The adjusted R² is equal to 0.10 in the bivariate specification (Column 1.1) and reaches 0.44 in the specification with all baseline control variables (Column 1.6).The instruments meet the relevance condition (the F-statistic of the first-stage is above 10 in each specification) and the Hansen J p-value is above 0.10 in each specification.

Column 1.1 presents the unconditional effect industrialization had on the vote outcomes in 1870. The coefficient is significant at the 5-percent level and equals −3.626. For example, doubling industrial employment in the Puy-de-Dôme département would have decreased support to autocracy by 2.5 percentage points. When controlling for various geographic controls (Column 1.2), industrialization still hampered adhesion to the Empire. The coefficient is higher in magnitude (−5.924) and significant at the 1-percent level.

The other specifications (1.4 and 1.5) include market integration in the 1790s as an indicator of pre-industrialization market potential. Market potential in the 1790s also increases opposition to the Empire in 1870. This effect does not confound the one of industrialization; the coefficients for industrialization remain of the same magnitude and significant at the 1-percent level. Specifications 1.6 and 1.7 add the share of yes ballots over votes cast in the 1852 plebiscite as an additional independent variable to control for time invariant factors determining political opposition.

Regardless of the set of controls, the coefficients of the “Industrialization” variable remain negative and significant at the 1-percent level in each OLS specification. The magnitude of the coefficient is stable over the various OLS specifications (between −3.5 in Column 1.6 and −5.9 in Column 1.4). In IV specifications, the effect of industrialization is larger (significant at the 1-percent level coefficients between −4.44 in Column 1.7 and −7.15 in Column 1.3).

7. Robustness Analyses

In order to test for the robustness of my results, I estimate several other specifications. First, I add controls for other dimensions of modernization. Second, proxies for pre-industrial modernization are added as control variables. Third, alternative explanations of the vote are also considered. Fourth, a final series of robustness tests shows that the results are also robust to using alternative estimations techniques and variables of interest.

7.1 Controlling for other dimensions of modernization

This section adds measures of modernization at the time of the plebiscite as control variables. It ensures that the relation between industrialization and opposition to autocracy does not emerge because of an omitted variable bias focusing on the three other dimensions of modernization cited in Lipset (1959) (urbanization, wealth and human capital). Industrialization indeed correlates with other dimensions of modernization. In turn wealth (Lipset 1959; Gassebner et al., 2013; Treisman 2015), human capital accumulation (Papaioannou and Siourounis 2008) and urbanization (Glaeser and Millett-Steinberg 2016) all have an effect on democracy. These results ensure that the impact of industrialization on the preference for democracy does not operate only via one of those channels.

The first dimension of modernization to be controlled for is wealth. Specifications in Table 2 include indicators of social conditions as control variables: the number of “hobos” per inhabitant in 1866 and infant mortality in 1868 (as a proxy for across départements differences in economic situation—Miller and Urdinola 2010; Baird et al. 2011).

After controlling for wealth, the baseline results remain unchanged. The set of instruments is still valid (F-stat above 10) and satisfy the J-test (Hansen J p-stat above 0.1). The coefficients for the “Industrialization” variable vary between −3.2 and −4.8 and are significant at the 5-percent or the 1-percent level, in line with baseline results.

New specifications also control for urbanization as it is often mentioned as parallel to industrialization (Rosenberg and Trajtenberg 2004). The specifications presented in Table 3 control for urban population in 1866 and/or population density in 1866. This does not impact the quality of the results. The “Industrialization” variable bears coefficients between −2.7 and −4.4, which are significant at the 5 or the 1-percent level.

Table 4 presents the results when controlling for measures of human capital accumulation. Industrialization fueled human capital accumulation (as described in Franck and Galor 2017 or de Pleijt et al. 2016) or resulted from it. The baseline results remain unchanged even after controlling for various facets of human capital accumulation (coefficients for industrialization between −3.4 and −5.6 and significant at the 5 or 1-percent level).

As a result, none of the alternative dimensions of modernization explains the effect of industrialization on opposition to autocracy. As industrialization in France took off in rural areas, the various dimensions of modernization have been decoupled in time and in space; disentangling the intrinsic effect of each of them.

7.2 Controlling for confounding factors

As an extra test, this section adds a series of pre-Second French Empire controls (Table 5). This test isolates the effect of industrialization from pre-Second Empire outcomes that would not have impacted the 1852 plebiscite but could have impacted the 1870 plebiscite. They include pre-industrial development (population density in 1801), presence of enlightened elites (universities in 1700), local counter-powers (number of independent communes in 1700, number of bishops in 1700), pre-industrial modes of production (number of iron forges in 1834, number of mills in 1834) and pre-Second Empire preference for Napoléon III (vote share to Napoléon III in the 1848 presidential elections).

Only two of these variables turn out to be significant. The number of mills in 1834 bears a significant and positive coefficient and the vote share to Napoléon III is positively correlated with later support to the Empire. These effects however do not confound the one of industrialization on the opposition to autocracy. The coefficients of the “Industrialization” variable are of the same magnitude as baseline results and all significant at the 5-percent level. Baseline results and the validity of the IV strategy do not depend on long-term dynamics.

7.3 Alternative explanations of the vote

This section controls for alternative explanations of the vote to guarantee that the effect of industrialization on the 1870 plebiscite was due to differences in the support for democracy.

7.3.1 Religion.

Religious groups may have prompted innovation leading to industrialization (e.g., the Protestant ethics, Weber 1904) and the opposition to Napoléon III. For example, the minister of Cults decided to control Protestants’ evangelization meetings in 1852, which spurred indignation in the community. Napoléon also later got in conflict with the Catholic Church. The effect of religious groups on the plebiscite is consequently ambiguous.

As evidenced in Table 6, controlling for the importance of Protestants and Jews alters neither the adjusted R² nor the validity of the instruments. In addition, the coefficients for the share of Protestants and the share of Jews in a département are never jointly significant. The coefficients for industrialization stay at baseline levels (between −3.0, Column 6.5; and −4.4, Column 6.4) and significant at the 5-percent or the 1-percent level.

7.3.2 Fraud.

If electoral frauds were more common in non-industrialized départements, the coefficient for industrialization would be upward-biased. The number of reported frauds is nevertheless quite low, maximum two in a département. Furthermore, La Presse¹⁸ and Le Siècle¹⁹ report only cases of small-scale frauds and pressures (from municipal administration pressure to double-voting). Moreover, the impact of frauds on the vote is likely orthogonal to previous industrialization. As denoted by Miller (2014), Napoléon III used semi-competitive elections to gather information on the opposition to the regime.

To make sure that frauds do not drive the results, Table 7 adds the number of reported cases of fraud appearing either in Le Siècle or La Presse as an additional control variable. Adjusted R² remain above 0.4 in each specification and the instruments stay valid (first-stage F-stat above 10). The variable “Fraud” is insignificant, whereas the coefficients for “Industrialization” lay between −3.3 and −4.5 and are significant at the 5-percent or the 1-percent level. Controlling for the number of alleged frauds alters neither the significance nor the magnitude of the coefficients for industrialization.

7.3.3 Threat of revolution/economic downturn.

Industrialized départements may have been more impacted by trade agreements (e.g., Cobden–Chevalier treaty) or by the liquidity crisis in 1867. An economic shock may have prompted the opposition to Napoléon III (see Brückner and Ciccone 2011). The Comptes généraux de l’Administration de la justice civile and commerciale provide each year a detailed account of bankruptcies at the department-level which allows controlling for département-specific shocks.

Table 8 presents results when controlling for the liabilities of bankrupting companies per inhabitant in the two years preceding the vote. This measure accounts for the average cost that the inhabitants of a département had to bear during the economic downturn before the vote. The adjusted-R² remains above 0.4 in each of the specifications and the instruments stay valid. The coefficient for the liabilities of bankrupting companies in 1868 is significant at the 10-percent or the 5-percent level and negative (as predicted since 1868 was the through of a liquidity crisis).

A dummy variable for départements affected by the phylloxera in 1870 (from Banerjee et al. 2010) is added as a last control variable (Columns 8.7 to 8.10). The “Phylloxera” dummy variable bears a negative and significant at the 1-percent level or 5-percent level coefficient. Yet, this effect does not confound the effect industrialization had in the plebiscite as the coefficients for the “Industrialization” variable remain consistent with the baseline results and are significant at least at the 10-percent level.

7.4 Alternative specifications

7.4.1 Alternative measures of industrialization.

Table 9 tests the robustness of the results to using alternative measures of industrialization: the adoption rate of steam engines in 1865 and in 1847, the number of steam engines in 1847 (as in Franck and Galor 2015, 2017), the number of persons financially dependent on the industrial activity, the increase in the number of industrial workers and the proportion of industrial workers in the population.

All these variables negatively correlate with the support to the Empire. For all of them, an increase by one-standard deviation reduces support to the Empire by around five percentage points, in line with baseline results.

7.4.2 Alternative measure of the support for democracy.

To ensure that the correlation between industrialization and the share of “Yes” ballots in the 1870 plebiscite is not incidental, this section uses a measure of democratic consolidation at the onset of the Third Republic as dependent variable: the results of Republicans in the 1876 general elections (Table 10).²⁰ It allows guaranteeing that baseline results are not driven by a specific measure of support to democracy.

The coefficients of the “Industrialization” variable are significant at the 5-percent level both in the OLS specification and the IV specification and bear a positive sign. Industrialization during the Second Empire spurred democratic consolidation at the onset of the Third Republic. The measure of industrialization consequently influences other measures of preference for democracy. Results in the 1870 plebiscite actually proxy well the preference for democracy over autocracy.

7.4.3 Alternative estimations methods.

This section uses alternative estimation methods in order to tackle the possible biases emerging because of the structure of the data.

As the dataset only contains a limited number of observations, the relation may be driven by outliers. Hence, this section performs jackknife estimation techniques (both for the OLS and the IV specifications). Tables 11 and 12 present the minimum, the mean and the maximum values of the coefficient for the “Industrialization” variable after dropping each observation in turn.

For both specifications, the coefficients for the “Industrialization” variable are stable over the exclusion of different observations (−2.6 to −3.8 for the OLS specification; −3.2 to −5.2 for the IV specification) and significant at least at the 5-percent level. Moreover, the validity of the instruments does not depend on any single observation, the F-stat remains above 10 over the exclusion of each of the observations.

The next estimations tackle the issue raised by Conley (1999). If industrialization and opposition to autocracy followed the same spatial diffusion, then spatial autocorrelation may drive the results. To tackle this issue, the next set of estimates uses Conley-type of errors. Using Conley standard errors corrects coefficients variance by explicitly modeling observations’ geographical dependence. Here, standard errors are considered as possibly correlated if a département centroid is less than 100/150/200 km far from another département centroid (Table 13).

Accounting for spatial-autocorrelation slightly decreases standard errors and hence does not affect the statistical significance of the coefficient for industrialization. In the generalized two-stages least squares procedure accounting for spatial autocorrelation, the coefficient for the “Industrialization” variable is equal to −5.8 and significant at the 1-percent level (with a spatial heteroroskedasticity robust t-stat equaling −4.27). Thus, spatial autocorrelation affects neither OLS estimates nor IV estimates.

8. Conclusion

Industrialization triggered opposition to the Second French Empire in the 1870 plebiscite even after controlling for various dimensions of modernization (urbanization, wealth, human capital). To give an example, if industrial employment would have doubled in the Puy-de-Dôme département, support to autocracy would have been 2.5–5.0 percentage points lower. Anecdotal evidence points to social turmoil due to poor working conditions as a plausible transmission channel of this effect. The results are robust to controlling for a series of confounding factors and remain valid when controlling for various explanations of the vote.

These empirical findings have numerous implications. Industrialization impacts voting behavior and opposition to autocracy. Autocrats impeding development and industrialization to secure their position may be right. The current deindustrialization might bear consequences for democracy. One should nevertheless stay cautious before having such an interpretation. This study focuses on structural dynamics. Therefore, it identifies neither the transmission channels of this effect; nor opposition groups. Future research ought to study those aspects to get a better understanding of the consequences of the current deindustrialization.

Funding

The research leading to these results has received funding from the People Programme (Marie Curie Actions) of the European Union’s Seventh Framework Programme FP7/2007-2013/ under REA grant agreement no. 608129.

Supplementary material

Supplementary material is available at European Review of Economic History online.

Acknowledgments

I thank Toke Aidt, Alexandra de Pleijt, Raphaël Franck, Pierre-Guillaume Méon, Gabriel Mesevage, Kim Oosterlinck, Khalid Sekkat, Mara Squicciarini, and Franz Zobl as well as seminars participants at the London School of Economics and Political Science, Université libre de Bruxelles, Université Paris I Sorbonne, and conferences participants at the European Public Choice Society Annual Meeting, the Beyond Basic Questions workshop and at the European Historical Economics Society Conference for useful comments and suggestions.

Conflict of interest statement. None declared.

Footnotes

References