EMPIRICAL STUDIES OF PART-YEAR OPERATION, PRODUCTION TECHNOLOGY, AND LABOR MARKETS IN THE UNITED STATES, GREAT BRITAIN, AND FRANCE

Accounts of part-year operation stress the relationship between part-year operation and seasonal agriculture and the consequences of part-year operation for factory production and performance, particularly when part-year operation relates to transitions from home production to artisanal shop manufacturing and to industrial factory production. For industrial leaders like the United States and Great Britain, the transition to full-year factory operation occurred in the early to mid-1800s. France transitioned several decades later. In Russia, such processes were still unfolding by the end of the nineteenth century. Notably, in the United States, Great Britain, and France, part-year operation emerged without the agricultural labor market distortions usually stressed in the Russian case.

Two key issues frame the economic history of industrial transitions: first, whether such changes in factory organization were associated with improvements in productivity or firm survival (e.g., Sokoloff Reference Sokoloff1984), and second, why such changes happened when and where they did. Sokoloff and Dollar (Reference Sokoloff and Dollar1997) show that, due to a higher land-to-labor ratio, agriculture in the United States was more varied and less seasonal. Workers could concentrate in manufactories without consistent seasonal pressure from the agricultural labor market. In the United Kingdom, however, a lower land-to-labor ratio resulted in a higher reliance on grains, which are highly seasonal, leading industrial entrepreneurs to adopt the cottage industry model, which allows workers the flexibility to accommodate varying demand for agricultural labor. Similarly, Sokoloff and Tchakerian’s (1997) study of manufacturing in the rural South and Midwest using the 1860 U.S. Census of Manufactures shows that manufacturers were less productive and more seasonal in locations practicing wheat agriculture.

Even though U.S. manufacturers were less sensitive to seasonality than British producers, manufacturing firms in the United States were still somewhat tied to the labor demands of agriculture. Atack, Bateman, and Margo (Reference Atack, Bateman and Margo2002) document that 38–40 percent of establishments operated for fewer than 12 full-time equivalent months a year, suggesting a substantial presence of seasonal operations that accommodated agricultural needs. They also find that larger firms with more labor and capital worked, on average, longer than smaller firms, and suggest that large, capital-intensive firms had higher termination costs as well as an ability to accommodate for business cycles without suspending operations.Footnote ¹¹ Overall, factory-like firms in the United States were less dependent on agriculture and operated longer on an annual basis.

Magnac and Postel-Vinay (Reference Magnac and Postel-Vinay1997) explicitly connect factory operating time to seasonal labor markets in mid-nineteenth-century France. They find that a factory’s likelihood of shutting down is related to variations in urban and rural wages due to agricultural seasonality. Though their study includes more detailed information on relative wages, ours paints a more detailed picture of the consequences of shutdowns for technology and performance. They also suggest that if firms had inputs that were costly to adjust, they would be more likely to work full-year, a pattern we observe in our data.Footnote ¹²

Changing labor composition of manufacturing establishments also influenced industrial development. In the United States, women and children participated actively in industry; in the 1800s, 40 percent of industrial laborers were women and children. Goldin and Sokoloff (Reference Goldin and Sokoloff1984) argue that because women and children had relatively lower productivity in agriculture in the Northeast than in the South, they entered the manufacturing labor force in large quantities in the Northeast, which helped develop the northern industrial sector. This logic tenuously suggests that firms with higher female-to-male ratios would be more likely to resemble industrial, factory-style companies and thus operate for longer periods of time.

SEASONALITY AND INDUSTRIAL DEVELOPMENT IN THE RUSSIAN EMPIRE IN CONTEMPORARY ACCOUNTS

Contemporary sources debate the prevalence of part-year operation in the Russian Empire and describe how factory managers handled several peculiar features of the Russian labor market. Though many observers belabor then-fashionable socialist debates over Russia’s stage of development, they help form a picture of factory life in Imperial Russia. For example, socialist economist Tugan-Baranovsky claims that recent evidence has refuted the stereotype of a typical Russian factory worker as “a peasant living on the land who makes up deficiencies of his agricultural income by occasional factory work” and hence refutes the claim that “there is no factory proletariat in Russia” (1970, p. 355).Footnote ¹³ As evidence of the development of the Russian proletariat, Tugan-Baranovsky cites Dement’iev’s (1897) earlier study of Moscow Province factories in 1884/85. Dement’iev collected exhaustive statistics on the factories he surveyed, painstakingly counting Sundays and holidays to explain that most factories by this period worked essentially full-year with well-justified breaksFootnote ¹⁴ and that “only the more primitive factories, where hand labor still prevailed, allowed their workers to depart in the summer” (Johnson Reference Johnson1979, p. 35). Walkin (Reference Walkin1954, pp. 163–4) argues that Tugan-Baranovksy may have been led astray by Dement’iev, who likely exaggerated the degree to which full-year operation prevailed in Russian factories.

Some contemporary accounts indicate that many late nineteenth-century factories worked full year and that workers had severed ties to villages, though some part-year operation endured. For example, the personal accounts of P. Timofeev (first name unknown), who visited urban factories in the early twentieth century recounted that “many skilled workers…have never seen a wooden plow and do not have the slightest idea how to plant wheat because they have lived in a town and worked in a factory for the past twenty-five to thirty years” (Timofeev 1906, quoted in Bonnell (1983, p. 82)). Even in such accounts that predominantly refer to what were likely full-year factories in urban areas, those workers who still had ties to villages sent money to their families in the countryside rather than traveling themselves. Some “would go back there every year or two themselves” (p. 82). Yet accounts of factories operating part-year due to agricultural seasonality appear frequently in the literature. Tugan-Baranovsky describes how, for example, at Prince Bariatinskii’s woolen-cloth mill in Pronsk district, Riazan’ Province, as well as in Okulov’s woolen-cloth mill in Skopin district, workers were routinely released for two months a year to participate in agriculture (Tugan-Baranovsky Reference Tugan-Baranovsky1970, p. 83). Resolving this debate about the prevalence and economic importance of part-year operation in Russia requires the kind of comprehensive factory-level data our study provides.

Part-year operation’s impact on factory productivity frustrated factory managers, who viewed full-time operation as an outcome of effective factory discipline.Footnote ¹⁵ To factory managers, workers’ return migration to home villages for seasonal agricultural work “delayed the workers’ full commitment to factory life, promoted labor instability and unreliability, and hampered the development of industrial skills” (Rimlinger Reference Rimlinger1960, p. 71). Managers’ strategies for coping with this particular problem of discipline included locking factory gates and setting strict rules for who could visit factory workers (p. 71). Some managers may have offered positive inducements like wage bonuses to keep workers from leaving in the summer, though the historical record does not reflect widespread use of this practice.Footnote ¹⁶ Russia’s agricultural institutions and how they evolved over time made this type of factory discipline problem particularly acute in Imperial Russia.

Seasonality in Russian factories impacted women and men differently within the household, which may have affected the relationship between part-year operation and labor force composition. In many accounts, when men left rural households during slow agricultural seasons to work in factories, women staying at home had to both perform their own tasks and their husbands’.Footnote ¹⁷ Female factory workers tended to be young and unmarried, widowed, or permanently settled in urban areas.Footnote ¹⁸ We investigated whether women and children tended to work in factories that operated more days annually or whether they tended to be employed in part-year factories facing labor scarcities.Footnote ¹⁹

FACTORY-LEVEL DATA AND AN OVERVIEW OF PART-YEAR OPERATION IN RUSSIA

Uniquely detailed data provided by the 1894 Russian Manufacturing Census (Russian Factory Production: List of Factories and Plants 1897) allows us to study the relationships among part-year operation, work-force composition, and production technology in the Russian Empire. The 1894 manufacturing data provides a factory-level snapshot of industrial activity in the Russian Empire at the end of the nineteenth century.

The database covers all industrial factories (i.e., no mines, farms, or oil refineries) in the Russian Empire during 1894/95, where a factory is defined as a productive entity with at least 15 employees, a steam boiler, a steam engine, or other factory devices.Footnote ²⁰ In other words, even a very small factory that otherwise met this definition by having factory tools or otherwise resembling a factory would be included. Indeed, the source’s authors seemed to interpret this criterion broadly: many factories with fewer than 15 employees and no listed machines appear in the census volumes. However, we acknowledge that our analysis only characterizes factories and thus excludes the multitude of small, handicraft establishments that likely represented a large share of Imperial Russian employment.Footnote ²¹ Given the patterns we observe for factories, we expect that such handicraft establishments may have been even more likely to work part-year, so our estimates likely represent a lower bound picture of the prevalence and importance of part-year operation in Russia.

Crucially, the database provides the number of days a factory worked in that year, our indicator of part-year operation. A shortcoming of this variable is the lack of information on working hours, which would provide a more exact measurement of total annual working time.Footnote ²² The database also lists a factory’s name; street address, which indicates the factory’s district location and whether the factory is located within or outside of a city;Footnote ²³ a complete list of production activities and products; number of workers by age and gender; types and power of all machines; fuels; and total revenue. In the data, the number of workers denotes the average of the total workforce in the factory each month the factory operated.Footnote ²⁴ This measure therefore gives us a good indication of the factory’s scale while fully operational, which is our main focus, even if it does not precisely measure the labor input in a production function sense.Footnote ²⁵

In the analysis that follows, we use total machine power as a proxy for the value of the kind of capital relevant to our story of part-year operation. To verify the correspondence of machine power to capital value, we can examine those factories for which the total value of capital is available: factories owned by corporations. We have matched corporate factories in the 1894 data to the RUSCORP Database (Owen Reference Owen1992). For these 545 factories, total machine power in the factory is correlated with the corporation’s total par value of capital (STCAP): the correlation coefficient is .1978, and in a regression of capital on total machine power, the estimated slope is .52 with a standard error of .11. Thus, total machine power and total value of capital are strongly related, at least for corporate factories, though total value of capital included additional components.

Table 1 Panel A displays descriptive statistics of key variables, including the number of days worked in a year, which we will use for most analyses in the paper. We will nonetheless focus several analyses on factories working more or less than the median number of days in each industry and region combination, which can be thought of as our definition of “full-year operation.” This is similar in spirit to Dement’iev’s (1897) approach since he also relied on empirical notions of full-time operation.

TABLE 1

AN OVERVIEW OF THE FACTORY DATA AND OPERATION DURATION

Notes: Total Machine Power is the factory’s total horsepower. Factories in dist-ind is the number of factories located in the same district and industry.

Source: Ministry of Finance, List of Factories and Plants (1897).

Table 1 Panel B breaks down part-year operation by industry. Given the large variation in working days per year by industry, we include industry controls whenever possible. Panel B also demonstrates that, generally, more mechanized industries like metals and machines, cotton, and chemicals tended to operate more days per year. Unsurprisingly, the least seasonal industries, like paper, tended to consist of factories mostly located in cities. Table 1 Panel C repeats this exercise by region.Footnote ²⁶ We find less variation in the number of working days by region than we do by industry, but some patterns exist. Factories located in northern regions, where the climate was harsher and less conducive to agriculture, tended to work more days per year than factories in the warmer and more fertile south. Thus, we also include region controls in most of our regressions.

Figure 1 Panel A reports a histogram for the number of working days in the factory per year, which shows considerable variation in the number of working days. We also observe some heaping, particularly at 100 days, 200 days, and 270 days, which would correspond to one season, two seasons, and full-time operation. Panel B of this figure displays separate histograms for factories located in cities (gorodi) versus in rural areas (uezdi). Urban factories tended to work more days than rural factories; a large mass of rural factories operated fewer than 250 days.

FIGURE 1

HISTOGRAM OF TOTAL WORKING DAYS

Source: Ministry of Finance, List of Factories and Plants (1897).

For some analyses, we also match the factories in our database to another manufacturing census that covers all of European Russia in 1900 (List of Factories and Plants of European Russia 1903). Factories are hand-matched by name, industry, and province. The 1900 source does not record information about factories’ working days per year, but we can use the source to test whether factories in European Russia that worked more days per year are more likely to survive and be recorded in the later manufacturing census. The 1900 census also generated an aggregate volume (Statistical Results on Factories and Plants by Industries Not Subject to the Excise Tax for 1900 1903) that provides some additional variables at the province-industry level, including the wage rates we use in some analyses.

The 1894 and 1900 factory lists are quite similar, as both of their creators attempted to capture information on all factory establishments in Russia,Footnote ²⁷ though there are two important differences between them: the 1900 list includes only factories in European Russia, and the 1894 list seems to include more small enterprises. Otherwise, the criteria for inclusion in the data were almost the same for both sources. The 1900 census excludes factories with fewer than 15 workers that did not use mechanical motors or had a general “handicraft character,” equivalently to the 1894 census. And similar to the 1894 census, in many cases, the line between factories and handicraft establishments was rather blurry, though the 1894 data seemed to have included more factories below the 15-worker threshold.Footnote ²⁸ In our survival analysis, we thus exclude 1894 factories outside European Russia and those with fewer than 15 workers to bring these sources to close parity. Although excluding firms with fewer than 15 workers would mean excluding a considerable number of smaller establishments, we believe we can still observe a meaningful effect of size on factory survival, even if the conclusions can only definitively be applied to larger establishments.

HYPOTHESES AND EMPIRICAL METHODS

Our study seeks to understand Russia’s industrial development at a moment in transition. We thus make two sets of predictions. First, we relate a factory’s working days to observable factory characteristics, particularly location, age, machinery, and labor force composition. Here we partly replicate Atack, Bateman, and Margo (Reference Atack, Bateman and Margo2002) to form comparisons with the better-known United States context. Second, we consider part-year operation as a transitional feature of Russian agriculture and test whether operating more days per year was associated with greater productive efficiency and the likelihood of survival. We also attempt to draw connections between the intensity of agricultural activity in a region and the likelihood of factories there operating fewer days. Though our results will be correlational, they support a particular narrative of part-year operation in Russia.

We argue that factories stayed open only part-year primarily because of labor market constraints due to seasonal agriculture, which has several implications for how part-year factories operate. First, and most simply, factories that operate part-year will be more likely to be located in rural districts rather than cities. Furthermore, we will investigate whether factories located near other factories, especially in the same industry, may take advantage of denser labor markets and operate more days per year. And given the many systematic differences across regions shown in Table 1, perhaps due to the timing of seasons and the degree of industrial development or concentration, our analyses will include controls for regional location.

Next, we investigate the relationship between working time and factory age. Because we hypothesize that part-year operation is costly, we predict that factories operating fewer days a year will be younger, because they will tend to fail at higher rates. Factories operating more days a year will be older and may appear to perform better simply due to survivorship bias.

To further investigate the relationship between operating days and the labor market, we also examine factory wages, as permitted by the available source, the 1900 province-industry aggregates. We hypothesize that, while factories that operate more days per year may pay higher total wages, factories experiencing the kind of tight labor market that would force them to operate part-year may also have to compensate workers at a higher wage per day rate (similar to predictions made in Atack, Bateman, and Margo Reference Atack, Bateman and Margo2002).

Furthermore, factories that operated part-year would have to leave any fixed capital, for example, large machines, idle when they did not operate. A factory manager could lessen the impact of a large machine purchase on the factory’s average total cost by staying open and producing more units.Footnote ²⁹ Shutting down and restarting machines after long idle periods may also be a costly process in and of itself. Thus, we predict a positive relationship between the number of days per year a factory works and the factory’s use of expensive, powerful machines. Likewise, factories with less expensive machinery were probably more likely to release workers to agriculture and shut down for part of the year. In the analysis that follows, we will only consider capital invested in machines and then only measure it as total power and not as a ruble value, so we acknowledge that we only capture a portion of a factory’s relevant fixed capital for the shut-down decision. Also, given the important differences in production technologies across industries, which likely underlie the patterns shown in Table 1, our analyses will include controls for industry.

Further, we can make a strong prediction about the level of capital (as measured by machine power) in the factory, but the hypothesized relationship with capital intensity seems ambiguous. Total working days may be positively related to the total machine power in the factory since machine power left idle is costly. Total working days may also be related to the factory’s overall capital intensity (machine power per worker or per unit of revenue) if factories that operated part-year would use less capital to produce the same amount of output. If part-year operation placed a constraint on the overall size of the factory, we may see a relationship with the size of total capital but not capital intensity. In fact, part-year operating factories may appear more capital-intensive if they are unable to substitute labor for capital as much as they would like.

Operating full-year vs. part-year may also have been related to the factory’s labor force composition, since part-year operation may have reflected constraints on the labor market. However, the direction of the relationship is somewhat ambiguous. Part-year factories located in the countryside may have chosen to employ women and children otherwise idle during the agricultural off-season. On the other hand, women who lived in cities may have had the opportunity to spend more time at manufacturing firms as they would have had no responsibilities related to agriculture.Footnote ³⁰ Furthermore, urban factories may have been larger and more mechanized, which facilitates employing women and children due to division of labor.Footnote ³¹ Thus, interactions among urban location, size, and machinery may be relevant to the relationship between labor force composition and working days.

Most likely, factories determined their schedule of working time as well as other features of operations like location, size, and mechanization simultaneously. It would be impossible to determine, for example, whether a rural factory was forced to shut down for part of the year or whether a factory that could operate part-year might decide to locate in the countryside. Similarly, we cannot rule out that factories with less expensive machines would find it less costly to shut down, rather than factories that knew they would have to shut down choosing smaller initial total capital. Our investigation, therefore, stresses correlations among these features rather than emphasizing a particular causal argument.

Furthermore, our paper emphasizes supply-side pressures on factories to operate full-year or part-year, which we argue is a first-order concern for Russian factories. We acknowledge, however, that there are other reasons, apart from these, for factories to operate part-year. For example, demand for certain products like building materials or certain kinds of food products may have been seasonal, or factories in certain industries like lumber would have operated seasonally for other reasons related to the production process. For these reasons, we include region and industry controls whenever possible, and we argue that the processes we describe characterize decisions made by a typical Russian factory but do not represent a universal explanation of part-year operation. Finally, we note that our argument does not assume the possibility of migration, which would only lessen the labor market pressures we may see reflected in our results. To test these hypotheses, we first consider a series of two-sample and two-proportion hypothesis tests comparing full-year and part-year factories. Then, we consider regressions of the form:

$$\eqalign{ & {\rm{log(}}Working{\rm{ }}Days{{\rm{)}}_{ijr}} = {\beta _0} + {\beta _1}\log {(Machine\,Power\, + \,1)_{ijr}} \cr & + {\beta _2}\log {(Total\,Workers)_{ijr}} + {\beta _3}Cit{y_{ijr}} + {\beta _4}Ag{e_{ijr}} + {\beta _5}Age_{ijr}^2 \cr & + {\beta _6}\,Number\,of\,Local\,Factorie{s_{dj}} + {\mu _j} + {\eta _r} + {\varepsilon _{ijr}} \cr} $$

This regression estimates the correlation between the number of working days in factory i in industry j, district d, and region r and the factory’s quantity of machinery (as measured by the log of total horse-power plus 1), its size as measured by the log of total workers, whether the factory was located in a city, the factory’s age and age squared, the number of factories in the same district and industry, and controls for industry and region. We estimate the regression with OLS and with a Tobit model right-censored at 6, since ln(365) = 5.90.

Next, we examine the relationship between a factory’s operating time and economic performance. Factories that operated fewer days per year may have paid more fixed costs from starting up and shutting down. Thus, we expect factories that operate more days per year to produce more revenue given the same labor and machine inputs. However, because revenue is measured annually, that relationship is likely to be mechanically true. We also examine, then, whether factories that operated more days per year produced more revenue per day, with or without controlling for inputs. Note that there are doubtless several sources of bias in such an investigation. For example, better managers might have decided to operate continuously and also have implemented more productive processes. We assume that such biases are relatively small.

We also consider whether factories that operated full-year were more likely to survive to 1900. A factory survives if it appears in the 1900 factory census. We expect that factories that operated for a greater proportion of the year may have been more likely to survive. Consider a model of factory exit in which a factory that performs below a certain threshold is more likely to exit the market. Any factory characteristic associated with greater economic efficiency, and hence greater profits, will be associated with a greater probability of survival. Given that full-year operation may have been more productive and more profitable, part-year factories may be more likely to shut down and full-year factories more likely to survive. If an exit is costly, of course, not all factories will exit if they make low profits. Moreover, in this context, many factories might have enjoyed monopoly rents and hence might have made profits despite inefficient methods. Region and industry controls somewhat alleviate differences across markets. We stress, therefore, that while survival is associated with improvements in profitability from operating a greater number of days, factories may survive for reasons other than working time.

We tested whether factories that operated longer were more likely to survive using several approaches. First, we compare the proportion of full-year factories that survive to the proportion of surviving part-year factories, where operating full-year is defined as operating more than the median number of working days in a given Industry-Region cell. Second, we estimate several probit regressions, where the left-hand side variable is the probability that an 1894 factory will appear in 1900.

Finally, we investigate the connections between factory operating times and the agricultural, economic, and political environments in the districts and provinces in which they operate. In regions where the soil was especially fertile and agriculture generated substantial revenue, factories may have been more likely to shut down during part of the year. Sokoloff and Dollar (Reference Sokoloff and Dollar1997) show that in regions where grains were more economically important, labor was more likely to be diverted into agriculture because of the high opportunity cost of opting out of agriculture during busy seasons. We explore the relationship between the amount of common seasonal grains procured in a province and the average number of working days in factories in that province. We use data on province-level agricultural output collected by Kessler and Markevich (Reference Kessler and Markevich2020).Footnote ³² We hypothesize that in regions that produced more seasonal grains, such as wheat and rye, or alternatively, in regions with more fertile soil, factories would operate fewer days a year, as workers would be more likely to engage in agriculture in the summer months. We also include controls for total population and proportion of urban population provided by the Ministry of Finance Yearbook data from 1900.

Because many important variables determining factory operating time are likely omitted at the province level, we continue our investigation at the district level. Though district-level crop-specific yields are difficult to find, district-level soil quality is available and provided by Dower et al. (Reference Dower, Finkel, Gehlbach and Nafziger2017 and 2018).Footnote ³³ Soil quality may be a better measure since it is less likely to be endogenously determined than yields. While soil quality is a critical determinant of the importance of agriculture, latitude is strongly related to both soil quality and overall weather patterns, while population characteristics capture economic conditions important for determining factory operations. We thus also include district-level latitude, population, urban population, and rural density controls provided by Dower et al. (Reference Dower, Finkel, Gehlbach and Nafziger2018).

Furthermore, we consider two additional potentially important correlates of operating time. First, we test an alternative theory that factories tend to shut down due to work stoppages or strikes, in which factories in districts with many strikes would on average have fewer working days per year.Footnote ³⁴ However, one could also consider a relationship in the opposite direction: a report by Varzar (Reference Varzar1905, pp. 48–50) on the statistics of manufacturing strikes listed longer operating times (in terms of hours) as one of the major reasons behind disturbances, in which case one might expect to find a positive relationship if the workers’ grievances also extended to working days. Given the relatively small number of strikes relative to the number of factories, and given these possible confounding relationships, we expect to observe only a weak relationship, if any, between strikes and working days. We measure the number of strikes at the district-level from chronicles published by Pankratova (Reference Pankratova1950–1963) between the years 1861 and 1894, of which we use the 1890–1894 subset to capture strikes in years relevant to the factories in our dataset. While likely not exhaustive, this is the best data source documenting strikes in the early 1890s known to us.Footnote ³⁵ Finally, we divide the number of strikes by the number of factory workers to ensure that our measure of strikes is not simply capturing the location of industrial clusters.

Second, we tie our work to discussions of Russia’s agricultural institutions by testing whether a higher share of land belonging to repartition communes in a district would be negatively related to factory working days in that district. Repartition communes could periodically redistribute land among peasants based on the size of the household and how well the household met obligations, which would have been a strong incentive for laborers to leave factory work in the summer and rejoin families in the countryside.Footnote ³⁶ We hypothesize a similar negative relationship between the share of total land for any type of commune and working days. The data on communes come from Dower et al. (Reference Dower, Finkel, Gehlbach and Nafziger2018), in the spirit of Chernina, Dower, and Markevich (Reference Chernina, Dower and Markevich2014). Note that all agricultural, demographic, social disturbance, and commune land data are only available for European Russia.

For district d, we estimate:

$$\eqalign{ & {\rm{log(}}Working{\rm{ }}Days{{\rm{)}}_{ijr}} = {\beta _0} + {\beta _1}\,Fertile\,Soi{l_d} + {\beta _2}Populatio{n_d} \cr & + {\beta _2}Urban\,Rati{o_d} + {\beta _4}Rural\,Densit{y_d} + {\beta _5}Strikes\,per\,Factory\,{Worker_d} \cr & + {\beta _6}\,Share{\mathop{\rm Re}\nolimits} partition\,Commune\,Lan{d_d} + {\beta _7}Share\,Commune\,Lan{d_d} \cr & \,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\, + {\beta _8}Latitud{e_d} + {\varepsilon _p} \cr} $$

Fertile Soil takes the value of the proportion of fertile soil in a district. Urban Ratio is the ratio of urban population to the whole population of a district. Rural Density is calculated as rural population over arable land.Footnote ³⁷ Strikes per Factory Worker is the sum of the number of factory disturbances by district in the years 1890–1894 divided by the number of factory workers in the district. Share Repartition Commune Land takes the value of the share of repartition communes to total communal land in a district, while Share Commune Land is the share of commune land to total land in a district.Footnote ³⁸