New technologies are a key driver of economic growth, but often diffuse at a remarkably slow pace.Footnote 1 The paradigmatic technology of the Industrial Revolution—steam power—is a case in point. Adopting steam substantially raised productivity (Atack, Margo, and Rhode Reference Atack, Margo and Rhode2019, Reference Atack, Margo and Rhode2022), yet it diffused slowly, and its aggregate growth impacts took more than a century to fully materialize (Crafts Reference Crafts2004). Adoption involved major capital investments and often required firms to operate at a greater size than before (Atack, Bateman, and Margo Reference Atack, Bateman and Margo2008). Thus, a potential barrier to diffusion was that most industrial establishments during the early stages of industrialization were too small to profitably adopt the new technology.
Our paper uses establishment-level data covering Sweden’s industrial take-off in the late nineteenth century to show that an institutional innovation—the limited liability corporation—contributed to the diffusion of steam technology, partly by enabling firms to expand the scale of their operations.
Limited liability was codified into law in most European nations around the mid-nineteenth century. In Sweden, the Companies Act of 1848 introduced the modern corporate form, which evolved into a liberal system of general incorporation by the 1860s. By granting owners limited liability, the corporate form may have facilitated the adoption of steam by lowering risks involved in making major capital investments as well as improving access to debt and equity markets.
To study the link between incorporation and adoption of steam, we use newly digitized data that allow us to track the organizational form and technology use of about 9,000 Swedish industrial establishments between 1864 and 1890. By the mid-nineteenth century, Sweden lagged far behind the industrial leaders, and few establishments relied on steam power. Larger establishments had a cost advantage in using steam power, yet the median industrial establishment employed about five workers—far too small to use steam. However, Sweden underwent a rapid industrial take-off in the 1870s, becoming one of the fastest-growing economies in Western Europe. Notably, the industrial breakthrough involved a widespread diffusion of steam technology, which became the most common power source among manufacturing firms within a decade.
Our analysis shows that the spread of steam technology was deeply intertwined with the rise of the modern corporation. We first document that corporations were more than three times as likely to have adopted steam compared to establishments owned by partnerships or sole proprietorships. OLS regressions show that incorporated establishments were significantly more likely to use steam also when flexibly controlling for industry and regional characteristics. A higher use of steam power among corporations is also evident conditional on establishment size and location in an urban area, that is, factors that are highly correlated with both incorporation and the use of steam technology.Footnote 2 However, the cross-sectional association captures both a potential causal effect of incorporation as well as underlying factors that may have determined selection into incorporation.
To identify the impact of incorporation on the adoption of steam, we use the fact that hundreds of establishments converted to the corporate form. We examine these conversions in a difference-in-differences framework, comparing relative changes in the adoption of steam among establishments that convert to corporate form relative to those that remain organized as partnerships or sole proprietorships. The probability that an establishment adopts steam increases sharply in the years after incorporation. While these estimates control for time-invariant differences across establishments, there may still be either positive or negative selection into incorporation due to time-varying factors. However, we show that the results are similar when directly controlling for pre-existing differences in size or allowing establishments to follow different growth trends. The same conclusion holds when drawing on recent advances in the event-study literature and solely relying on the timing of incorporation among the subset of establishments that, at some point, incorporated (Sant’Anna and Zhao Reference Sant’Anna and Zhao2020; Callaway and Sant’Anna 2021).
We lastly show that Sweden’s low barriers to incorporation facilitated the adoption of steam partly by enabling smaller firms to expand their operations to a size where it became feasible to adopt steam technology. Examining the heterogeneous impacts of incorporation, we find that the link between incorporation and steam adoption is mainly driven by initially smaller establishments, often located in rural areas with less developed capital markets. Typically, these establishments were too small to reap the rewards of steam prior to incorporation. However, after incorporation, they approximately doubled their workforce, reaching a scale required to adopt steam. To further bolster this interpretation, we draw on case study evidence underlining that smaller firms often actively used incorporation to simultaneously expand their operations and invest in steam, which was facilitated both by establishing limited liability among firm owners and improved access to long-term financing.
Our paper provides micro-level evidence of the corporation’s role in accounting for the spread of steam technology during the nineteenth century. Thus, we advance the literature that studies the diffusion of steam, which has mainly relied on regionally aggregated data where firm-level determinants—such as organizational form—cannot be identified (e.g., Atack, Bateman, and Weiss Reference Atack, Bateman and Weiss1980; Nuvolari, Verspagen, and von Tunzelmann Reference Nuvolari, Verspagen and von Tunzelmann2011; Gutberlet Reference Gutberlet2014; Bogart et al. Reference Bogart, Bottomley, Satchell and Shaw Taylor2017; de Pleijt, Nuvolari, and Weisdorf Reference de Pleijt, Nuvolari and Weisdorf2020; Franck and Galor Reference Franck and Galor2022; Ridolfi, Salvo, and Weisdorf Reference Ridolfi, Salvo and Weisdorf2022).Footnote 3 A seminal exception is Atack, Bateman, and Margo (Reference Atack, Bateman and Margo2008), who use cross-sectional establishment-level data from the U.S. Census of Manufactures, documenting that the use of steam technology is associated with greater establishment size.Footnote 4 We leverage the longitudinal nature of our data to contribute complementary evidence that firms actively used incorporation to achieve greater scale, which arguably was central in enabling many initially small firms to adopt steam.
Our findings also contribute to a contentious literature on the role of the modern corporation in accounting for late nineteenth-century growth (e.g., Landes Reference Landes1966; Chandler Reference Chandler1977; Rosenberg and Birdzell, Jr. 1986). Due to the scarcity of data, few studies have been able to identify how incorporation affected firm-level outcomes, and most of the recent literature has still relied on country- or industry-level data (e.g., Guinnane et al. Reference Guinnane, Harris, Lamoreaux and Rosenthal2007; Foreman-Peck and Hannah Reference Foreman-Peck and Hannah2015; Hilt Reference Hilt, William and Robert2015).Footnote 5 Gregg and Nafziger (Reference Gregg and Nafziger2019) and Gregg (Reference Gregg2020) are important recent exceptions, with the latter showing that incorporation among Russian factories led to increases in labor productivity and aggregate machine power. However, Russia lacked a general incorporation law and required firms to undergo a costly process to incorporate. In contrast, Sweden had introduced a general incorporation system by mid-century—similar to Britain, Germany, and the United States—where incorporation involved a simple process of registration, which arguably extended the benefits of the corporate form to a broader set of firms. Indeed, establishments that incorporated in Sweden were smaller than their Russian counterparts, and we provide evidence that the low barriers to incorporation enabled marginal firms to incorporate and expand their operations to a scale where adopting steam became feasible.
More broadly, these findings highlight how underlying institutional differences across countries may shape the choice of organizational form and technology use among firms (e.g., Porta et al. Reference Porta, Lopez-de-Silanes, Shleifer and Vishny1998; Kuran Reference Kuran2005; Foreman-Peck and Hannah Reference Foreman-Peck and Hannah2015). Our findings indicate that general incorporation laws that had been enacted in Europe and the United States by the mid-nineteenth century may have facilitated industrial growth partly by enabling firms to leverage the corporate form to grow and adopt new technologies.
Data
Our analysis uses data drawn from the Historical Manufacturing Census of Sweden (Fabriksberättelserna), which contains information on Swedish manufacturing establishments during the latter half of the nineteenth century.Footnote 6 To oversee the industrial development in the country, the Swedish National Board of Trade (Kommerskollegium) tasked local authorities in the cities and the countryside with gathering reports on the performance of all manufacturing establishments in their jurisdiction. Each establishment received a questionnaire every year (see Online Appendix Figure A.1), which had to be returned to the local offices. These questionnaires are now digitized, allowing for an in-depth analysis of Sweden’s industrialization at the establishment level.
Our data provide detailed annual information on Swedish manufacturing establishments, which are defined as specific units of production in a given place. While there is an important distinction between a firm and an establishment (e.g., multiple establishments can be owned by the same firm), the distinction between firms and establishments is less relevant for our study since multi-establishment firms were relatively uncommon during our period of analysis (see Online Appendix Figure A.2). For each establishment, we know the name, location, and industry.Footnote 7 Each establishment also reported the number of workers and its output, measured as the total sales value each year. We use this information to calculate a measure of labor productivity as the value of output divided by the number of workers.
Crucially for our purposes, each establishment also reported whether it used steam, water, or animal power in a given year. We focus on the extensive margin of adoption of whether an establishment used a given power source, as contemporaries argued that the reported information on horsepower is likely measured with significant error (Key-Åberg Reference Key-Åberg, Reuterskiöld, Alin, Blomberg and Pontus1898, pp. 193–94). To have a consistent measure of steam usage over the whole period, we define a dummy taking the value of one when an establishment reported using at least one steam engine. We define analogous dummies for animal and water power to measure the use of different power technologies.Footnote 8
Industrial establishments are only categorized as being owned by an individual (male/female) or a company in our data, so corporations are not specifically delineated. However, as emphasized by Jörberg (Reference Jörberg1961, p. 197), one can reliably identify corporations based on the recorded establishment names on the original questionnaires that establishments submitted. To identify corporations, we search the names of establishments for terms such as Aktiebolag, AB, or Limited, which indicate that an establishment is owned by a corporation.Footnote 9 We complement this approach by using two registers that record all Swedish incorporations between 1848 and 1881 (van der Hagen and Cederschiöld Reference van der Hagen and Cederschiöld1875, Reference van der Hagen and Cederschiöld1882). From these registers, we obtain the year of incorporation, which is matched to the establishments reported in our data. Based on this information, we define a dummy equal to one starting in the first year an establishment is incorporated.
As the raw data do not contain an identifier to track establishments over time, we use automatic record linkage algorithms in the spirit of, for example, Ruggles (Reference Ruggles2002) or Abramitzky, Boustan, and Eriksson (2012) to link establishments into a yearly panel. We briefly outline the underlying procedure here, which is further described in Almås et al. (Reference Almås, Thor Berger, Konrad Burchardi, Björn Eriksson, Hannes Malmberg, Olsson and Ostermeyerforthcoming). We start in the first year of our panel and compare all establishments in one region with all establishments in the same location in the following year in terms of industry as well as establishment and owner names. To assess whether such names are similar, we use Jaro-Winkler and Levenshtein distances.Footnote 10 We link two observations between years if the industry and name(s) of the owner or establishment are very similar, which is the case if the scores exceed our cutoff of 0.9.Footnote 11 Crucially, we keep only unique links and repeat this process across all years. Finally, we manually check the resulting panel to correct incorrect links and identify additional ones that were not automatically generated. We show later that potential biases in the linking approach are unlikely to affect our results.
We complement our establishment-level dataset with regional information on bank density. Specifically, we calculate the number of active banks per establishment in each of the 24 Swedish counties (län), based on data provided by Häggqvist et al. (Reference Häggqvist, Hedberg, Karlsson and Larsson2019), which state the name of each limited and unlimited liability bank operating in Sweden since 1866, as well as their period of existence.Footnote 12 To calculate bank density, we divide the number of active banks by the number of establishments active in these regions over the same period and use the median value as a cut off to identify regions with relatively better access to banking.
Our main sample is an unbalanced panel of approximately 9,000 industrial establishments that we observe for a total of about 66,000 establishment-years between 1864, when virtually all barriers to entry and incorporation had been removed, and 1890, after which electricity became an increasingly viable alternative to steam power. The sample is limited to observations where an establishment is not reported as closed and where the number of workers is recorded.Footnote 13 Table 1 provides summary statistics for the main sample used in the analysis, which includes 8,924 establishments, out of which 996 (or about 11 percent) were at some point owned by a corporation. About two-thirds of these establishments were organized as corporations upon entry, while about a third were observed as partnerships or sole proprietorships prior to converting into corporate form. On average, corporations were considerably larger, exhibited higher labor productivity, and were more likely to be located in urban areas compared to non-incorporated establishments. Notably, incorporated establishments also used steam to a much greater extent.
Table 1 SUMMARY STATISTICS
Notes: This table presents summary statistics for the establishments included in our main sample. Sales (also per worker) is reported in 1,000s SEK and is missing for about 3,500 observations. Note that the reported summary statistics include establishments that were active for only one year, which are, by design, excluded in the panel regressions. See the main text for more details about the data and sample restrictions.
Source: Fabriksberättelserna.
CONTEXT AND DESCRIPTIVE EVIDENCE
The Rise of the Corporation
Sweden’s industrial takeoff in the late nineteenth century occurred against the backdrop of a range of liberal institutional reforms.Footnote 14 Among the most important was the Companies Act of 1848, which codified the limited liability corporation into law. To incorporate, the 1848 law required firms to divide up the firm’s capital into shares, register with the local court, and obtain a charter from the state (Broberg Reference Broberg, Gerald and Hertner2008). While the state retained the right to decline a charter, incorporation was liberally granted to businesses. By the 1860s, the procedure had evolved into a general incorporation system that involved a simple registration process (Nilsson Reference Nilsson1959). Sweden’s transition to a de facto system of general incorporation coincided with a similar transition in other European countries.Footnote 15
The corporation became an increasingly dominant organizational form in Swedish industry over the latter half of the nineteenth century. Figure 1 shows that the share of industrial establishments owned by corporations was about 2 percent in the mid-1860s, which rapidly increased starting with the industrial boom of the 1870s. By 1890, the share of establishments owned by corporations had increased to about 20 percent.Footnote 16 Although most corporations were newly formed, about a third of all establishments owned by corporations had existed in another organizational form as partnerships or sole proprietorships prior to incorporation (see Figure 1). We use these conversions into corporate form as the basis for our main empirical analysis.
Figure 1 THE RISE OF THE CORPORATION
Notes: This figure displays the share of establishments owned by corporations and the yearly number of establishments that enter our data as corporations and convert to the corporate form, respectively. See the main text for more details about the data and sample restrictions.
Source: Fabriksberättelserna.
Why do firms choose to incorporate? Two main advantages have been emphasized. First, incorporation grants owners the right to operate under limited liability so that owners are not liable for claims exceeding their original investments. Second, incorporation facilitates access to debt and equity markets, as shares can be used as collateral to obtain long-term financing.
Historians have emphasized that establishing limited liability for owners was a central reason for incorporation (Broberg Reference Broberg, Gerald and Hertner2008). Its role can be exemplified in the engineering industry, where many small firms built their businesses on exploiting new products or inventions. Without the status of limited liability, few individuals would have been willing to invest their own capital in these risky ventures (Jörberg Reference Jörberg1961, p. 209). Moreover, the role of limited liability is evident from the fact that incorporation often involved no change in the ownership of firms. After incorporation, firms were often closely held, with ownership remaining in the hands of the initial owner or family (Jörberg Reference Jörberg1961, p. 198).Footnote 17
However, contemporaries also emphasized that incorporation facilitated access to capital.Footnote 18 In particular, incorporation enabled firms to use shares as collateral for loans to secure long-term financing for investments.Footnote 19 While lending backed by shares was of minor importance in the 1860s, commercial banks quickly expanded share-backed lending as the rate of incorporation increased in the 1870s and late 1880s (Broberg and Ögren Reference Broberg and Ögren2019). The rise of the corporate form led to increased involvement of banks in industrial development, which had become an increasingly important source of industrial capital by the end of the century (Gårdlund Reference Gårdlund1947; Broberg Reference Broberg, Gerald and Hertner2008; Ögren Reference Ögren2009; Schön Reference Schön2012).
Corporations and the Diffusion of Steam Power
Steam technology diffused slowly in Sweden during the first half of the nineteenth century, but it gradually grew to become the most common source of power by the end of the century. Figure 2A shows that only about 10 percent of industrial establishments reported using steam in the early 1860s. By 1880, about 27 percent of establishments relied on steam power. While direct comparisons should be made with care, these numbers place Swedish industry slightly ahead of American manufacturing, where 22 percent of establishments used steam in the same year (Atack, Bateman, and Margo Reference Atack, Bateman and Margo2008). A decade later, about 41 percent of Swedish industrial establishments had installed at least one steam engine.Footnote 20
Figure 2 THE DIFFUSION OF STEAM POWER
Notes: Panel A: This figure displays the yearly share of establishments using steam, water, and animal power. Note that an establishment can use more than one power source at a time. Other power sources played a negligible role and are thus excluded. Panel B: This figure displays the share of establishments that use steam, water, and animal power across different size classes between 1864 and 1890. See the main text for more details about the data and sample restrictions.
Source: Fabriksberättelserna.
The adoption of steam required establishments to operate at a relatively large scale and often involved a complete reorganization of the factory.Footnote 21 Larger establishments had a cost advantage because they could rely on more cost-efficient and larger engines (Devine Reference Devine1983). Moreover, the productivity gains after adopting steam were greater in large establishments because they enhanced the division of labor (Atack, Bateman, and Margo Reference Atack, Bateman and Margo2008; Atack, Margo, and Rhode Reference Atack, Margo and Rhode2022). On cost grounds, larger establishments thus had clear incentives to prefer steam over water in the latter half of the nineteenth century (Atack Reference Atack1979).
Figure 2B shows that larger establishments, indeed, were substantially more likely to rely on steam, while traditional power sources (animals or water) were less scale-dependent. Steam use was considerably higher among establishments with 15 or more workers, yet the median industrial establishment in our sample employed just five workers.Footnote 22 The fact that most industrial establishments were too small to benefit from adoption has sometimes been cited to explain the slow diffusion of steam in Sweden (Prado Reference Prado2014).Footnote 23
Corporations may have had an advantage in adopting steam given their larger size and the greater scale requirements of steam technology (see Table 1). Indeed, corporations spearheaded the adoption of steam. Industrial establishments owned by corporations were more than three times as likely to rely on steam power compared to establishments owned by partnerships or sole proprietorships (see Table 1).Footnote 24 Similarly, the diffusion of the corporate form and steam power was deeply intertwined across industries. Figure 3 displays the share of establishments that were incorporated as well as the share of establishments using steam between 1864 and 1890 across 12 broad industry groups. Industries where the corporate form was more prevalent—such as the food and beverages, metals, or wood industries—also had adopted steam power to a greater extent.Footnote 25 However, such differences in steam use could also be attributed to the larger establishment size in these industries or the fact that corporations, for example, were larger and more likely to be located in urban areas (see Table 1). We next provide econometric evidence showing that the link between an establishment’s incorporation and adoption of steam likely captures a causal relationship.
Figure 3 INCORPORATION AND STEAM USE ACROSS INDUSTRIES
Notes: This figure displays the share of industrial establishments that report using steam power and that are incorporated across 12 broad industry groups between 1864 and 1890, as well as a best-fit line from a linear regression. The size of the circles corresponds to the mean size of establishments within each industry group, as measured by the log number of workers. See the main text for more details about the data and sample restrictions.
Source: Fabriksberättelserna.
EMPIRICAL ANALYSIS AND RESULTS
This section documents that incorporation was a key driver of the adoption of steam power among Swedish manufacturing establishments in the late nineteenth century. We first examine pooled OLS regressions showing that corporations were more likely to use steam power. To identify the impact of incorporation, we then examine establishments that converted to the corporate form in a difference-in-differences and event-study framework, showing that the probability of an establishment adopting steam increases significantly after incorporation. Exploring the heterogeneous impacts of incorporation, we find that it particularly benefited smaller establishments in rural areas characterized by lower bank density. Lastly, we provide evidence that incorporation allowed such marginal establishments to expand to a scale where steam became a feasible technology to use.
Incorporation and the Adoption of Steam Power
OLS ESTIMATES
We begin by estimating simple pooled OLS regressions of the following form:
$${Y_{it}} = {\rm{\phi} _t} + {\rm{\beta }}Corporatio{n_{it}} + {\boldsymbol{\rm{X}}}{'_{it}}{\rm{\delta\,+\,}}{{\rm{\varepsilon }}_{it}}$$
where Y is a dummy taking the value of one if establishment i used steam power in year t and zero otherwise. Corporation is similarly a dummy variable taking the value of one starting in the year that an establishment is incorporated. The baseline regressions include a full set of year fixed effects (ϕt) that control for the fact that steam use increased over time. In additional specifications, the vector X contains a set of controls capturing other factors that may have shaped the use of steam or industry- and region-by-year fixed effects controlling for time-varying shocks across industries and regions. Throughout the analysis, we cluster standard errors at the establishment level.
Table 2 reports OLS estimates of Equation (1). Column (1) shows that establishments owned by corporations were 42 percentage points more likely to use steam power compared to non-incorporated establishments, which is a large difference considering that only about one in four establishments used steam in the sample. However, the bivariate association between incorporation and steam use may reflect the fact that incorporated establishments, for example, were considerably larger and more likely to be located in urban areas (see Table 1). We next examine the robustness of the link between incorporation and steam use while controlling for such characteristics.
Table 2 CORPORATIONS AND THE USE OF STEAM POWER: OLS ESTIMATES
Notes: OLS regressions. Standard errors clustered at the establishment level are reported in parentheses. See the main text for more details about the data and sample restrictions. *** p < 0.01, ** p < 0.05, * p < 0.1.
Source: Fabriksberättelserna.
A key virtue of the steam engine was that it reduced the locational constraints imposed by water power, which may have facilitated a clustering of establishments in urban locations (Rosenberg and Trajtenberg Reference Rosenberg and Trajtenberg2004; Kim Reference Kim2005). Column (2) adds a dummy taking the value of one if an establishment is located in an urban area. Establishments in urban areas were about 13 percentage points more likely to use steam. However, controlling for an establishment’s geographical location has a marginal impact on the association between incorporation and steam use.
Given that steam power requires establishments to operate at a larger size, the number of workers may be a key predictor of steam use (Atack, Bateman, and Margo Reference Atack, Bateman and Margo2008). Column (3) adds a set of dummies capturing whether establishments employed 6–15, 16–50, and above 50 workers. The probability of using steam increases with establishment size. For example, an establishment employing more than 50 workers was about 57 percentage points more likely to use steam relative to establishments employing five or fewer workers (the omitted reference category). Notably, controlling for establishment size sharply reduces estimated differences in steam use between incorporated and non-incorporated establishments. Thus, a potential mechanism explaining the higher steam use among corporations is that they operate at a larger size.Footnote 26
A range of omitted variables may still explain the association between corporate status and steam use. For example, steam use differed considerably across industries (see Figure 3). To flexibly control for a variety of such industry factors, we add a full set of industry-by-year fixed effects corresponding to our 12 industry groups in Column (4). While the association between incorporation and steam use is reduced, there remains a large positive and statistically significant association. Similarly, spatial differences in, for example, population, fuel prices, or market access may correlate with both the organizational form of establishments and steam use. Therefore, we add a full set of region-by-year fixed effects for the 24 Swedish counties (län) to absorb time-varying regional shocks in Column (5). Notably, the association between incorporation and steam use remains virtually unchanged.
In sum, these results document a statistically and economically significant association between corporate ownership and the use of steam technology. However, interpreting this association as a causal relationship is challenging since it may still reflect a range of unobserved establishment characteristics that we cannot control for. Therefore, we next leverage within-establishment variation in incorporation and steam adoption.
DIFFERENCE-IN-DIFFERENCES ESTIMATES
To identify the impact of incorporation on steam adoption, we use the fact that about a third of all corporations in our sample are observed either as a partnerships or sole proprietorships prior to incorporation. Such conversions provide an opportunity to leverage within-establishment variation in corporate form to isolate the impact on the use of steam technology in a difference-in-differences framework.
Our primary difference-in-differences specification takes the following form:
$${Y_{it}} = {{\rm{\gamma }}_i} + {\rm{\phi} _t} + {\rm{\beta}}Corporatio{n_{it}} + {\boldsymbol {\rm{X}}}{'_{it}}{\rm{\delta\,+\,}}{{\rm{\varepsilon }}_{it}}$$
where Y is a dummy taking the value of one if establishment i uses steam power in year t. Corporation is a dummy taking the value of one starting in the year an establishment incorporates and zero for the years prior. We control for establishment fixed effects (γi), capturing time-invariant and establishment-specific factors, meaning that the identifying variation is driven by establishments that we observe both before and after incorporation. We also include a full set of year fixed effects (ϕt) that absorb differences in steam use across years. In additional specifications, the vector X includes other establishment-level controls as well as industry- and region-by-year fixed effects to absorb temporal shocks that are common to all establishments within the same industry or region. Again, standard errors are clustered at the establishment level.
Table 3 presents estimates of Equation (2). Column (1) shows that incorporation significantly raised the probability that an establishment adopted steam by about 7 percentage points, which is roughly a third of the sample mean. In Columns (2) and (3), we include a full set of industry- and region-by-year fixed effects. While the estimated link between incorporation and adoption of steam is slightly reduced, there remains an economically and statistically significant effect. We provide additional evidence in the Online Appendix that our baseline results are similar when using alternative samples.Footnote 27
Table 3 INCORPORATION AND THE ADOPTION OF STEAM: DID ESTIMATES
Notes: OLS regressions. Standard errors clustered at the establishment level are reported in parentheses. Initial establishment size is measured through dummies denoting whether a given establishment employed 1–5, 6–15, 16–50, or 50+ workers in the first year we observe it. See the main text for more details about the data and sample restrictions. *** p < 0.01, ** p < 0.05, * p < 0.1.
Source: Fabriksberättelserna.
A central empirical concern is that selection into corporate form may drive our results.Footnote 28 While the baseline results control for time-invariant establishment-level characteristics that may affect selection, they do not account for potential selection into incorporation due to time-varying characteristics. A priori, selection into corporate status could be positive or negative. To examine the extent of selection into corporate form, we estimate establishment-level OLS regressions, where we regress employment, output, and labor productivity on a dummy that takes the value of one if an establishment incorporates in the next period (see Online Appendix Table A.4).Footnote 29 These results reveal that establishments that incorporated in the next period were slightly larger in terms of employment and output, though there is no evidence that they had higher labor productivity. While the evidence is not conclusive, event-study estimates in Online Appendix Figure A.9 suggest that incorporating establishments did not experience a differential trend in employment, output, or labor productivity prior to incorporation.
To examine whether selection on establishment size may affect our estimates, we first interact initial establishment size before incorporation, as measured by the number of workers, with a full set of year fixed effects in Column (4) of Table 3. While the estimated magnitude declines somewhat, it still suggests that incorporation was associated with about a 5 percentage point increase in the probability of adopting steam. Another way to address potential selection is to allow for differential growth paths. In Column (5), we include a full set of linear establishment trends, which suggests that incorporation led to about a 4 percentage point increase in steam adoption.
Taken together, the fact that establishments selecting into incorporation were slightly larger prior to incorporation does not seem to account for the fact that the adoption of steam increased after incorporation.Footnote 30 However, to further reduce concerns of selection into incorporation, we next present event-study estimates that also allow us to rely solely on the timing of incorporation among the subset of establishments that, at some point, incorporated for identification.
EVENT-STUDY ESTIMATES
To estimate our event-study models, we follow Sant’Anna and Zhao (Reference Sant’Anna and Zhao2020) and Callaway and Sant’Anna (2021). Their approach involves estimating multiple difference-in-differences regressions to obtain individual treatment effects for groups of establishments that are treated in a specific year. Crucially, only never-incorporated or not-yet-incorporated establishments are used as comparison groups. In the second step, a weighted average is used to summarize the individual effects into an overall effect, with the weights corresponding to the size of the groups. Alternatively, one can also specify a dynamic version to conduct event studies, where the time since treatment is standardized across groups before the corresponding treatment effects are averaged.Footnote 31
Figure 4 displays event-study estimates and 95 percent confidence intervals. Figure 4A compares steam adoption among establishments that incorporate relative to establishments that never incorporate. Prior to incorporation, there were no significant differences in steam use. After incorporation, however, there was a sizable and statistically significant increase in the probability of adopting steam. The average of the group-time average treatment effects on the treated for the five years after incorporation is about 6 percentage points (p-value < 0.01), which is very similar to the difference-in-differences estimates presented in Table 3.
Figure 4 INCORPORATION AND THE ADOPTION OF STEAM: EVENT-STUDY ESTIMATES
Notes: This figure presents event-study estimates of the impact of incorporation on the adoption of steam using the approach developed by Sant’Anna and Zhao (Reference Sant’Anna and Zhao2020) and Callaway and Sant’Anna (2021). We report average treatment effects on the treated and 95 percent confidence intervals and use as control group never treated units in Panel A and not yet treated units in Panel B. See the main text for more details about the data and sample restrictions.
Source: Fabriksberättelserna.
To further reduce concerns that selection drives the results, Figure 4B compares incorporated establishments with those that are not yet incorporated (i.e., the sample is restricted to establishments that become corporations at some point). These estimates identify the effect of incorporation solely from the differential timing of incorporation, which further reduces concerns about selection. Figure 4B shows that there were no significant differences in the use of steam engines in the years prior to incorporation. After incorporation, there was an increase in adoption, although these estimates have less precision, presumably due to the lower number of observations. However, the magnitudes are broadly similar to those in Figure 4A with an average across the group-time average treatment effects on the treated for five years after incorporation of about 4 percentage points (p-value = 0.12).
Taken together, these results show that industrial establishments were not significantly more likely to rely on steam prior to incorporation, though there was a sharp increase in adoption in the years following incorporation. However, these estimates capture the average effects of incorporation and do not clarify the potential heterogeneous impacts of the corporate form.
Incorporation and the Adoption of Steam: The Role of Scale
THE HETEROGENEOUS IMPACTS OF INCORPORATION
Adopting steam involved large capital investments and required many establishments to operate on a greater scale. While incorporating establishments were often large already prior to incorporation, incorporation may have been particularly beneficial for smaller or more credit-constrained establishments if the corporate form facilitated the expansion of an establishment’s scale and investment in new technology. To shed light on this question, we examine the heterogeneous impacts of incorporation.
First, we examine whether the impact of incorporation differed among establishments that varied in terms of size prior to incorporation. To do this, we categorize establishments based on the mean number of workers before incorporation.Footnote 32 Figure 5 presents estimates from Equation (2) separately for initially small (1–15 workers) and large (16 or more workers) establishments. Notably, a positive and significant effect of incorporation is evident among initially small establishments, while there is no significant increase in steam adoption after incorporation among already large establishments.
Figure 5 HETEROGENEOUS IMPACTS OF INCORPORATION: DID ESTIMATES
Notes: This figure reports OLS estimates and 95 percent confidence intervals from nine individual regressions based on Equation (2). The outcome is a dummy taking the value of one if an establishment uses steam power, which is regressed on a dummy taking the value of one if an establishment is organized as a corporation, controlling for establishment and year fixed effects. In each regression, the sample is respectively restricted to: (i) initially small (1–15 workers) and large (16 or more workers) establishments measured as the mean size of establishments in the years before incorporation (note that this corresponds to all years of existence for establishments that never incorporate); (ii) whether an establishment is located in a rural or urban area; (iii) whether an establishment is located in a region with low/high bank density; and (iv) to selected yearly periods (1864–1869, 1870–1879, and 1880–1890). See the main text for more details about the data and sample restrictions.
Source: Fabriksberättelserna.
Second, Figure 5 compares the impact of incorporation among rural and urban establishments. It may have been more challenging for rural establishments to gain access to long-term financing required to adopt steam, or they may have lacked the technical expertise necessary to install, maintain, and use steam engines. While differences are relatively small, the estimated impact of incorporation was larger among establishments located in rural areas.
Third, we examine whether the adoption of steam differed among establishments with varying access to commercial banks. Figure 5 reports estimates for establishments in regions with above and below median bank density, respectively. A stronger association between incorporation and the adoption of steam in areas with lower bank density suggests that incorporation particularly benefited establishments in areas with less developed local capital markets. Relatedly, Figure 5 shows that the link between incorporation and steam adoption is stronger during the 1860s and 1870s, while it declines in magnitude in the 1880s.Footnote 33 A relatively larger effect during early industrialization, when capital markets were less developed, further highlights the potential of the corporate form in overcoming capital constraints.
To bolster the interpretation that incorporation facilitated the adoption of steam among more marginal establishments, we show that incorporation did not lead to the adoption of traditional power sources—animal and water power—that typically did not require firms to operate on a greater scale (Online Appendix Table A.5).Footnote 34 Moreover, the finding that more marginal establishments benefited from incorporation in terms of adopting steam is not evident when examining investments in either animal or water power (see Online Appendix Figure A.11).
Taken together, these findings suggest that incorporation primarily facilitated the adoption of steam technology during early industrialization among initially smaller establishments, often located in rural areas with less developed capital markets. Next, we document that incorporation enabled these small firms to expand to a size where steam became a feasible technology to adopt.
INCORPORATION, ESTABLISHMENT SIZE, AND ADOPTING STEAM
A key barrier to the adoption of steam was that many establishments were required to operate at a larger scale than before. Indeed, the use of steam was considerably more common among establishments that employed 15 or more workers (see Figure 2B). However, the fact that particularly smaller establishments seem to have benefited from incorporation raises the question of whether incorporation enabled these establishments to expand their operations to a scale where it became feasible to adopt steam.
Figure 6 displays the establishment size distribution, before and after incorporation, among establishments that converted into the corporate form. After incorporation, the size distribution shifted to the right for establishments that adopted steam. However, establishments that did not adopt steam after incorporation did not expand, which suggests that the increase in scale and the adoption of steam were closely intertwined.
Figure 6 ESTABLISHMENT SIZE BEFORE AND AFTER INCORPORATION
Notes: This figure shows Kernel density estimates (using an Epanechnikov kernel with a half-width of 0.25) of establishment size measured as the log number of workers for: (i) establishments before incorporation that did not use steam; (ii) establishments that after incorporation used steam; and (iii) establishments that after incorporation did not use steam. We exclude establishments that used steam in all years and those that never incorporated. See the main text for more details about the data and sample restrictions.
Source: Fabriksberättelserna.
To more formally evaluate this mechanism, Table 4, Panel A, presents estimates similar to Equation (2) where the outcome is the number of workers in each establishment. Column (1) shows that an establishment after incorporation saw an increase of about 24 additional workers, which implies that an establishment approximately doubled in size considering that the average establishment employed 25 workers.Footnote 35 Columns (2) and (3) present separate regressions for establishments differentiated by their average size before incorporation, again defined as small (1–15 workers) and large (16 or more workers) establishments.Footnote 36 While the absolute increase in size after incorporation was higher in initially larger establishments, the relative increase was largest in initially small establishments. More importantly, the average small establishment employed just four workers and was therefore too small to use steam technology. However, after incorporation, such establishments approximately doubled in size, thus approaching the size threshold where steam use was substantially higher (see Figure 2B).
Table 4 INCORPORATION, ESTABLISHMENT SIZE, AND STEAM ADOPTION: DID ESTIMATES
Notes: OLS regressions. The independent variable Workers measures the sum of workers employed at the establishment in 100s. Standard errors clustered at the establishment level are reported in parentheses. See the main text for more details about the data and sample restrictions. *** p < 0.01, ** p < 0.05, * p < 0.1.
Source: Fabriksberättelserna.
Can increases in size among initially smaller establishments account for the increased use of steam after incorporation? Table 4, Panel B presents estimates of Equation (2), where the outcome is a dummy capturing whether an establishment uses steam. Columns (4) and (6) show that incorporation is only associated with an increased probability of steam adoption among initially small establishments. In Columns (5) and (7), we include a time-varying control for the number of workers employed. Controlling for size reduces the association between incorporation and steam adoption, both in magnitude and statistical significance among initially small establishments, suggesting that increases in size partly account for the higher probability of adopting steam.
Together, these findings suggest that Sweden’s liberal incorporation system initially benefited smaller establishments by enabling them to expand to a size where adopting steam became feasible. To support this interpretation, we provide case study evidence in Online Appendix B showing that many small firms actively used incorporation to grow and adopt steam power.Footnote 37 We further corroborate this argument by comparing Swedish corporations with the only country for which similar data is available: Imperial Russia (Gregg Reference Gregg2020). Russia lacked a general incorporation law, and firms were required to undergo a costly process to incorporate. Consistent with the lower barriers to incorporation in Sweden, Swedish establishments that incorporated during our period were much smaller compared to Russian establishments that incorporated in the 1890s (Online Appendix Figure A.13).Footnote 38 While suggestive, these size differences further underline the fact that Sweden’s liberal system of incorporation seemingly enabled marginal firms to access the benefits of the corporate form.
CONCLUSIONS
A puzzling fact is that technologies that can substantially increase firm productivity often diffuse at a slow pace. We examine the adoption of steam technology among manufacturing establishments in Sweden during its industrial breakthrough. While few manufacturing establishments used steam around the mid-nineteenth century, it had become the most common source of power by the end of the century. We show that the transition to steam was spearheaded by corporations. By leveraging the fact that hundreds of establishments converted to the corporate form, we document a plausibly causal link between incorporation and the adoption of steam. A key barrier to the diffusion of steam was that most establishments were too small to adopt steam technology. Yet we find that Sweden’s liberal system of incorporation enabled many smaller establishments to leverage the benefits of the corporate form to expand to a greater scale where steam became a feasible technology to adopt. The institutional changes embodied in Sweden’s system of incorporation thus contributed to technological diffusion and ultimately to its catch-up with the leading industrial nations.
Open access
