Abstract
In a simulation experiment, building on the abductive simulation approach of Brenner and Werker (2007), we test historical explanations for why German firms came to surpass British and France firms and to dominate the global synthetic dye industry for three decades before World War 1 while the U.S. never achieved large market share despite large home demand. Murmann and Homburg (J Evol Econ 11(2):177–205, 2001) and Murmann (2003) argued that German firms came to dominate the global industry because of (1) the high initial number of chemists in Germany at the start of the industry in 1857, (2) the high responsiveness of the German university system and (3) the late (1877) introduction of a patent regime in Germany as well as the more narrow construction of this regime compared to Britain, France and the U.S. We test the validity of these three potential explanations with the help of simulation experiments. The experiments show that the 2nd explanation—the high responsiveness of the German university system— is the most compelling one because unlike the other two it is true for virtually all plausible historical settings.
Similar content being viewed by others
References
Allen RC (2003) Progress and poverty in early modern Europe. Econ Hist Rev 56(3):403–443. doi:10.1111/j.1468-0289.2003.00257.x
Arthur WB (1994) Increasing returns and path dependence in the economy. University of Michigan Press, Ann Arbor
Beer JH (1959) The emergence of the German dye industry. University of Illinois Press, Urbana
Brenner T, Duschl M (2014) Modelling firm and market dynamics: a flexible model reproducing existing stylized facts. Working Papers on Innovation and Space, vol 07.14. Philipps-Universität Marburg
Brenner T, Werker C (2007) A taxonomy of inference in simulation models. Comput Econ 30(3):227–244
Buenstorf G, Klepper S (2009) Heritage and agglomeration: the akron tyre cluster revisited. Econ J 119(537):705–733
Caro H (1892) Über die Entwicklung der Theerfarben-Industrie (About the develpment of the coal-tar industry). Ber Dtsch Chem Ges 25(3):955–1105
David PA (1985) Clio and the economics of QWERTY. Am Econ Rev 75(2):332–337
De Vaan M, Boschma R, Frenken K (2013) Clustering and firm performance in project-based industries: the case of the global video game industry, 1972–2007. J Econ Geogr 13(6):965–991
Freeman C (1982) The economics of industrial innovation, 2nd edn. The MIT Press, Cambridge
Heimeriks G, Boschma R (2014) The path-and place-dependent nature of scientific knowledge production in biotech 1986–2008. J Econ Geogr 14(2):339–364
Hofmann AW (1873) Einleitung. In: Amtlicher Katalog der Ausstellung des Deutschen Reiches. Wiener Weltausstellung. Gruppe III: Chemische Industrie (Official Catalog of Exhibit by the German Empire. Vienna World Exhibition. Group III. Chemical Industry). Berlin, pp 95–139
Homburg E, Murmann JP (2014) Global database of synthetic dye firms and plants, 1857–1914. http://professor-murmann.net/dye-database.html. Accessed 28 Oct 2014.
Klepper S (2007) Disagreements, spinoffs, and the evolution of Detroit as the capital of the US automobile industry. Manag Sci 53(4):616–631. doi:10.1287/mnsc.1060.0683
Kopp E (1874) Wiener Weltaustellung 1873 (Vienna World Exhibition 1873). In: Bericht über Gruppe III, Chemische Industrie (Report on Group III, Chemical Industrie), vol 8. C. Baader, Schaffhausen
Krugman P (1995) Development, Geography, and Economic theory. MIT Press, Cambridge
Landes DS (1969) The unbound prometheus: technological change and industrial development in Western Europe from 1750 to the present. Cambridge University Press, New York
Leprieur F, Papon P (1979) Synthetic dyestuffs: the relations between academic chemistry and the chemical industry in nineteenth-century France. Minerva 17(2):197–224
Malerba F, Nelson R, Orsenigo L, Winter S (1999) ‘History-friendly’ models of industry evolution: the computer industry. Ind Corp Chang 8(1):3–40
Malerba F, Nelson R, Orsenigo L, Winter S (2008) Vertical integration and disintegration of computer firms: a history-friendly model of the coevolution of the computer and semiconductor industries. Ind Corp Chang 17(2):197–231
Murmann JP (2003) Knowledge and competitive advantage: the coevolution of firms, technology, and national institutions. Cambridge University Press, New York
Murmann JP (2013) The coevolution of industries and important features of their environments. Organ Sci 24(1):58–78. doi:10.1287/orsc.1110.0718
Murmann JP, Homburg E (2001) Comparing evolutionary dynamics across different national settings: the case of the synthetic dye industry, 1857–1914. J Evol Econ 11(2):177–205
Reader WJ (1970) Imperial chemical industries: a history. Vol. I. Oxford University Press, London
Redlich F (1914) Die volkwirtschaftliche bedeutung der deutschen teerfarbenindustrie. Staats-und sozialwissenschaftliche forschungen 180. Duncker & Humblot, München-Leipzig
Sydow J, Schreyögg G, Koch J (2009) Organizational path dependence: opening the black box. Acad Manag Rev 34(4):689–709
Teltschik W (1992) Geschichte der Deutschen grosschemie: entwicklung und einfluss in staat und gesellschaft (History of the large-scale German chemical industry: development and impact on state and society). VCH, Weinheim
Thissen F (1922) Die Stellung der deutschen Teerfarbenindustrie in der Weltwirtschaft (vor, in, und nach dem Kriege) [The Position of the German Tar Color Industry in the Global Economy before, during, and after the War]. Doctoral Dissertation. University of Giessen
Wenting R, Frenken K (2011) Firm entry and institutional lock-in: an organizational ecology analysis of the global fashion design industry. Ind Corp Chang 20(4):1031–1048
Wurtz A (1876) Progrès de L’industrie des Matières Colorantes Artificielles (The progress of the synthetic dye industry). G. Masson, Paris
Author information
Authors and Affiliations
Corresponding author
Additional information
We would like to thank Peter Bearman, Sendil Ethiraj, Shayne Gary, Steve Klepper, Ray Levitt, Dan Levinthal, Richard Nelson, Charles Tilly, participants of a Seminar at the Max Plank Institute for Economics and participants at the Schumpeter conference for helpful comments on earlier, significantly different versions of this paper. We also like to thank two anonymous reviewers for making suggestions that helped us to improve the paper. We would like to thank in particular Ernst Homburg for helping us locate the historical information that was needed to specify important parameters of the simulation experiment that could not be specified from earlier work by Homburg and one of us (Murmann).
Appendices
Appendix 1: some data on the synthetic Dye industry
Table 8
Appendix 2: some additional description of the simulation model and approach
2.1 A 2.1 Firm strategy
In the model by Brenner and Duschl (2014) each firm is characterized by a strategy variable that ranges from 0 to 1 and determines whether a firm is more an imitator (0) or an innovator (1). In the original model (Brenner and Duschl 2014) this variable is randomly drawn when a firm is founded and remains constant during its existence. From historical records we know that in the dye industry many firms started as imitators and developed into innovators. Therefore, the strategy value is set to 0.01 for each firm at its foundation. Then, after each year it is tested whether the sales of a firm decreased by more than Φ strat,react %. If this is the case, the strategy variable is increased by Φ strat,increase as long as the value of 1 is not reached. The reaction level Φ strat,react is randomly drawn for each firm at its foundation between 0 and 1. The strategy increase Φ strat,increase is randomly drawn for each simulation between 0.05 and 0.25.
2.2 A 2.2 Innovation processes and start-ups
The original model by Brenner and Duschl (2014) considers innovation as the appearance of new market packages that firms compete for, whereby those firms that follow an innovative strategy have a higher probability to win the new markets. We change two aspects of their original modeling: In our model market packages are characterized by two technological variables, namely technological advancement a m and type y m (see Section A 2.3) and firms are more active and determine the appearance of new market packages.
For the detection of a new market an basic innovation rate Φ inno,0 is defined. The probability of a firm to innovate (detect a new market) is given by this rate multiplied by the firms total market share and the free capacity of the firm as well as the R&D activities of the firm (as in Brenner and Duschl 2014). The R&D activities of a firm are defined to depend on the number of chemists employed (see above). If a firm is able to innovate, a new market package appears. The technology type y m and technological advancement a m are randomly drawn from an area that deviates maximally by 1 from the technological advancement and maximally by Φ type,inno from the technology type of the existing products of the firm. A country is randomly assigned to the market package. Before the market package is finally created in the simulation it is checked whether an existing patent forbids the firm to produce this product. Only if this is not the case the market package is created and the innovative firm is the new owner.
We still assume that new firms might enter the industry with innovations (new market packages). To this end, the procedure above is repeated with a potential new firm innovating on the basis of an existing product (market package). The likelihood of such an event is given by
where Φ inno,start-up is a parameter reflecting the probability of new firms entering with innovations in contrast to the innovation probability of incumbents, Φ start-up,chem is a parameter defining the dependence of the start-up probability on the number of unemployed chemists in a country reg (“reg” standing for the countries UK, G, F, S and USA) (drawn randomly from a range between 0 and 1 for each simulation group), and c reg,unempl represents the number of unemployed chemists. This means that we make the number of start-ups dependent on the number of chemists that might found a firm. To determine the technological characteristics of the new market package, one existing market package is drawn randomly and the procedure described above is repeated including the check for patents. If the resulting technology is not covered by an existing patent, a new market package and firm is generated in the simulation model.
2.3 A 2.3 Market space
Dyes differ with respect to many characteristics. The most obvious is their color. Further aspects are the underlying chemical technology, textiles for which it can be used or durability. In our simulation model we simplify the situation by using only one value, y m , which ranges between 0 and 1 to reflect these characteristics. Each market package has a unique value (see above) reflecting the demand for one specific combination of characteristics.
Innovations lead to new values of y m (see above). However, innovations usually do not change all characteristics at once. Hence, to estimate the technological distance, Φ type,inno , that can be bridged by an innovation, we use the following considerations. Historically there are around 5 product classes that differ in their underlying chemical technology (aniline, alizarin, azo dyes, sulfur dyes, and synthetic indigo). Furthermore, seven classes of color (red, orange, yellow, green, blue, violet, black) exist, which leads to 35 classes. Most innovations will appear within these classes. However, within these classes the synthetic dyes still differ in their durability, the textiles for which they can be used, and similar characteristics. In order to restrict our parameter not too much, we assume between 40 and 1000 different technologies leading to 0.001 < Φ type,inno <0.025.
2.4 A 2.4 Market size
The initial (1857) potential market sizes can be estimated as between 530 and 1610 for UK, between 390 and 1270 for Germany, between 310 and 950 for France, between 70 and 210 for Switzerland, between 240 and 720 for the US and between 1400 and 4200 for the rest of the world. We draw random sizes from these ranges for each simulation run. The values reflect the whole demand for dyes. Of course, at the beginning most of this demand is not satisfied by synthetic dye producers (see above). If firms innovate (see above) and an existing market packages that is currently not owned by a synthetic dye firm (but by a natural dye firm, not explicitly modeled here) is technologically reachable, they will not generate a new market package but occupy such an existing market package. Such an innovation is by a factor Φ exist,inno more likely than an innovation leading to a new market package. The total potential market size is assume to increase linearly over time to reach 2300 in UK, 2000 in Germany, 900 in France, 300 in Switzerland, 2600 in US and 8100 in the rest of the world by 1913. Since new market packages appear randomly due to innovation processes by the firms, we have to control the development of the market size. If the market size in the simulation model exceeds the presupposed market size development in a country, market packages are deleted until the values fit again. To this end, for each existing market package in a country the following value is calculated:
Each time the market package with the lowest value disappears. The first term implies that market packages with a lower technological advancement are more likely to disappear. The second term implies that market packages with more technologically similar other market packages are more likely to disappear. Φ dens determines the relative importance of these two aspects. In contrast to the model by Brenner and Duschl (2014), market packages do not disappear randomly in our model but their disappearance is modeled such that the total market size follows the historically known path. This implies that a higher innovative activity (creating more new market packages) of firms increases the probability of the disappearance (in this sence replacement) of existing market packages.
2.5 A 2.5 Chemists
Historical records show that the number of chemists per tons of synthetic dyes sales ranges between 0.0006 and 0.01 (these are the minimum and maximum share of chemists that were employed by the firms Bayer, BASF, Jäger and Levinstein at the end of the period of time studied. For a detailed study of these firms, see Murmann 2003). Therefore, for each firm a random chemist rate Φ chem,rate is drawn from this range when the firm is founded. In addition, the number of chemists also depends on a firm’s innovation strategy s f , because innovation processes require chemical expertise. Thus, the desired number of chemists of a firm f is given by
where T f denotes the firm’s sales.
If a firm cannot employ the desired number of chemists it employs as many chemists as it can get. The real strategy s f,real of a firm depends on the number of chemists that it is able to employ. If this number is lower than the one that matches the intended strategy, the strategy has to be adapted downwards and the firm acts less innovative than intended.
We assume that chemists prefer employment in a domestic dye company. Hence, if firms require additional chemists, they first employ unemployed chemists from the own country. If no unemployed chemists are available in the own country, firms search for chemists in other countries. Only a certain share of all chemists are willing to move to foreign countries. This share is fixed according to historical records as given in Table 9.
If the need of chemists decreases in a firm, chemists are fired and become unemployed.
2.6 A 2.6 Simulation approach
Rights and permissions
About this article
Cite this article
Brenner, T., Murmann, J.P. Using simulation experiments to test historical explanations: the development of the German dye industry 1857-1913. J Evol Econ 26, 907–932 (2016). https://doi.org/10.1007/s00191-015-0430-8
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00191-015-0430-8
Keywords
- Simulation experiment
- Historical development
- Dye industry
- Industrial development
- University education
- Patent law