1
The long-term impact of mergers and the emergence of a merger wave in
pre-World-War I Germany
Gerhard Klinga
a
University of Southampton
THIS IS NOT THE FINAL (POST-REVIEW) VERSION
YOU FIND THE FINAL VERSION HERE:
Kling, G. (2006) The long-term impact of mergers and the emergence of a merger wave in
pre-World War I Germany, Explorations in Economic History 43(4), 667-688
Using annual data on mergers for 35 leading German companies from 1870 to 1913, my study
tries to explain the first merger wave that emerged 1898. My panel probit model that
accounted for economies of scale, macroeconomic conditions, success of former mergers, and
market structure revealed that previous mergers made subsequent mergers more likely. The
propensity to merge was higher for larger companies that increased their market power. In the
banking industry, managers imitated mergers, although these mergers were not successful,
and hence followed the minimax regret principle. Rational information-based herding caused
the serial dependency of mergers in other industries.
Keywords: merger, pre-World-War I, merger wave, minimax regret, herd behavior
2
1. Introduction
During the first phase of globalization that lasted from 1895 to 1914,1 the first merger wave
occurred and changed the industrial structure in Europe and the U.S. remarkably. Famous
companies like U.S. Steel, DuPont, General Electric, and Eastman Kodak are a product of
these early mergers.2 Merger waves are a recurrent phenomenon, for five waves occurred so
far, and the burst of the “New Economy Bubble” ended the last merger mania in 2000. The
emergence of the next merger wave is just a matter of time; hence, it is of great importance to
assess why merger waves started and whether shareholders benefited from them. Merger
waves occurred under different regulatory and institutional frameworks during the last century
– but what were the common driving forces. As the human nature did not change over time,
its limitations concerning rationality might explain this puzzle. My study tries to find answers
to these questions by using historical evidence of mergers from 1870 to 1913 in Germany.
During this period, the new exchange law in 1896 changed the regulatory framework for
German companies and could have triggered the first merger wave.
Economic historians concentrated mostly on debates about the interrelation between the
expansion of large-scale enterprises, external growth, and mergers.3 Maintenance of size and
survivorship were seen as major factors of success. This way of looking at success was
probably motivated by several cross-country studies that pointed out that size was essential.
Especially, large German enterprises were seen as a main guarantee for superior economic
1
Williamson (1996) and Tilly (1999) besides others coined the term `first phase of globalization´; however, the
exact location of this period, which exhibited an increase in international integration, is disputable.
2
3
I refer to Nelson (1959) and Chandler (1990).
Tilly (1982, 1986), Gerschenkron (1962), Huerkamp (1979), and Feldenkirchen (1988) discussed the
concentration process in different lines of business for Germany. Davis (1966), Nelson (1959), Eis (1979),
Hannah (1974), and Chandler (1977) conducted similar studies for the U.S. and UK case.
3
development in the pre-World-War I period.4 After reviewing new statistical material,
however, Kinghorn and Nye (1996) corrected the picture regarding the superiority of German
large-scale enterprises. They found evidence that German firms were smaller compared to
U.S. or French companies, and the concentration process was less developed.5 Besides these
astonishing results, additional doubts emerged regarding the alleged superiority of large
enterprises. Baten (2001a, b, c) showed that small firms exhibited larger total factor
productivity and provided evidence that the median of firm size stayed stable over time. This
contradicted the common opinion that firm size increased between 1895 and 1912.6
Accordingly, mergers that increased firm size were often seen as successful, regardless
whether they reduced share prices: this view changed in the literature. Mergers occurring in
the 1980s and later periods were regarded as successful if they increased shareholder value.
Based on principal-agent theory, Jarrel and Poulsen (1989) stated that managers favor
shareholder value-destroying mergers to increase firm size, as a larger company reduces the
possibility for shareholders to monitor the management effectively. Shleifer and Vishny
(1988) argued that mergers are driven by `empire building´, as managers enjoy controlling a
larger company. The definition of success seems to be crucial, and my study concentrates on
shareholder value as relevant indicator for success. Thus, my empirical model tries to quantify
4
I refer to the most famous contribution made by Chandler (1990). For late Victorian Britain, Elbaum and
Lazonick (1986) argued in favor for large enterprises that had the ability to adopt the modern form of
corporation. They stressed that the high firm size in Germany and the U.S. contributed to countries’ success.
5
Kinghorn and Nye (1996) calculated the number of workers in different industries for Germany, U.S., and
France and showed that German companies were relatively small. Especially, in the iron and steel industry, U.S.
companies employed on average about two times more workers than German counterparts.
6
Due to data availability, his research is restricted to the state `Baden´; however, his empirical evidence can also
be interpreted as a general statement regarding firm size. An exception, however, are some regions in which
large-scale enterprises (iron and steel, mining etc.) predominated the industrial structure.
4
the change in total stock returns due to mergers.7 Generally, the impact of mergers on stock
characteristics was seldom studied for historical periods.8 My estimates of the success of
mergers were used as variable for explaining the emergence of the first merger wave.
Theoretical models for explaining merger waves are still in an `infant stage’. In particular,
theoretical approaches to understand horizontal mergers – typical for the first merger wave –
were seldom developed.9 An exception, however, is Boeckem’s (2002) game theoretical
model – but there is no way to transfer her heterogeneous Cournot game into an empirical
model.10 The driving force for her horizontal merger wave stems from the heterogeneity in
marginal costs. Yet measuring marginal costs is an unsolvable task, especially for the banking
industry that is mainly responsible for the merger wave.
The recent contribution by Rhodes-Kropf and Viswanathan (2004) modeled the positive
interrelation between market valuation and mergers, and Maksimovic and Phillips (2001)
among others confirmed that mergers were more likely in periods of overvalued stocks. My
empirical model accounts for the market valuation of stocks.
Schenk (2001) tried to explain merger waves by a game theoretical model in which
managers apply the minimax-regret decision rule. He argued that a booming economy is the
prerequisite for starting a merger wave. The first merger creates incentives to imitate this
decision and trigger additional mergers irrespective of the success of the first merger. This
7
Total stock returns account for changes in share prices and dividend payments.
8
Exceptions are the studies of Borg et al. (1989), Borg and Leeth (1994, 2000), and Banerjee and Eckard (2001)
that analyzed U.S. data. These event-studies measured the impact of mergers on shareholder value.
9
Boeckem (2002) stated that this is mainly due to the lacking profitability of horizontal mergers in a simple
Cournot setting with homogenous goods if more than two firms interact, as a free-rider problem arises. Hence,
cost asymmetry is necessary to induce profitable mergers in the case of more than two firms (see Salant et al.,
1983, Levin, 1990, and Farrell and Shapiro, 1990).
10
Consider that Boeckem’s (2002) model assumed that the firm with the lowest marginal costs moves first. So
the merger wave is not explained endogenously, which makes empirical proofs difficult.
5
behavior follows the assertion that one prefers to fail conventionally than to succeed
unconventionally. Correspondingly, imitating unsuccessful mergers causes less regret than
overlooking the opportunity for a successful merger. Palacios-Huerta (2003) suggested that
detecting serial dependency of mergers would confirm that managers follow the minimaxregret principle. However, serial dependency of mergers could be due to rational herding
behavior and is not necessarily a hint for the minimax-regret principle. Reputation and
compensation-based herding is not relevant for the pre-1914 period – but information-based
herding apply to this period, as mergers were observable, and imperfect information could be
assumed.11 Banerjee (1992), Bikhchandani et al. (1992) and Welch (1992) developed
theoretical models for information-based herding, and Avery and Zemsky (1998) modified
these models for explaining herding behavior on financial markets. Observing mergers of
other companies can be used to learn something about the private information and signals that
these companies received; hence, it is rational to imitate others’ decisions. When decisionmakers enter such an information cascade, they tend to imitate previous mergers – but they
overlook their own private information and signals, which creates a negative externality. How
can one distinguish between managers that follow the minimax-regret principle and managers
that enter an information cascade? Schenk’s (2001) minimax-regret game predicts that
mergers are imitated even if previous mergers were not successful. As the success of mergers
can be observed, information-based herding does not support imitating unsuccessful mergers.
My paper is organized as follows. Section 2 discusses the method of sampling, which is
oriented towards Baltzer and Kling (2004). Thereafter, a panel vector autoregression (VAR)
reveals the long-lasting impact of mergers on total stock returns. As mergers were not always
11
Information-based herding requires imperfect information, i.e. not everyone can observe specific signals about
the market valuation of a potential target firm. Even if all decision-makers have full access to information,
Simon (1955) pointed out that “bounded rationality” suggests that economic agents do not have unlimited
information processing capabilities. This “lack of brainpower” justifies imitating the decision of others.
6
profitable for shareholders during the whole period, rolling regressions allow time-varying
coefficients. To uncover motives for merging, a panel probit model controls for
macroeconomic conditions, economies of scale, market structure, regulatory changes, success
of previous mergers, and firm specific effects. The decision to merge exhibited serial
dependence; hence, decision-makers imitated the behavior of other companies. Finally, I
discuss whether my empirical findings fit to the minimax regret principle or to informationbased herd behavior.
2. Method of sampling
2.1 Constructing a representative sample for German companies
Sampling in economic history means meticulous work. Therefore, there are two commonly
applied methodologies to draw a sample in historical periods, namely choosing a limited
period of time and collecting all available information during this period, or selecting
companies that fulfill specific criteria, i.e. firm size, to limit the number of cross-sectional
units. For a long-term study, the latter procedure is more appropriate, and hence my sampling
follows Tilly (1982), Weston (1953), and Huerkamp (1979). The former studies by Tilly
(1982) and Huerkamp (1979) focused on surviving companies, and I do the same, albeit I am
aware of a potential survivorship bias. The appendix shows that the survivorship bias is not
relevant for my results.
My aim is to construct a sample consisting of 35 leading companies listed on the Berlin
stock exchange from the early 1870s to the beginning of the First World War in 1914. The
data set contains only acquiring companies, as targets were not observable after mergers.
Restricting the sample to the largest companies ensures that I captured the most active
7
acquirers.12 A simple approach that includes the 35 largest companies would lead to an overrepresentation of banks, and newly developing industries like the chemical industry would be
neglected. Thus, it is essential to construct weights for different industries to guarantee that
infant industries are represented. To obtain weights of industries, I divided all listed
companies into four major sectors. This procedure is in line with the contemporary division
made in `Saling’s Börsen-Papiere´ since the early 1870s, namely banks, mining companies,
transport, and the other-industries-sector.13 Moreover, I refined the other-industries-sector into
seven sub-sectors: breweries, real estate companies, chemical industry, metalworking
industry, mechanical engineering, textile industry, and other minor industries. The insurance
sector exhibited very illiquid trading due to regulations and hence was not considered for my
sample.14 To determine weights for these four main industries, I have to think about criteria
that reflect the importance of the respective category. Paid-in nominal capital could be used as
criterion – but it changed during the 44 years decisively. Most notably, the transport industry
and the other-industries-sector underwent a pronounced development. Due to the
nationalization of nearly all railway companies, the contribution of the transport sector to the
total amount of nominal capital decreases considerably from about 30% to 10%. In contrast,
the more and more diversifying other-industries-sector exhibited an increase from 15% in
1873 to more than 40% in 1912. Table 1 provides an overview with regard to the change in
12
Tilly (1982) argued that the companies’ laws of 1884 and the new exchange law established 1896 favored the
acquisition of smaller companies by larger companies. This assertion can be justified by the legal requirement
that the minimum issue volume had to exceed one million Mark. Hence, a larger company had advantages to
finance acquisitions by issuing new shares.
13
Strictly speaking, Saling (1874-1876) combined the mining sector and the other-industries-sector into one
category. However, considering the overwhelming importance of the mining industry in the pre-World-War I
economy, I prefer to build up an own category for mining companies.
14
Changes of ownership had to be announced and permitted by the board of directors of the respective company.
8
nominal capital of the four main industries. Relying solely on nominal capital as criterion
would not guarantee an appropriate representation of infant industries like the chemical
industry. Accordingly, I follow Tilly (1982), who used the number of listed companies for
determining industry weights. Table 1 shows that the other-industries-sector played the
biggest role in 1873 as well as in 1912, when one considers the number of companies.
(Insert Table 1)
Besides technological progress, the legal environment paved the ground for new
industries. The new law concerning the foundation of companies passed in 1870 caused a real
flood of new companies. During the subsequent period, which is called `Gruenderboom’,
more than 50% of the founded companies belonged to the expanding other-industries-sector.
These companies possessed only a low amount of paid-up nominal capital, as the new law
admitted that investors could pay only a fraction of the `official´ nominal capital. In 1873, the
turning point was reached on the stock market accompanied by decreasing founding activities
in the following years. To capture this wave of newly founded companies from 1870 to 1873,
I fixed the year 1873 as reference year to collect the data for both criteria, namely the number
of listed companies and paid-in nominal capital, from `Saling’s Börsenpapiere (1874-1876)´.
To take the changes during the whole period into account, I fixed the year 1912 as second
reference point, which is close to the end of my investigated period. The `Vierteljahreshefte
zur Statistik des Deutschen Reichs (1913)´ provided data for 1912. Based on these criteria,
my sample consists of 11 banks, 4 mining companies, 5 transport companies, and 15 firms
belonging to the other-industries-sector.
After determining industry weights, the largest companies – based on nominal capital –
in every main industry were selected. These `blue chips´ should represent the most actively
traded stocks on the Berlin stock exchange. `Saling’s Börsen-Jahrbuch (1913/1914)’ and the
`Handbuch der deutschen Aktiengesellschaften (1900/1901 and 1913/1914)’ provided
9
information on mergers, annual share prices, dividends and nominal capital for these 35 `blue
chips´. Table 2 summarizes the merger activity in my sample.
(Insert Table 2)
2.2 Inflation and economic growth rates
Considering that my investigation covers 44 years, time series should be deflated – but which
price deflator should be used. Jacobs and Richter’s (1935) price index was constructed from
wholesale prices, whereas Hoffmann (1965) calculated his private consumption index using a
larger base of time series. Despite the discussible weak points in constructing the data for the
1870s (see Fremdling, 1991), Tilly (1992) used Hoffmann’s (1965) private consumption index
for deflating his indicator for asset returns of the German stock market. Tilly (1992)
concluded that the almost identical course of nominal and real series pointed to the fact that
the development of prices did not dominate the time series. Encouraged by Tilly (1992), I
used Hoffmann’s (1965) private consumption index. A comparison of my real and nominal
data supports Tilly’s (1992) view, which is not surprising because the Mark was tied to the
Gold Standard. Prices increased only by 30% over the whole period, and the average annual
inflation rate reached 0.62%. Nevertheless, inflation rates exhibited tremendous fluctuations
during this period. Besides inflation rates, I also used a time series of economic growth rate
provided by Hoffmann (1965).15
15
Burhop and Wolff (2003) tried to correct for some biases of the Hoffmann (1965) time series. However, using
these corrected time series leads to similar outcomes. My impression is that the differences among alternative
price series are more important in levels than in first differences.
10
2.3 Merger activity during the investigation period from 1870 to 1913
My descriptive finding deviates from Tilly (see p.634, table 1, 1982) in that my data show that
mergers were centered on a peak in 1906. This finding, however, is in line with Huerkamp
(1979), who uncovered a strong increase in merger activities from 1887 to 1907. In contrast to
Tilly (1982) and my study, she also included firms that were not listed on stock exchanges.
Tilly (1982) did not include the banking industry, which accounted for 70% of all mergers in
my sample, and he covered only the period from 1880 to 1913. It is unclear why his sample
size varied from 38 to 49. To take the changing sample size into account, Fig.1 reports the
number of mergers divided by the sample size, which can be regarded as merger intensity.
(Insert Fig.1.)
3. Panel vector autoregression with macroeconomic variables and mergers
Decision-makers could observe the success of former mergers, which could affect their
decisions concerning future expansions. Hence, knowing whether mergers were shareholder
value creating during the pre-1914 period is crucial. For that purpose, a panel VAR analysis is
conduced using first differences in nominal capital and total stock returns as endogenous
variables. Inserting macroeconomic time series controls for their influence on share prices and
dividends, which in turn affected total stock returns.16 As Germany joined the Gold Standard
in the first half of the 1870s, “inflation illusion” discussed by Modigliani and Cohn (1979)
seems to be likely. The “inflation illusion” hypothesis stated that higher inflation rates reduce
16
It is possible to show by applying Granger causality tests that macroeconomic variables can be treated as
exogenous in this setting (see Baltzer and Kling, 2004).
11
real stock returns.17 Besides this alleged influence of inflation rates, one should control for
economic growth rates, as good economic conditions increase earnings, which enables higher
dividend payments and higher share prices.
Event-studies, i.e. Magenheim and Mueller (1988),18 were quite common to investigate
the long-term effect of mergers – but there are two main limitations: first, Morse (1984)
showed that the statistic power of event studies declines when the frequency of data is
reduced, i.e. annual instead of daily stock returns; second, omitted variable bias would turn
inference into a risky venture because event studies cannot control for macroeconomic
shocks. Almost all event studies are based on daily stock returns and measure the market
response triggered by merger announcements. On a daily basis, macroeconomic shocks could
be neglected – but not for measuring the long-term impact of mergers. Thus, an event-study
approach should be replaced by more sophisticated methods that allow to separate different
kinds of shocks, namely mergers and macro-level shocks.
The first step is to test for stationarity, which is a prerequisite for applying a panel VAR
approach. For the sake of increasing the number of observations, unit-root tests can be applied
to panel data. In the literature, two different kinds of tests are common that differ with regard
to the null hypothesis.19 The Hadri (2000) LM test has as null hypothesis that all series are
stationary, whereas the pooled ADF test by Levin et al. (2002) and the procedure by Im et al.
(1997) use the null hypothesis that all series exhibit a unit root and are hence non-stationary.
If both tests point in the same direction, the result is clear. The outcomes for total stock
returns and changes in nominal capital presented in Table 3 are unambiguous. The Levin et al.
17
Baltzer and Kling (2004) confirmed the “inflation illusion” hypothesis for pre-1914 Germany.
18
Armitage (1995) wrote a detailed survey on event studies and their limitations.
19
Ho (2002) stated that ADF and KPSS tests should be used jointly to compare the results.
12
(2002) and the Im et al. (1997) test reject the null hypothesis, and the Hadri (2000) LM test
cannot reject that all series are stationary.20
(Insert Table 3)
The prerequisite of a VAR model is fulfilled; thus, I can estimate a simple VAR model in
reduced form.21
p
Δz it  Σ 0   Σ j Δz it  Ξg t  Mmit  η it
(1)
j 1
The vector Δzit contains the total stock returns returnsit and the change in nominal capital
Δnit, the vector gt captures economic growth rates (gdpt) and inflation rates (inflationt), and
the binary variable mit takes the value one if firm i merged in year t. The lag length p of the
VAR model is determined by information criteria that are calculated based on Hamilton
(1994). The Akaike criterion favors five lags, whereas the Schwarz and the Hannan Quin
criteria prefer a simpler model with three lags. For the sake of simplicity, I choose three lags,
and table 4 summarizes the outcomes of the VAR analysis.
(Insert Table 4)
Generally, macroeconomic variables possessed a considerable impact on stock
characteristics. Mergers increased the expansion of nominal capital, as new shares were
issued to finance mergers. Yet mergers did not affect total stock returns directly – but mergers
triggered higher nominal capital, and higher nominal capital in turn had a negative impact on
total stock returns. Hence, there was an indirect effect on total stock returns – but this effect
was economically small. Putting these outcomes in other words, my model confirms Tilly’s
(1982) finding that mergers played an important role for the expansion of enterprises.
20
For growth rates of net national product and inflation rates, I carry out standard KPSS and ADF tests. Both
time series are stationary.
21
Consider that a VAR has the advantage that total stock returns and nominal capital can influence each other. In
a standard regression model, one has to specify the dependent variable.
13
Interestingly, mergers did not increase total stock returns during the whole period – but
mergers were not shareholder value destroying as found by Borg et al. (1989) for the second
merger wave in the 1920s.
The success of mergers might have changed from 1870 to 1913; hence, I used rolling
regressions for model (1) to determine the partial impact of mergers on total stock returns for
every year of my investigation period.22 Mergers that occurred not in the banking industry
were very successful from 1898 to 1904. This coincides with my descriptive finding that the
merger wave started around 1898. Noteworthy, banks that were the most active acquirers
during this period did not have high benefits from mergers in term of higher total stock
returns for their shareholders. Fig. 2 plots the coefficients of the variable mit for the banking
industry and the rest of the sample. These time-varying coefficients were inserted as
explanatory variable into the probit model that tries to explain the decision to merge.
(Insert Fig.2.)
4. What drives merger during the first phase of globalization in Germany?
4.1 Constructing proxies for market structure and economies of scale
During the first merger wave without effective regulations, mergers could be used to gain
market power; hence, the market structure of the respective industry might play a
predominant role for explaining merger activity.23 However, it is difficult to obtain good
proxies for market structure of industries, as information about market shares (i.e. sales) is not
available on an annual basis and severe forms of collusive behavior, i.e. syndicates and
22
I construct a five year window that is shifted forward after each estimation by one year; thus, the impact of
mergers on total returns can be determined for every year.
23
Stigler (1950) called the first merger wave `merging for monopoly’ and stressed the lack of regulations as a
ground for the merger wave.
14
pooling agreements (`Interessengemeinschaften´) were common – but difficult to measure.24
The relative size measured by nominal capital of a company divided by the 25% percentile
(the smallest companies in the industry) serves as proxy for market structure.25
Jensen (1988) among others stressed the importance of economies of scale for conducting
mergers, for cost advantages related to firm size generate benefits from mergers. When one
focuses on shareholder value, total stock returns of companies relative to total stock returns of
the smallest companies (25% percentile) can be regarded as a proxy for economies of scale.
An alternative approach applied by Spoerer (1996) is to quantify the return to equity, which
requires a thorough balance-sheet analysis. Based on the data provided by Spoerer (1996) and
my stock market data, I calculated correlations between returns to equity and performance
indices for different industries.26 This is, however, a crude proof, as the samples do not match
perfectly. The correlation reached 0.63 for the mining industry and 0.60 for the otherindustry-category. Henceforth, a company with high returns to equity exhibited a high total
stock return – but the relationship indicates that either the stock market was not highly
informationally efficient or the reliability of balance-sheet data for the pre-1914 period was
not sufficient – or both problems existed. Due to the reliability of data, I decided to use the
relative deviation of total stock returns as proxy for economies of scale. Fig. 3 depicts the
24
25
See also Feldenkirchen (1988).
Besides this proxy, I constructed alternative proxies based on the “Vierteljahreshefte zur Statistik des
Deutschen Reichs”. As data before 1895 were not available for all industries, I used self-collected industry data
from Baltzer and Kling (2004) based on “Saling’s Börsen-Papiere, 1874-1876” for 1870 to 1874. With these
initial values, the Holt-Winter (1966) method can replace missing values. For the banking industry, “Der
Deutsche Oekonomist” provided data from 1883 to 1913. These data sources only allow determining the average
firm size in an industry; thus, one can divide individual firm size by the average size to obtain proxies.
Interestingly, these proxies are highly correlated (>0.80) with the proxy based on the 25% percentile that I used
for my empirical models.
26
The indices were value-weighted using market capitalization as weight.
15
economies of scale for the banking industry and all other industries and it shows that both
time series exhibited a strong comovement. Over the whole period, however, the banking
industry had very low economies of scale, as total returns were only 0.96% higher compared
to the smallest companies. In contrast, when one aggregates all other industries and calculates
average economies of scale, these industries outperformed the smallest companies by
17.95%.27
(Insert Fig.3.)
4.2 Model selection of a dynamic panel probit model with random effects
The decision to undertake a merger can be analyzed as count data model because companies
could acquire more than one target within the same year. However, such multiple acquisitions
were seldom; hence, it seems to be appropriate to model the binary decision – “to merge or
not to merge” – of a company. As I deal with panel data, a panel discrete choice model should
be applied. Fortunately, in recent years considerable progress in estimating panel probit and
logit models was made.28 Random or fixed-effects model can account for company specific
effects that are inherent when using panel data. As fixed-effects model do not indicate jointly
significant coefficients and log-likelihood ratio tests reveal the relevance of random effects, I
propose a panel probit model with random effects. Lagged explanatory variables up to lag p
are permitted. The intuition behind the model with lagged variables is that the decision “to
merge or not to merge” may depend on current and past macroeconomic, industry, and
company specific factors as well as behavioral aspects, namely the imitation of previous
27
Total stock returns also reflected other stock characteristics and not just economies of scale; thus, I thought
about using alternative proxies. “Der Deutsche Oekonomist” provided balance-sheet data for the banking
industry, and one can determine total assets. Yet proxies based on total assets are highly correlated with proxies
based on nominal capital (correlation coefficient > 0.9).
28
Hsiao (1992) provided an excellent overview of panel probit models.
16
mergers. Based on this probit model, one can predict mergers and the probability to merge for
different companies and industries.29
p
p
p
k 0
k 0
k 0
m it     βk x t k    k m t k   k B  success jt k  B  D1896  BD1896  ui   it
(2)
The binary variable mit takes the value one if company i executed a merger in year t and zero
otherwise. The column vector xt contains inflation rates (inflationt), growth rates of net
national product (gdpt), economies of scale (scaleit), market structure (structureit), and
previous success of mergers in industry j (successjt). As the success of mergers in the banking
industry deviated strongly from other industries shown in Fig. 2, it is reasonable to allow a
shift in the slope coefficient for banks by inserting the interaction term Bsuccessjt. Schenk
(2001) stressed that good macroeconomic conditions are a prerequisite for starting a merger
wave; hence, economic growth rates are included. As discussed in the previous section on the
long-term success of mergers, inflationary shocks could affect the real valuation of
companies. Modigliani and Cohn (1979) argued that higher inflation trigger a period of real
undervaluation of stocks, as real share prices fall more than real dividends. The valuation of a
company is relevant for the merger process, and it is commonly believed that overvalued
markets make a merger wave more likely.30 When stocks of an acquirer are overvalued, an
29
Granger causality tests indicate that mergers did not influence the explanatory variables used in my panel
probit model. Particularly, mergers did not Granger cause inflation (p-value: 0.655), economic growth (p-value:
0.578), success (p-value: 0.684), the interaction term (banks and new exchange law 1896) (p-value: 0.598), the
number of mergers (p-value: 0.485), economies of scale (p-value: 0.829), and market structure (p-value: 0.963).
The lacking impact of mergers on these explanatory variables shows that there is no causality problem, when I
use explanatory variables without lags (k=0) in my panel probit model.
30
Recently, Rhodes-Kropf and Viswanathan (2004) constructed a theoretical model that explains the emergence
of a merger wave by high market valuations. As standard cointegration tests and panel cointegration tests
rejected a long-term equilibrium of price-dividend ratios, I inserted inflation rates to control for the real valuation
of companies, instead of using price-dividend ratios.
17
acquisition becomes cheaper, as the company can use its own overvalued stocks to finance the
expansion. To test the hypothesis that decision-makers imitated the behavior of other
companies, the number of mergers m t conduced by the 35 companies in the previous year
enters the equation on the right-hand side.
Regulatory changes during my investigation period could affect the probability to merge.
In particular, the new exchange law established in the year 1896 contributed to the creation of
the German universal banking system, which is widely discussed in the current literature.31
Without any doubts, there are some good qualitative arguments why the exchange law should
affect the concentration process especially in the banking industry. For instance, the
restrictions regarding forward dealings on exchanges imposed by the new law created a new
profitable niche for large banks to conduct their own forward trading. Furthermore, the
required minimum issue volume for new share was one million Mark; thus, Tilly (1982)
argued that acquisitions by larger companies were favored. To quantify the impact of the new
exchange law in 1896, I specify a dummy variable D1896 that takes the value one for the
period after the year 1896 and zero otherwise. To test whether the banking industry had a
higher propensity to merge over the whole period, I include a dummy for banks (B). The
interaction term between both dummies (BD1896) controls for alleged special advantages of
banks because of the new regulatory environment.
To determine the lag length of this dynamic model, I carry out the maximum likelihood
estimation of model (2) with different lag specifications. Thereafter, the Akaike and the
Schwarz criteria are calculated together with the log likelihood of the respective model. To
ensure reliable comparisons among different lag structures, the relevant sample is fixed. Table
5 reports the outcomes of model (2) with different lag length p. The Akaike and Schwarz
criteria reach their minimum in the case of the model without lags; thus, lags are not required.
31
By far the most elaborate study on this issue is Fohlin’s (2002) contribution.
18
Moreover, log-likelihood ratio tests indicate that random effects matter, for the null
hypothesis of no random effects can be rejected in all cases.
(Insert Table 5)
My findings point out that macroeconomic conditions and economies of scale did not
affect the probability to merge, in spite of the assertion mentioned by Schenk (2001) that good
macroeconomic surroundings are the prerequisite for starting merger waves. Yet economic
conditions were not bad during the first merger wave – but not mainly responsible for the
increase in merger activity. Rosenberg (1967) among others divided the period before 1914
into a depressive period from 1873 to 1897 characterized by deflationary tendencies and a
period of economic growth from 1898 to 1914. Market structure mattered in that mergers
were more likely if the company was large relative to the smallest companies (25%
percentile); hence, gaining more market power was a strong motive for merging. The previous
success of mergers in the respective industry had a positive impact on merger activity;
however, the banking industry differed in that respect, as the interaction term (Bsuccessjt) is
significant at least on the 10% level of significance. Consequently, banks conducted more
mergers, in spite of the lacking success of former mergers. The new exchange law in 1896
was only relevant for the banking industry, as the interaction term (BD1896) is highly
significant, and the insignificant dummy D1896 does not demonstrate an impact on other
industries. Thereby, the new law supported mergers among banks, while doing little for other
industries. Imitating behavior can be confirmed, for former mergers ( m t ) in the economy
triggered future transactions. Accordingly, one can state that mergers made subsequent
mergers more likely, and hence the first merger wave emerged. This imitating behavior of
decision makers can be due to the minimax-regret principle or to rational information-based
herding. As mergers in the banking industry were more likely even if previous mergers were
not successful, one can state that banks followed the minimax-regret principle. In contrast,
previous mergers increased total stock returns in other industries and were successful. The
19
coefficient of the successjt variable is positive and significant on the 5% level of significance;
consequently, successful mergers made future mergers more likely. The imitating behavior in
other industries could be explained by rational information-based herding because only
successful mergers created incentives to imitate.
To illustrate the probit model, I calculated one-step-ahead forecasts regarding the
probability that a firm will merge in the next year. The model provides for every year and
every company forecasted probabilities. Fig. 4 depicts the average probability of the whole
sample for the respective year based on the model without lags shown in table 5. To highlight
whether other economic factors, i.e. market structure, justified imitating behavior, I derived
forecasts without imitation by excluding the number of previous mergers ( m t ) as explanatory
variable. By comparing these two series, one can evaluate whether imitating the move of
others was a good strategy in that other economic factors like market structure and the success
of previous mergers stimulated additional mergers in future. Since 1898 – two years after the
new exchange law was introduced – to 1904, rational economic factors caused a high
probability to merge. The model with imitating behavior predicted lower probabilities than the
model without imitation. During the starting phase of the merge wave imitating other mergers
was justified by economic factors (i.e. mergers increased total stock returns) – but from 1904
to 1913 imitating behavior led to mergers not justified by economic surroundings. Fig. 2
shows that after 1906 mergers were no longer profitable for shareholders, and imitating
yielded shareholder value destroying mergers. Accordingly, in the beginning of a merger
wave like in the case of other bubbles and manias on stock markets `riding the wave’ by
imitating others was profitable. This finding is in line with the literature on rational herd
behavior, as decision-makers that imitate others and neglect their own information create a
negative externality. In the end of such an information cascade, the information content of
previous mergers declines rapidly; thus, imitation might lead to bad decisions.
(Insert Fig.4.)
20
5. Conclusion
Based on my descriptive findings, I confirm the emergence of the first merger wave in
Germany that started around 1898 accompanied by the introduction of the new exchange law
in 1896. The panel VAR model uncovers that mergers were not successful over the whole
period, albeit periods of successful mergers can be identified applying rolling regressions.
From 1898 to 1904, mergers affected total stock returns positively – but shareholders of banks
did not benefit from mergers. This period coincided with the beginning of the merger wave
and points to the fact that banks merged, in spite of generating losses for their shareholders.
The panel probit model reveals that mergers were more likely when the company was large
compared to the smallest companies (25% percentile) in the industry; hence, mergers could be
used to increase market power further. Interestingly, the previous success of mergers had a
positive influence on merger activity – except for the banking industry. This means that even
if previous mergers were not successful, banks tended to increase their merger activity.
As the number of previously executed mergers increased the probability for merging in
future, serial dependency between merger decisions can be confirmed. Palacios-Huerta (2003)
argued that this is already a confirmation that decision-makers follow a minimax-regret
principle – but serial dependency might be caused by rational information-based herd
behavior.32 As the success of previous mergers could be observed by decision-makers,
imitating unsuccessful mergers cannot be justified by rational information-based herding – but
is an indication that Schenk’s (2001) minimax-regret principle can explain this firm behavior.
Correspondingly, one can distinguish between the imitating behavior in the banking industry
and in other industries. Banks tended to imitate unsuccessful mergers, and hence bank
managers followed the minimax-regret principle because they valued potential failures less
32
I refer to Banerjee (1992), Bikhchandani et al. (1992), Welch (1992) and Avery and Zemsky (1998).
21
than the regret to overlook attractive mergers. Contrarily, in the case of the three other main
industries mergers were more likely if former mergers increased total stock returns;
consequently, rational information-based herd behavior can explain the behavior of these
companies. Noteworthy, the introduction of the new exchange law in 1896 only increased the
probability of merging in the banking industry and contributed to the emergence of the merger
wave of banks.
This historical example underlines the importance of imitating behavior motivated by the
minimax-regret principle, information-based herding and bounded rationality for the
emergence of the first merger wave. Institutions and laws might change – but the human
nature remains the same. Henceforth, the next merger mania will occur with certainty.
22
References
Armitage, S., 1995, Event study methods and evidence on their performance, Journal of
Economic Surveys 9, 25-52.
Avery, C. and P. Zemsky, 1998, Multidimensional uncertainty and herd behavior in financial
markets, American Economic Review 88, 724-748.
Baltzer, M. and G. Kling, 2004, Resiliency of the pre-World War I German stock exchange:
Evidence from a panel vector autoregression, conference proceedings of the 5th World
Congress of Cliometrics 2004, 169-180.
Banerjee, A., 1992, A simple model of herd behavior, Quarterly Journal of Economics 107,
992-1026.
Banerjee, A. and E.W. Eckard, 2001, Why regulate insider trading? Evidence from the first
Great Merger Wave (1896-1903), The American Economic Review 91, 1329-1349.
Baten, J., 2001a, Produktivitätsvorteil in kleinen und mittelgroßen Industrieunternehmen,
Sicherheit in Großunternehmen? Die Gesamtfaktorproduktivität um 1900, Tübinger
Diskussionsbeitrag 217.
Baten, J., 2001b, Große und kleine Unternehmen in der Krise von 1900-1902, Tübinger
Diskussionsbeitrag 216.
Baten, J., 2001c, Expansion und Überleben von Unternehmen in der `Ersten Phase der
Globalisierung´, Tübinger Diskussionsbeitrag 215.
Bikhchandani, S., Hirshleifer, D. and I. Welch, 1998, A theory of fads, fashion, custom, and
cultural change as informational cascades, Journal of Political Economy 100, 992-1026.
Boeckem, S., 2002, Merger mania or after you - a theory of endogenous merger waves,
Working Paper, Universität Dortmund.
Borg, J.R. and Leeth, J.D., 2000, The impact of takeovers on shareholder wealth during the
1920s merger wave, Journal of Financial and Quantitative Analysis 35, 217-238.
23
Borg, J.R. and Leeth, J.D., 1994, The impact of mergers on acquiring firm shareholder
wealth: The 1905-1930 experience, Empirica 21, 221-244.
Borg, J.R., Borg, M.O. and Leeth, J.D., 1989, The success of mergers in the 1920s: A stock
market appraisal of the second merger wave, International Journal of Industrial
Organization 7, 117-131.
Burhop, C. and G. Wolff, 2003, A compromise estimate of the net national product and the
business cycle in Germany, 1851 – 1913, conference proceedings of the 5th World
Congress of Cliometrics 2004, 97-108.
Chandler, A.D., 1990, Scale and scope: The dynamics of industrial capitalism, Cambridge.
Chandler, A.D., 1977, The visible hand, Cambridge.
Davis, L., 1966, The capital markets and industrial concentration; the US and the UK, a
comparative study, Economic History Review 19, 255-272.
Elbaum, B. and W. Lazonick, 1986, An institutional perspective on the British decline, in:
The decline of the British Economy, Elbaum, B. and W. Lazonick (ed.), Oxford.
Eis, C., 1971, The 1919-1930 merger movement in American industry, Arno Press, New
York.
Farrell, J. and C. Shapiro, 1990, Horizontal mergers: An equilibrium analysis, American
Economic Review 80, 107-126.
Feldenkirchen, W., 1988, Concentration in German industry, 1870-1939, in: The
concentration process in the entrepreneurial economy since the late 19th century, Pohl, H.
(ed.), Steiner-Verlag, Wiesbaden.
Fohlin, C., 2002, Regulation, taxation and the development of the German universal banking
system, 1884-1913, European Review of Economic History 6, 221-254.
Fremdling, R., 1991, Productivity comparison between Great Britain and Germany, 18551913, Scandinavian Economic History Review 39, 28-42.
Gerschenkron, A., 1962, Economic backwardness in historical perspective, Cambridge.
24
Hadri, K., 2000, Testing for stationarity in heterogeneous panel data, The Econometrics
Journal 3, 148-161.
Hamilton, J.D., 1994, Time series analysis, Princeton, Princeton University Press.
Hannah, L., 1974, Mergers in British manufacturing industry, 1880-1918, Oxford Economic
Papers 26, 1-20.
Ho, T., 2002, A panel cointegration approach to the investment-saving correlation, Empirical
Economics 27, 91-100.
Hoffmann, W.G., 1965, Das Wachstum der deutschen Wirtschaft seit der Mitte des 19.
Jahrhunderts, Berlin.
Hsiao, C., 1992, The econometrics of panel data: Handbook of theory and applications,
Advanced Studies in Theoretical and Applied Econometrics, vol. 28. Norwell, Mass.
and Dordrecht: Kluwer Academic.
Huerkamp C., 1979, Fusionen der 100 größten Unternehmen von 1907 zwischen 1887 und
1907, in: Horn and Kocka (ed.), Recht und Entwicklung der Grossunternehmen,
Goettingen.
Im, K., Pesaran, M.H. and Y. Shin, 1997, Testing for unit roots in heterogeneous panels.
unpublished working paper, Department of Applied Economics, University of
Cambridge.
Jacobs, A. and H. Richter, 1935, Die Großhandelspreise in Deutschland von 1792 bis 1934,
Hamburg.
Jarrel, G.A. and A.B. Poulsen, 1989, The returns to acquiring firms in tender offers:
Evidence from three decades, Financial Management 18, 12-19.
Jensen, M.C., 1988, Takeovers: Their causes and consequences, Journal of Economic
Perspective 2, 21-48.
25
Kinghorn, J.R. and J.V. Nye, 1996, The scale of production in western economic
development: A comparison of official industry statistics in the United States, Britain,
France, and Germany, 1905-1913, Journal of Economic History 56, 90-112.
Kocka, J. and H. Siegrist, 1979, Die hundert größten Industrieunternehmen im späten 19. und
frühen 20.Jahrhundert. Expansion, Diversifikation und Integration im internationalen
Vegleich, in: Recht und Entwicklung der Großunternehmen im 19. und frühen
20.Jahrhundert, Horn, N. and J. Kocka (ed.), Vandenhoeck&Ruprecht, Göttingen.
Levin, A., Lin, C. and C. J. Chu, 2002, Unit Root Tests in Panel Data: Asymptotic and Finite
Sample Properties, Journal of Econometrics 108, 1-24.
Levin, D., 1990, Horizontal mergers: The 50-percent benchmark, American Economic
Review 80, 1238-1245.
Magenheim, E.B. and Mueller, D.C., 1988, On measuring the effect of acquisitions on
acquiring firm shareholders’, in Coffee, J.C., Lowenstein, L. and S. Rose-Ackerman
(eds.), Knights, raiders, and targets: The impact of the hostile takeover, New York,
NY, Oxford University Press.
Maksimovic, V. and G. Phillips, 2001, The market for corporate assets: Who engages in
mergers and asset sales and are there efficiency gains? Journal of Finance 56, 2019-2065.
McFadden, D., 1974, The measurement of urban travel demand, In Zarembka, P. (ed.),
Frontiers in econometrics, New York, Academic Press.
Modigliani, F. and R.A. Cohn, 1979, Inflation, rational valuation and the market, Financial
Analysts Journal 35, 24-44.
Morse, D., 1984, An econometric analysis of the choice of daily versus monthly returns in
tests of information content, Journal of Accounting Research 22, 605-623.
Nelson, R.L., 1959, Merger Movements in American Industry 1895-1956, Princeton
University Press; Princeton, New Jersey.
26
Palacios-Huerta, I., 2003, Professionals play minimax, Review of Economic Studies 70,
395-415.
Rhodes-Kropf, M. and S. Viswanathan, 2004, Market valuation and merger waves, Journal of
Finance 59, 2685-2718.
Rosenberg, H., 1967, Große Depression und Bismarckzeit: Wirtschaftsablauf, Gesellschaft
und Politik in Mitteleuropa, Berlin.
Salant, S., Switzer, S. and R. Reynolds, 1983, Losses from horizontal mergers: The effect of
an exogenous change in industry structure on Cournot-Nash equilibrium, Quarterly
Journal of Economics 98, 185-199.
Schenk, H., 2001, Mergers and the economy: Theory and policy implications, presented at the
workshop
entitled
`European
Integration,
Financial
Systems
and
Corporate
Performance´, UN University, Institute for New Technologies Kasteel Vaeshartelt.
Shleifer, A. and R.W. Vishny, 1988, Value maximization and the acquisition process, Journal
of Economic Perspective 2, 7-20.
Simon, H.A., 1955, A behavioral model of rational choice, Quarterly Journal of Economics
69, 99-118.
Spoerer, M., 1996, Von Scheingewinnen zum Ruestungsboom: Die Eigenkapitalrentabilitaet
der deutschen Industrieaktiengesellschaften 1925-1941, Stuttgart.
Stigler, G. J., 1950, Monopoly and oligopoly by merger, The American Economic Review
40, 23-34.
Tilly, R., 1999, Globalisierung aus historischer Sicht und das Lernen aus der Geschichte,
Köln.
Tilly, R., 1992, Der deutsche Kapitalmarkt und die Auslandsinvestitionen von 1870
bis
1913, in: ifo Studien 38, 200-225.
Tilly, R., 1986, German banking, 1850-1914. Development assistance for the strong, Journal
of European Economic History 15, 113-152.
27
Tilly, R., 1982, Mergers, external growth, and finance in the development of large-scale
enterprise in Germany, 1880-1913, Journal of Economic History 42, 629- 658.
Welch, I., 1992, Sequential sales, learning, and cascades, Journal of Finance 47, 695-732.
Weston, J. F., 1953, The role of mergers in the growth of large firms, Berkeley.
Williamson, J. G., 1996, Globalization, convergence and history, Journal of Economic
History 56, 277-306.
Williamson, J. G., 1996, Globalization, convergence and history, Journal of Economic
History 56, 277-306.
Winters, P. R., 1966, Forecasting sales by exponentially weighted moving averages,
Management Science 6, 324-342.
Data sources
Der Deutsche Oekonomist, 1883-1913, Berlin.
Handbuch der deutschen Aktien-Gesellschaften, issue 1900/1901 and 1913/1914, volume I
and II.
Kaiserliches Statistisches Amt, 1911/1912, Vierteljahreshefte zur Statistik des Deutschen
Reichs – Ergänzungsheft zu 1913, II, Berlin.
Saling’s Börsen-Jahrbuch, 1913/1914, Ein Handbuch für Kapitalisten und Effektenbesitzer.
2nd ed., Berlin, Leipzig.
Saling’s Börsen-Papiere, 1874-6, vol. 1-5, 4th ed., Berlin.
Vierteljahreshefte zur Statistik des Deutschen Reichs, 1892-1912, Berlin.