Reevaluating the conglomerate discount in Germany: the role of design choices

Advocates of diversification state that conglomerates exhibit lower firm risk because they combine segments with imperfectly correlated earnings streams. Founders or founding families in particular benefit from this firm’s risk reduction because they typically possess a relatively undiversified personal portfolio (Anderson and Reeb 2003). In addition, diversification reduces the default probability, thereby increasing the market value of debt (Ammann et al. 2012). This coinsurance effect increases the firms’ debt capacity and creates value through two channels. First, it enables conglomerates to increase leverage and hence the interest tax shield. Second, the increased debt capacity enables conglomerates to make more investments than comparable focused firms (Berger and Ofek 1995; Lewellen 1971).

Furthermore, conglomerates can not only make more investments but also allocate capital more efficiently within firms. This is due to the creation of internal capital markets (“bright side of capital”) that enable segments with high cash flow and poor investment opportunities to finance segments with less cash flow but better investment opportunities (“winner picking”). Consequently, conglomerates can make more value-increasing investments than their individual segments would be capable of making independently (Stein 1997; Stulz 1990; Weston 1970). Moreover, internal markets also allow for more efficient allocation of other resources, including human capital (Lang and Stulz 1994).

Finally, conglomerates are considered more efficient due to synergies and economies of scope. By exploiting firm-specific assets across segments, conglomerates can become more efficient and more profitable than focused firms (Berger and Ofek 1995; Chandler 1977; Weston 1970).

There are also several arguments that suggest conglomerates trade at a discount. Although economies of scope are expected to increase efficiency, they entail costs that may reduce or even reverse the benefits of synergies.

Opponents of diversification also emphasize the dark side of internal capital markets, which describes their inherent inefficiencies. Within internal capital markets, divisional managers can exert influence to increase assets under their control. This practice can result in the subsidization of less profitable segments at the expense of more profitable ones (Rajan et al. 2000; Scharfstein and Stein 2000; Stulz 1990). Moreover, agency problems and managers’ rent-seeking tendencies can not only lead to inefficient cross-subsidization but also induce firms to retain or pursue a value-decreasing diversification strategy. Managers derive private benefits from diversification, as diversification increases the value of their relatively undiversified personal portfolio (Jensen and Murphy 1990), causes them to be indispensable to the firm (Shleifer and Vishny 1989), and allows them to exploit the firm for their own purposes (Jensen 1986; Purkayastha et al. 2022; Stulz 1990). Moreover, managing a larger firm is associated with more power, prestige, and compensation (Jensen 1986; Jensen and Murphy 1990; Stulz 1990).

Another explanation for a conglomerate discount is based on different assessments made by investors and analysts compared to focused firms. Given that conglomerates’ accounting data are less transparent, they are more difficult to evaluate, impairing the accuracy of analysts’ forecasts (Bushman et al. 2004; Feldman 2016; Gilson et al. 2001).

Although Germany is among the largest economies in the world, most studies on the conglomerate discount rely on U.S. samples. Prior literature on the conglomerate discount in Germany is sparse, marked by inconclusive findings, and potentially outdated.¹ As institutional differences across countries are found to affect the conglomerate discount (e.g., Fauver et al. 2003; Lee et al. 2008; Lins and Servaes 1999; Rudolph and Schwetzler 2013; Weiner 2005), prior evidence from the U.S. is not necessarily applicable to the German market.

The earliest empirical study examining the effect of diversification on firm value in Germany comes from Lins and Servaes (1999). The authors analyze a sample of firms from Germany, Japan, and the UK between 1992 and 1994. While they observe an average discount of approximately 10% in Japan and approximately 15% in the UK, no significant valuation difference can be identified for German firms. The working paper by Schwetzler and Reimund (2003) confirms that German conglomerates are not discounted. They examine German firms between 1998 and 2001 and find insignificant effects of diversification on market valuation. However, they argue that prior research did not adequately account for cash holdings and find weak evidence of a conglomerate discount using an adjusted measure of market value.

Fauver et al. (2003) analyze 35 countries, including Germany, during the period from 1991 to 1995. Their results suggest that both the degree of development of the capital markets and the legal and regulatory environment affect the valuation of conglomerates. While Fauver et al. (2003) find evidence for a conglomerate discount, which varies across countries, in the case of German firms, this discount transformed into a premium of 2–10.7%.

Although these results diverge from U.S.-based studies (Berger and Ofek 1995; Mansi and Reeb 2002; Sturm and Nüesch 2019), the conglomerate discount remained relatively underexplored in the subsequent years. Univariate results in the discussion paper by Weiner (2005) suggest that German conglomerates are traded at a discount of approximately 3–10% and Beckmann (2006) finds a conglomerate discount, which increases with the number of unrelated segments.

The most recent paper on the conglomerate discount in Germany is the study by Glaser and Müller (2010). Building on the work of Mansi and Reeb (2002) and using a sample of 4070 firm-years between 2000 and 2006, they analyze whether the conglomerate discount is caused by the book value bias of debt. The valuation differences between focused and diversified firms are usually analyzed by using excess values (Berger and Ofek 1995). However, these excess values rely on the book value of debt, which does not capture the enhanced bondholder value resulting from risk reduction. In their initial analysis, they document a conglomerate discount, which ranges from 7.7 to 13.9%. This discount decreases once the market value of debt is employed instead of the book value of debt and ranges from 6.7 to 8.2%.

To our knowledge, only two dissertations have analyzed the conglomerate discounts since the publication of the study by Glaser and Müller (2010). While Kluge (2014) identifies a conglomerate discount in the period from 2004 to 2010, Liu’s (2016) results indicate a conglomerate premium during the period from 2005 to 2014.²

Our sample consists of listed German firms between 2000 and 2019. We obtained data from Datastream. The sample period starts in 2000, as German firms have been required to disclose reliable business segment data starting in 2000. We do not consider years after 2019 due to the impact of the COVID-19 pandemic. Our initial sample consists of 13,207 firm-years among 1180 unique firms.

Following prior research, we exclude 2928 firm-years from the financial sector (i.e., SIC 6000–6999), as our valuation method requires variables that are often not reported by financial firms. Missing financial data that are necessary to calculate control variables restrict our sample to 9935 firm-years among 894 firms (Panel A). We use three different proxies for market valuation that require additional financial data and are subject to further exclusion criteria as described in Sect. 3.3 (EV_Sales, EV_Merton, and EV_Goodwill). Thus, our final sample ranges from 4455 to 5630 firm-years depending on the measure of market value.

Various approaches can be employed to operationalize conglomerates and the degree of diversification. Prior literature on diversification usually utilizes a binary variable that takes a value of 1 when the firm is a conglomerate and 0 when the firm is focused (Campa and Kedia 2002; Chang et al. 2016; Mansi and Reeb 2002). We follow Glaser and Müller (2010) and classify firms as focused that have (1) only one operating segment or (2) more than one operating segment but all operate in the same two-digit SIC industry or (3) no business segment information was published. Firms are categorized as (3) if there is no information available on segment assets or segment sales or no specific segment descriptions in the database. Segments are treated as nonoperating segments if the segment description contains strings that indicate that the segment is nonoperating (i.e., holding, central division, or corporate), the segment SIC is 9999 (nonclassifiable establishment), or segment assets or sales are negative or zero because such segments can be regarded as adjustment segments. Thus, our measure of diversification indicates whether a firm is unrelated diversified or not.

We employ different types of excess values as proxies for market valuation. A firm’s excess value is calculated as the natural logarithm of the ratio of a firm’s actual value to its imputed value. A positive excess value suggests that the firm trades at a premium (i.e., the conglomerate’s actual value is higher than its imputed value if each of its segments operated as a single-firm segment), while a negative excess value implies that the firm trades at a discount (i.e., the conglomerate’s actual value is lower than its imputed value).

Our primary measure of market value, EV_Sales, is the traditional Berger and Ofek (1995) excess value. The firm’s actual value is the sum of the market value of equity and the book value of debt. We calculate the imputed value based on sales multiples, where the imputed value is the sum of the imputed values of its segments; each segment’s imputed value is equal to the segment’s sales multiplied by its industry median ratio of total capital to sales of focused firms. The industry median ratios are based on a 2-digit SIC grouping that includes at least five focused firms. Excess values are also calculated for focused firms. By construction, the median excess value of focused firms is zero. For some firms, the sum of segments’ sales and the firm’s sales differ. Following prior research, we exclude conglomerates whose segment sales deviate by more than 5% (Ammann et al. 2012; Hoechle et al. 2012). The segment sales are adjusted up or down if the deviation is less than 5%. Finally, we exclude extreme values, i.e., actual values that are either more than four times the imputed value (> 1.386) or less than one-fourth of the imputed value (< − 1.386).

Additionally, we calculate EV_Merton and EV_Goodwill to account for two common biases in the diversification discount literature. Glaser and Müller (2010) show that measures of firm values based on book values of debt systematically undervalue conglomerates. Consistent with Eberhart’s (2005) application of the Merton (1974) model, we calculate the market value of debt to account for the fact that diversification enhances bondholder value due to a reduction in firm risk. Furthermore, Custódio (2014) argues that assets are typically reported at their transaction-implied value which often exceeds the target’s pre-merger book value resulting in lower market-to-book ratios. To mitigate this measurement bias, we subtract goodwill from the book value of assets in measuring the firm value.³

To investigate the association between diversification and market value, we replicate the empirical design in Glaser and Müller (2010):

We employ measures of excess value based on Berger and Ofek (1995) (EV_Sales), Glaser and Müller (2010) (EV_Merton), and Custódio (2014) (EV_Goodwill) as proxies for firms’ market value. diversified firm (dummy) is a binary variable that takes a value of 1 when the firm is diversified and zero when the firm is focused. Consistent with Glaser and Müller (2010), we control for firm size, profitability, capital expenditures, and accounting standards.⁴ Appendix 2 provides the definitions of all variables along with their Datastream identifier.

Because our sample includes heterogeneous firms, which differ in size and thus cause heteroskedasticity, we use robust standard errors. We also employ year fixed effects to control for time effects influencing the diversification discount, which have been documented in prior literature (Berger and Ofek 1995; Chang et al. 2016; Denis et al. 2002). We do not employ industry fixed effects because excess values reflect a firm’s value relative to the median in an industry and are thus almost analogous to an industry fixed effects estimator (Campa and Kedia 2002).⁵

Table 1 provides descriptive statistics and univariate results for our sample. The means of EV_Sales, EV_Merton, and EV_Goodwill are − 0.104, − 0.060, and − 0.103, respectively. Consistent with the existence of a conglomerate discount, t-tests of means suggest that conglomerates have a lower EV_Sales (difference = − 0.119, p < 0.01), EV_Merton (difference = − 0.065, p < 0.01), and EV_Goodwill (difference = − 0.125, p < 0.01). Furthermore, univariate results suggest that conglomerates hold more assets, generate more operating income, and have fewer capital expenditures.

Table 2 presents correlations between the variables in our models. We find significant negative Spearman and Pearson correlations between diversified and market value (EV_Sales, EV_Merton, and EV_Goodwill) that could indicate the existence of a conglomerate discount. In addition, we find significant correlations between explanatory variables. As these correlations are low and mean variance inflation factors range from 2.15 to 2.18, multicollinearity is not a problem in our models.

Although t-tests of means and correlations indicate that conglomerates are on average traded at a discount, we find 37.21% of the conglomerates in our sample to have an average sales-based excess value above zero (i.e., to be traded at a premium).⁶ One potential reason is that the benefits and costs of diversification can differ among firms (Bushman et al. 2004; Glaser et al. 2013). Consistent with this argument, we find that conglomerates traded at a premium are more likely to benefit from increased debt capacity by having higher leverage (difference = 0.016, p < 0.01). While conglomerates are expected to benefit from better investment opportunities through the creation of internal capital markets, the inability to increase leverage may inhibit the exploitation of these opportunities. Accordingly, conglomerates traded at a premium invest more in R&D relative to sales (difference = 0.027, p < 0.01) and have higher capital expenditures relative to sales (difference = 0.017, p < 0.01). We also find that conglomerates traded at a premium are more efficient (e.g., due to synergies and economies of scope), as evident in higher performance in terms of EBIT to sales (difference = 0.024, p = 0.068), return on assets (difference = 0.042, p < 0.01), growth of sales (difference = 0.081, p < 0.01), and growth of assets (difference = 0.069, p < 0.01).⁷

Table 3 presents ordinary least squares regressions of diversification on market value. In line with prior research, we find conglomerates to be valuated at a discount (Berger and Ofek 1995; Glaser and Müller 2010; Sturm and Nüesch 2019). The coefficient on diversification is negative and significant at the 1% level in each regression. In particular, there is a discount of 11.5% in model (1). Consistent with prior evidence from the U.S. (e.g., Custódio 2014), the conglomerate discount decreases to a range of 7.9% and 11.4% when we account for debt value and goodwill measurement biases. This finding underscores the importance of accounting for measurement biases when calculating excess values, because neglecting these biases likely leads to an overestimation of the conglomerate discount.⁸ The direction of all other associations with our control variables is consistent with prior literature.⁹

Our results contradict the findings of two studies on the German market. While Lins and Servaes (1999) identify no effect of diversification on market valuation in 1992 and 1994, Fauver et al. (2003) find evidence for a conglomerate premium between 1991 and 1995. On the one hand, German firms have been required to disclose segment information comparable to U.S. accounting rules since 2000. Consequently, differences between Germany and the U.S. before 2000 may be attributed to different accounting standards. On the other hand, the valuation of conglomerates in Germany may have become similar to the valuation of conglomerates in the U.S. due to globalization and the increasing activities of foreign investors.

To gain more insights into the conglomerate discount, we estimate the effect of diversification on market value for each year separately. Figure 1 presents coefficients and confidence intervals from those regressions. The coefficients on diversification are mostly negative and vary from − 23.1 to + 5.4% depending on the year and the measure of market value. This broad range of estimates may explain ambiguous results in prior literature. Lee et al. (2008) argue that the valuation of conglomerates is affected by a country’s institutional setting which changes over time and can cause studies on the conglomerate discount to find different results when analyzing different sample periods. Moreover, the differences among measures evident across the sample period call research designs into question that solely rely on one measure of market value.

Volkov and Smith (2015) and Garrido-Prada et al. (2019) argue that globally diversified firms benefit from easier access to external capital and a more efficient allocation of capital during periods of increased financial constraints. Contrarily, industrially diversified firms are as negatively affected by local recessions as focused firms. Consistent with these studies, we continue to observe a conglomerate discount during the financial crisis 2008. Our results in Fig. 1 further indicate that changes in segment reporting resulting from the mandatory adoption of IFRS 8 in 2009 have not affected the valuation of conglomerates.¹⁰ Interestingly, our analysis reveals mostly insignificant effects of diversification after 2014 and partly positive coefficients on diversification when EV_Merton is the dependent variable. This is of particular interest because we are not aware of any study that examines the conglomerate discount in Germany after 2014.

Finally, we analyze whether the conglomerate discount varies across industries. Table 4 presents ordinary least squares coefficients on diversification for each two-digit SIC industry. Note that we do not tabulate industries with less than 100 observations in any of the three regressions. Consistent with Erdorf et al. (2013) and Santalo and Becerra (2008), our results suggest that the valuation of conglomerates varies across industries. In particular, we find negative and significant coefficients on diversification across all three specifications for SIC codes 20, 35, 37, 59, and 73. However, we also identify industries where conglomerates are not traded at a discount (SIC codes 36, 38, 49, 80, and 87). Conglomerates operating in these industries are expected to suffer less from the disadvantages of diversification. Interestingly, we find firms operating in the motion pictures industry (SIC code 78) to be more likely to realize the advantages of diversification. We find positive and significant coefficients across all three specifications for firms operating in this industry indicating a conglomerate premium of 25.2–37.8%.

The heterogeneity of conglomerate valuations across industries is of particular interest in small industries. As the traditional excess value measure of Berger and Ofek (1995) requires at least 5 focused firms in each industry, the consideration of conglomerates depends on the sample selection process and the availability of data in the respective database. For example, we find 4 industries that fall just below this threshold (i.e., industries with 4 focused firms) resulting in missing excess values for 101 conglomerates that report segments operating in these industries.

As excess values impute the conglomerate value based on multiples from focused firms operating within the same industry, the selection of appropriate benchmark firms is important. However, studies outside the U.S. are often limited in the availability of comparable benchmark firms especially for the largest conglomerates. To explore the impact of the selection of benchmark firms, we add data from large European firms to our sample and re-estimate excess values. Specifically, we add STOXX Europe 600 firms and analyze both the effect of diversification on market value within a European sample and the effect within the German subsample that considers European focused firms for the calculation of excess values.

Table 5 presents ordinary least squares regressions of diversification on market value within this enlarged sample. We continue to find a conglomerate discount in the European sample ranging from 16.3 to 20.4%, indicating that the restriction to German firms may have underestimated the conglomerate discount by 8.4 to 8.9% points. The conglomerate discount changes from 11.2 to 16.1% when we restrict our sample to German firms but keep European firms as benchmark firms indicating a difference of 3.3 to 4.6% points compared to our main models.

Though these results may indicate that the limited availability of comparable focused firms affects our results, the inclusion of firms from other institutional settings may also bias regression outcomes, because country-specific differences are found to affect the valuation of conglomerates (Fauver et al. 2003; Lee et al. 2008; Lins and Servaes 1999; Rudolph and Schwetzler 2013; Weiner 2005). Regardless of the source of these differences, the results emphasize that the selection of benchmark firms, particularly for studies outside the U.S., has an impact on the results.

Another bias induced by the selection of benchmark firms has been explored by Boguth et al. (2022). They argue that measurement errors may arise due to differences between conglomerates and focused firms and propose estimating valuation multiples of conglomerates based on cross-sectional quantile regressions of conglomerates’ value on their sales exposure to 10 Fama French industries. Using this measure, we find the conglomerate discount to become insignificant raising concerns about whether focused firms are appropriate benchmark firms for the calculation of excess values.

Several studies have shown that the decision to diversify is endogenous, resulting in biased estimates of the conglomerate discount. However, prior literature from the U.S. provides mixed evidence on the endogeneity-adjusted conglomerate discount, ranging from studies that find a decrease in the conglomerate discount to studies that find no conglomerate discount or even a premium (Ammann et al. 2012; Chang et al. 2016; Hoechle et al. 2012; Villalonga 2004). The diversity in results may be attributed to both the type of endogeneity addressed and the employed methods.

We begin to analyze the impact of endogeneity by focusing on the omission of relevant factors. To the extent that omitted variables are correlated with both the diversification decision and market valuation, our estimates of the conglomerate discount are biased. In Table 6, we show that the conglomerate discount is affected by the selection of control variables. Specifically, we re-estimate our empirical model, including 1- and 2-year lags of our control variables, as suggested by Campa and Kedia (2002). Even though the inclusion of lagged control variables adds little information to the model and requires no additional data, the conglomerate discount decreases to between 6.7 and 10.4%. In other words, we find a reduction of at least 1% point in each specification by adding little information to our research design. We further include firm fixed effects in our empirical model. The inclusion of firm fixed effects also requires no additional information but has higher informative value as it accounts for (unobservable) firm-specific characteristics that remain constant over time. Our results in Table 6 suggest that firm-specific characteristics partially cause the conglomerate discount, as evident by a reduction in the conglomerate discount of 2–5.1% points. Specifically, we find a conglomerate discount of 6.3% (8.5%) when EV_Sales (EV_Goodwill) is the measure of market valuation and no conglomerate discount when EV_Merton is the dependent variable.

In the next step, we account for a potential self-selection bias by estimating instrumental variables regressions. Estimates on differences between conglomerates and focused firms are only unbiased if the diversification status is randomly assigned. However, this assumption is unrealistic in the context of managerial decisions. 2SLS is a possible approach to eliminate this self-selection bias.¹¹ Following prior research (e.g., Ammann et al. 2012; Campa and Kedia 2002; Villalonga 2004), we analyze four different categories of instruments. First, we include two instruments capturing the attractiveness of the industry in a given year: the percentage of firms that are conglomerates and the percentage of sales accounted for by conglomerates. Industry-specific factors that affect the likelihood to diversify include, for example, industry regulation, market structure, technology, and business risks. Second, we consider time trends such as the existence of M&A waves by including the number and volume of M&A per industry-year. Third, we account for trends in macroeconomic conditions. As 2SLS estimates the effect of all instruments and control variables on the endogenous variable, we already capture time trends that are constant across firms through year fixed effects. Thus, we include the regional growth in GDP and its lagged value to capture time trends that vary across firms. We use the first-digit postal codes of the firms’ headquarters to assign a firm to a specific region and access data on regional GDP from the Federal Statistical Office of Germany. This approach assumes that economic changes within the firm’s primary region have the highest impact on the firm. Fourth, we include a binary variable measuring whether firms are listed on a major exchange (i.e., DAX) as these firms are more visible and have higher analyst coverage, which in turn facilitates M&A activities and raising external financing.¹²

Table 7 presents our first-stage results on the determinants of diversification, considering the instruments and the control variables from Table 6. To ensure robustness, we analyze each combination of the four instrument categories separately, resulting in 15 distinct instrument combinations. However, we only tabulate tuples of instrument categories that sufficiently correlate with our diversification measure and do not produce overidentified models. Specifically, we require F statistics for the joint significance of instruments to exceed 10 and perform Wooldridge’s robust score test of overidentifying restrictions. These criteria are consistent with existing research and are intended to ensure the accuracy of our results (Lal et al. 2023). We find 4 sets of instruments that are valid and have high explanatory power for the diversification decision.¹³ Specifically, our instruments capture industry attractiveness in model (1), industry attractiveness and M&A activities in model (2), industry attractiveness and macroeconomic conditions in model (3), and industry attractiveness, M&A activities, and macroeconomic conditions in model (4). Our results suggest that the fraction of conglomerates and the number of M&A within the industry significantly affect the diversification decision (p < 0.01, respectively).

Our second-stage results are presented in Table 8. We analyze the effect of diversification on EV_Sales, EV_Merton, and EV_Goodwill for each set of instruments separately. Our results suggest that the self-selection bias is responsible for the conglomerate discount. Across all 12 regressions, we consistently observe insignificant effects of diversification. Despite the variety of instruments employed in the literature, our results indicate that each set of instruments that meets the requirements for 2SLS is capable of addressing the self-selection bias.

Consistent with prior literature (e.g., Ammann et al. 2012; Chang et al. 2016; Hoechle et al. 2012; Villalonga 2004), our results indicate that firms decide whether to operate as a focused firm or as a conglomerate and the firm’s decision to diversify is driven by exogenous changes in the firm’s environment, which subsequently impact firm values. Thus, the valuation difference between focused firms and conglomerates cannot be attributed to a causal relationship.