Exercising a firm’s growth options: A portfolio approach

Journal of Business Research 132 (2021) 571–585 Available online 15 May 2021 0148-2963/© 2021 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). Exercising a firm’s growth options: A portfolio approach Pablo de Andr´ es a , Gabriel de la Fuente b , Pilar Velasco b , * a Universidad Aut´ onoma de Madrid and ECGI, Faculty of Economics and Business, Cantoblanco Campus, 28049 Madrid, Spain b Universidad de Valladolid, Faculty of Economics and Business, Avenida Valle del Esgueva, 6, 47011 Valladolid, Spain ARTICLE INFO JEL classification: L25 G32 D22 C25 Keywords: Opportunity exploitation Portfolio of businesses Investment strategy Options exercise Real options ABSTRACT This paper investigates a firm’s decision to exercise its growth options within the current scope of business. We contend that the point at which a firm chooses whether or not to exercise such options depends on several strategic characteristics associated to its portfolio of businesses (i.e. level of diversification, average volatility of its current business, relatedness among them, and rivalry in its core business). Using a sample of U.S. companies from 1998 to 2013, we find evidence that a firm’s level of diversification and the volatility of its current business deter immediate growth option exercise. In contrast, relatedness and rivalry in its core business trigger growth options exercise. Our portfolio perspective differs from the bulk of the literature on real options, which evaluates each growth option exercise in isolation, and contributes to furthering knowledge on the drivers underlying a diversified firm’s strategic investments. 1. Introduction This study investigates a firm’s decision concerning whether or not to exercise its growth options (hereinafter, GO) within the current scope of business. Prior literature has tended to explore the exercise of GO individually, overlooking the strategic characteristics associated to a firm’s portfolio of businesses. Recent research surveys, such as Ipsmiller, Brouthers and Dikova (2019), urge the need to develop a portfolio perspective of option investments, a perspective which has remained largely overlooked. This paper seeks to narrow this gap by adopting a real options portfolio approach to rationalize the relationship between the risk features of the firm’s whole set of businesses and the propensity to GO exercise. Our focus is on GO exercises not involving entry into a new business. By limiting our study to such investments, we discard those arising from the decision to enter a new business, whose determinants/antecedents may be of a different nature. According to the real options (RO) approach, a firm’s value is the sum of the value of its assets-in-place (AiP) and the value of its GO. Assets-in-place are full-scale currently operating businesses. GO are platforms for growth that provide the firm with a right, but not an obligation, to make a full-scale investment in the future. GO do not generate large cash flows, but limit a firm’s commitment while conferring preferential access to future investment (Bowman & Hurry, 1993; Doh & Pearce, 2004; Kulatilaka & Perotti, 1998). The decision to exercise a growth option implies an irreversible full-scale investment (i. e., its exercise price) in exchange for the asset-in-place (underlying asset) that is added to the company’s portfolio of business. The decision concerning whether or not to exercise a firm’s GO by materializing them into major investments (assets-in-place) and the factors affecting the value of such a decision are a central concern in strategic management literature. Previous studies that have adopted the RO approach have provided important insights by delving into the circumstances which may spark GO exercise in a wide range of contexts, such as new market entry (Brouthers & Dikova, 2010; Folta & O’Brien, 2004), venture capital (Ceccagnoli, Higgins, & Kang, 2018; Vrande & Vanhaverbeke, 2013), R&D investments (Huang & Jong, 2019; Huchzermeier & Loch, 2001), internationalization (Brouthers, Brouthers, & Werner, 2008; Rivoli & Salorio, 1996), or strategic alliances (Chi & McGuire, 1996; Estrada, Fuente, & Martin-Cruz, 2010). This stream of literature focuses on the trade-off between the cash flows from an immediate and (to some degree) irreversible investment, and flexibility from exercise postponement (option to wait). Engaging in large sunk investments involves cash flows at the expense of sacrificing flexibility and increasing risk exposure. As a consequence, optimal timing is seen to depend on the irreversibility and risk of investment, lost cash flows that would have been generated if the firm had opted for immediate commitment, the threat of pre-emption in the underlying business, and whether or not the option is simple or compound, * Corresponding author. E-mail addresses: [email protected] (P. Andr´ es), [email protected] (G. Fuente), [email protected] (P. Velasco). Contents lists available at ScienceDirect Journal of Business Research journal homepage: www.elsevier.com/locate/jbusres https://doi.org/10.1016/j.jbusres.2021.04.074 Received 1 June 2020; Received in revised form 23 April 2021; Accepted 29 April 2021 Journal of Business Research 132 (2021) 571–585 572 proprietary or shared (Miller & Folta, 2002). One common feature of this prior literature is that GO exercise decisions are evaluated solely in terms of GO own characteristics, regardless of the portfolio of businesses in which they are embedded. However, GO are part of a broader portfolio at the corporate level, in which the current businesses (assets-in-place) and GO are interrelated and combine to maximize overall firm value. Studies such as Trigeorgis (1993), Vassolo, Anand and Folta (2004), and Yang, Narayanan and De Carolis (2014) show that GO are particularly interconnected, such that decisions taken on any single component impact the value of the others. Unlike financial call options, exercising a growth option may not only influence the value of its underlying real asset but also the value of other underlying GO projects within the same portfolio and even their exercise conditions. Such an interrelation might intensify in the case of a diversified firm as a result of relations between the different businesses (Andr´ es, Fuente & Velasco, 2017b; Yang et al., 2014). Indeed, GO are seen to play a particular role in a diversified firm due to this sort of connection, as first stated by Raynor (2002), who argues that diversification strategies can have “option value” since they might create options on future synergies between businesses. This “options-based diversification” can add value, even if it fails to increase a firm’s performance, in that it reduces risk in a way investors cannot replicate. The hedging role of GO within a portfolio of businesses is no longer a matter of spreading risk (as provided by business diversification), but about the possibility of developing strategic flexibility (Miller, 1998) in response to a declining market or an unpredictable failure of current assets-in-place, by investing in new projects whose performance might preserve a firm’s prior returns (Yang et al., 2014). A firm’s economic exposure arises from a complex mixture of competitive, input supply, output demand, technological, and political risks (Miller, 1998). Such exposure affects all the current businesses in which the firm operates and, eventually, overall corporate performance. GO may help to hedge such external contingencies, thereby affecting a company’s performance. On the one hand, GO are small ‘bets’ on potential growth. Undeveloped businesses frequently require intensive investments while delivering narrow sales and low or negative outcomes. By limiting its commitment to securing just a ‘foot in the door’ of such a business, the firm manages its exposure against certain contingencies. Hence, holding GO within the current scope of business might constitute a good way of undertaking a stepwise commitment in those industries and, therein, limiting downside risk while the firm develops the necessary experience and skills (Bowman & Hurry, 1993). On the other hand, GO maintain the potential for realizing large gains that might offset potential losses from other major full-scale segments. For example, depending on the actual direction of contingencies, a mature and so far profitable segment may decay and start to underperform. Under such circumstances, GO exercise could trigger their underlying businesses in order to generate sufficient gains to make up for such losses, thus reducing a firm’s exposure to corporate risk. Taking the portfolio of businesses as our baseline unit of analysis, we argue that GO exercise should depend on the risk characteristics of the diversified portfolio as a whole. If GO play a role as strategy insurance, exercising such GO would then be governed by the firm’s overall risk. It is not just a question of moving from flexible positions to sunk investments (as might be the case in a financial call option), but of encompassing a broader set of considerations regarding the risk balance between AiP and GO within a diversified portfolio. In particular, we contend that the decision to exercise GO is contingent upon four different but interrelated dimensions of a firm’s risk exposure: level of diversification, average volatility of its current businesses, relatedness between them, and rivalry in its core business. Using a sample of U.S. companies from 1998 to 2013, we identify upticks in a firm’s level of investment which are likely to signal GO exercise. To do so, and following prior literature (Grullon, Lyandres, & Zhadanov, 2012; Lee, Park, & Folta, 2018), we apply the spike concept by Whited (2006), which refers to firm-years of abnormally high investment activity, and we also require a significant decrease in a firm’s GO relevance the year following the spike in order to ensure that those investments arise from exercising previously held GO. We exclude those non-option strategies based on setting up a new business through an immediate full commitment of resources in order to restrict our focus to those large investments arising from previously underdeveloped entries undertaken by the company. Therefore, we focus on large investments within the existing scope of businesses and that do not coincide with a new diversifying entry. In order to better identify this event, we apply Whited’s (2006) different thresholds to pinpoint a large investment, the requirement of a firm’s number of segments to remain unchanged and a significant reduction in a firm’s GO relevance, signalling options exercise. Consistent with our hypotheses, we find that a firm’s degree of diversification and the average volatility of its businesses deter GO exercise, whereas relatedness of a firm’s businesses and rivalry in its core business positively drive investment decisions. The remainder of the paper is organized as follows. The second section sets out the literature review and our hypotheses. The next section describes the sample, variables and econometric methodology. The following section presents our main empirical findings and robustness analyses. The last section discusses and concludes. 2. Literature review and hypotheses Through RO lenses, a distinction can be drawn between two types of investments within a diversified portfolio of a given set of businesses: AiP and GO. AiP are full-scale committed investments and comprise the main current businesses in which a firm operates. In contrast, GO are minor commitments in businesses that could be expanded at some point in the future. AiP are the main source of cash for firms and enable them to exploit economies of scale and scope. However, they have limited or null flexibility value (Smit & Trigeorgis, 2009). The threat of a future decline in a firm’s AiP performance encourages managers to invest in platforms for future growth. GO provide such platforms, whose exercise would offset a future decline in AiP performance. Therefore, GO establish a threshold limit to downside risk (‘strategic insurance’ (Raynor, 2002)), albeit at the expense of cost-scale efficiency. Given these different properties, there should be an optimal combination of AiP and GO that maximizes performance from current full-scale businesses (AiP) whilst hedging against an uncertain future through GO. Such an optimal balance of AiP and GO would thus explain GO exercise and is likely to be affected by the risk the firm is exposed to. Risk exposure of a firm’s portfolio of existing businesses is taken to be a main driver of strategic decisions. In fact, risk management is recognized as lying at the heart of corporate strategy (Bettis, 1983) and as a main factor driving a better understanding of a firm’s success and failure (Lane, Cannella, & Lubatkin, 1998). Even if investors are able to replicate corporate diversification in their own portfolios, this holds true for at least two reasons. First, some stakeholders other than shareholders P. Andr´ es et al. Journal of Business Research 132 (2021) 571–585 573 (such as managers, employees, customers or suppliers) often hold undiversified portfolios with a large portion of their non-financial wealth invested in the firm. Moreover, many of their investments in the company are firm-specific since they cannot be redeployed to other business settings without a significant loss in value (Wang, Barney, & Reuer, 2003). In this regard, Wang et al. (2003) stress the importance of a firm’s risk management vis-` a-vis reducing the probability of financial distress and, as a result, stimulating additional firm-specific investments from its stakeholders. These investments are of paramount importance since they are usually a key source of competitive advantage and a platform for value creation (Lim & Wang, 2007; Miller, 1998; Wang et al., 2003). Second, the risk that determines a firm’s market value is not equivalent to the risk that is relevant for investment portfolio selection. Securities investors cannot usually affect the expected cash flows from each asset in their portfolio. This should not be the case for a firm and its portfolio of businesses, where midcourse strategic decisions are adopted to shape cash flows –increasing gains and limiting losses– in response to how uncertain events unfold (Bowman & Hurry, 1993; Lubatkin & Chatterjee, 1994). This capacity to affect a firm’s cash flows is cornerstone to the RO approach, and offers a view of strategies that is closer to actual decision processes. One key tool for managing the whole range of risk exposures across a firm’s portfolio of businesses is GO. In contrast to AiP diversification, which relies on volatility reduction by combining uncorrelated full-scale assets within a portfolio, GO provide strategic hedging for corporate downside risk by developing flexibility (Andr´ es, Fuente, & Velasco, 2017a; Miller, 1998; Raynor, 2002). GO offer the possibility of taking advantage of current foothold investments in the event that full-scale, mature businesses become exhausted (Miller & Waller, 2003). By exercising its GO, a firm seeks to maintain above-average returns should their products become obsolete or simply less competitive as a result of unexpected shifts in demand, competitors’ decisions, or government regulations, among others (Miller, 1998; Williamson, 2001). Both AiP diversification itself and GO can allow a firm to reduce its dependence on the uncertainty of any single industry, albeit differently. By combining two or more negatively correlated (or weakly correlated) businesses, diversification helps to smooth a firm’s income stream, thereby reducing its return volatility. This risk-reducing outcome stems from an overall averaging effect. By contrast, GO allow a firm to reduce its exposure to economic slumps in any of its main full-scale businesses. In the event of economic downturns, GO provide managers with the ability to intervene by investing in developing businesses. Therefore, the main outcome from GO emerges in the form of minimizing a firm’s downside risk (Del Viva, Kasanen, & Trigeorgis, 2017; Miller, 1998; Raynor, 2002). Corporate risk is also reduced, but as a result of the greater flexibility to respond to environmental uncertainties in a timely fashion. 1 Based on this different role played by AiP diversification and GO, we can expect both strategies to be complementary in hedging a firm’s risk exposure. As a consequence, the decision regarding whether or not to exercise a firm’s GO will depend on the characteristics of its portfolio of businesses as a whole. In particular, we consider four risk features which can impact the propensity to GO exercise: a firm’s level of diversification, the average volatility of its current business, the relatedness among them, and rivalry in its core business. First, managers can reduce a firm’s idiosyncratic risk by distributing its resources across the various full-scale businesses. However, the greater the range of these businesses within a firm’s portfolio (i.e. the diversification level), the greater the number of GO needed to hedge their possible failure. Greater diversification implies a lower volatility of a firm’s income stream, but also a wider scope of market positions that need to be covered by GO in order to limit downside risk. As a result, we expect a negative relationship between the degree of diversification and a firm’s GO exercise decision. A reduction in the range of businesses releases the need for GO hedging and, therefore, encourages GO exercise. Hypothesis 1 posits this relationship: H1. The lower (higher) the degree of diversification, the more (less) likely a GO exercise is to occur. A second variable which might determine the weight of a firm’s GO within its diversified portfolio is the average volatility of its AiP. Lower AiP volatility might encourage GO exercise. Holding unexercised GO will provide insurance against exposure to environmental and industry uncertainties and therefore be preferable to option exercise. Consistent with this argument, Lyandres and Zhdanov (2013) confirm that the mix of AiP and GO is an important determinant in the likelihood of survival, particularly because a firm’s GO portfolio value alleviates the risk of default. As a consequence, the lower the volatility of a firm’s current businesses, the less necessary and valuable its GO in order to provide a response to changeable environmental contingencies. Accordingly, we propose Hypothesis 2: H2. The lower (greater) the average volatility of a firm’s businesses, the more (less) likely a GO exercise is to occur. Another important characteristic of a firm’s portfolio of business that is likely to influence GO exercise is diversification relatedness. Related diversification is considered to outperform unrelated diversification (Becerra, 2009; Markides & Williamson, 1994; Schommer, Richter, & Karna, 2019) since it can take greater advantage of synergies across multiple divisions, thus alleviating the threat of a future decline in AiP and reducing a firm’s need to hedge through GO. Lubatkin and Chatterjee (1994) argue that relatedness provides synergies that other diversification types cannot, and that these advantages allow the firm to respond more effectively, and in a more timely manner, to environmental uncertainties. Complementarily, related diversification fosters firm flexibility through inter-temporal economies of scope from resource redeployment between businesses over time, which can play a further substitutive role of the flexibility granted by GO to hedge corporate exposure in response to future shocks to the firm. Such an internal reallocation advantage favours the most attractive use of corporate resources, which proves less costly and less irreversible the more related those business units are (Helfat & Eisenhardt, 2004; Sakhartov & Folta, 2014). Relatedness reduces the costs required to shift resources from one business to another existing one and makes it easier to reverse resource allocation (if required) in the future, thus increasing its flexibility to respond to environmental pressures (Dickler & Folta, 2020). Recent research has pointed to the hedging role of redeployability within existing business units of a diversified firm (Aivazian, Rahaman, & Zhou, 2019; Lieberman, Lee, & Folta, 2017). For example, Lieberman et al. (2017) find that redeployability reduces the cost of failure in businesses and promotes experimentation within them because the potential of internal reallocation entails lower sunk costs and facilitates exits. Since this complementary flexibility granted by redeployability across related businesses reduces dependency on any single business, fewer GO need to be held and the more likely it is that GO exercise will occur. Accordingly, we state Hypothesis 3: H3. The more (less) related the diversification, the more (less) likely a GO exercise is to occur. 1 While spreading risk could be replicated by financial investors, skewness from GO (only exercised in the case of positive outcomes) provides a strategy insurance against unexpected change, such as those emerging from technological innovation and competitive dynamics, which cannot be replicated with a portfolio of focused firms (Raynor, 2002). Recent research reveals that financial investors might be rationally willing to accept lower average returns in exchange for skewness (Mitton & Vorkink, 2007). In accordance with such rational preferences, Del Viva et al. (2017) show that investors try to increase their portfolio skewness by including a number of stocks with more GO. P. Andr´ es et al. Journal of Business Research 132 (2021) 571–585 574 Our last proposed dimension of a firm’s diversified portfolio that is likely to affect its GO exercise is rivalry in its core business. Strategic management literature has often drawn on measures of risk as proposed by the financial portfolio theory that might not fully correspond to the risk concept which non-financial stakeholders are concerned with. While efficiently diversified investors might pay attention to measures of return volatility, other stakeholders may be more interested in the likelihood of creating and maintaining superior levels of performance in their firms (Ruefli, Collins, & Lacugna, 1999), which directly depend on rivalry in their core businesses. Prior literature suggests the relevance of rivalry in explaining corporate-level resource allocations (McGrath, Chen, & Macmillan, 1998; Morandi-Stagni, Santal´ o, & Giarratana, 2020). A firm can intensify GO exercise in its secondary industries as a means of compensating declining performance in its core business (Miller, 2004; Penrose, 1959; Rumelt, 1974). Aguerrevere (2009) points out that the strategic behaviour of market participants shapes the relative importance of AiP and GO on a firm’s value. Similarly, recent research such as D´ ecaire, Gilje and Taillard (2020) shows that real options exercises are related to peer exercise behaviour. Meslier, Tacneng and Tarazi (2014) and Vo (2017) show that increased competition usually operates as a signal of such a threat of decline, thereby triggering GO exploitation. By GO exercise in its minor industries, a firm reduces its dependence on its core business and is able to acquire competitive advantages from synergistic interrelationships. This allows the firm to mobilize advantages such as low cost, high buyer loyalty or administrative know-how (Lubatkin & Chatterjee, 1994) in order to offset some of the disadvantages stemming from greater rivalry. These competitive advantages allow the firm to uphold its market positions against its rivals in alternative industries in which the firm is already operating, thereby reducing its exposure in the core business. Consistent with these arguments, recent studies such as Ljubownikow and Ang (2020), support the notion that firms shape their diversification strategy as a strategic response in an effort to escape greater competitive intensity. Following on from these arguments, rivalry in a firm’s core business of a diversified firm is likely to accelerate engagement in its embryonic businesses, prompting GO exercise decisions. Thus, we hypothesize: H4. The greater (lower) the rivalry in a firm’s core business, the more (less) likely a GO exercise is to occur. 3. Research design: Data, variables and econometric model 3.1. Dataset and sample selection Our empirical analysis departs from an unbalanced panel sample of U.S. listed companies from 1998 through 2013. Our initial sample covers all publicly traded U.S. companies included in the Worldscope database for that period. Our dataset comprises both active and inactive firms to mitigate potential survivorship bias. We obtain annual firm data from Worldscope and market data from Datastream. These databases are accessed through Thomson One package by Thomson Reuters, which offers full coverage of U.S. companies filing with the Securities Exchange Commission. Additionally, we collect industry information from the U.S. Census Bureau (Statistics of U.S. Businesses), which provides annual data for U.S. business establishments by geography, industry, and enterprise size. Following prior literature, we use Berger and Ofek’s (1995) criteria to select the sample. Accordingly, we exclude firm-years with any business segment in the financial industry (SIC codes 6000–6999) or with non-positive sales. We discard observations containing missing data on total capital, total sales, and segment-level sales. We also restrict our sample to those firm-years with total sales greater than or equal to $20 million, and whose sum of segment sales is within the 99–101% range of the firm’s reported total sales. Finally, we rule out firm-years with missing values for our model variables as well as those reporting only one business segment 2 , since relatedness can only be calculated for diversified firms. Our final panel sample for estimating our full models consists of 8524 firm-year observations corresponding to 1499 firms (1438 active firms and 61 inactive firms). To mitigate outlier problems, we winsorize our continuous variables at the 1% and 99% levels. 3.2. Variables One key issue in our study is to capture whether or not a firm pursues GO exercises in a given year. According to previous literature, such as Grullon et al. (2012) and Lee et al. (2018), GO exercise is captured by investment spikes, which are upticks in a firm’s level of investment. Following those studies, we draw on Whited’s spike concept (2006) that identifies firm-years of abnormally high investment activity. The first problem to tackle is how to delimit the levels of investment that may signal GO exercise, since firms may also invest certain amounts each period for other reasons, such as maintaining or creating new GO. According to Whited (2006), an investment spike occurs if the ratio of investment to total assets is alternatively two or three times greater than the firm-median of said ratio. 3 Moreover, in order to avoid mixing evidence from GO exercise and GO creation, we discard those investments whose main outcomes are new GO. Accordingly, we only select those spikes that lead to a significant reduction in a firm’s GO value. In particular, we require a firm’s Tobin Q to decrease the year following the investment spike. Moreover, such a variation in absolute terms must be equal to or above the sample median. 4 Secondly, we need to isolate those large investments within the existing scope of businesses that do not coincide with entry into a new business. As explained earlier, this is a key issue for our research purpose in order to avoid noise in our analyses and discard those diversification patterns based on full-scale initial investments, which previous works have shown to be inconsistent with a real-option logic of investment (Andr´ es et al., 2017a; Bowman & Hurry, 1993). With this aim in mind, for a spike to be considered as an option exercise, we require the firm’s number of businesses to remain unchanged relative to the period before the spike. In this way, we ensure that the options being exercised come from a strategy pursued by the company within its portfolio of businesses and, therefore, that the abnormal level of investment is not simply the result of setting up a new business. 5 To compute the number of businesses, we employ disaggregation of industries at the 3-digit and 2-digit SIC code level alternatively to check the robustness of our results. We avoid using the 4-digit SIC code classification since such narrower segment data is more likely to suffer from managerial discretion due to the closer similarity between some of the industries, thus leading to a more inconsistent definition of the business segment over time (Folta & O’Brien, 2004; Villalonga, 2004). Taking these criteria, we define two sets of alternative specifications of our dependent variable (GO exercise, EXERCISE). On the one side, based on 3-digit SIC codes, we define: EXERCISE2_3d (EXERCISE3_3d), which equals 1 if the firm’s ratio of capital expenditures to total assets is two times (three times) greater than the firm-median, and the number of segments at the 3-digit SIC code level remains constant, and 0 otherwise. 2 In order to avoid severe restrictions in the number of observations in our sample, we consider businesses at the 4-digit SIC code level when applying this filter. 3 Results remain robust when applying Whited’s (2006) alternative investment threshold of 2.5 times greater than the firm-median. Results are available upon request. 4 As a result of this requirement for computing our dependent variable, our sample for estimation purposes is restricted to the period 1998–2012. 5 Note that even GO exercises which do not involve entry into a new business imply modifying the level of a firm’s diversification and relatedness, since such investments affect sales distribution among current businesses. We thank a reviewer’s suggestion for highlighting this issue. P. Andr´ es et al. Journal of Business Research 132 (2021) 571–585 575 On the other, from a broader industry classification such as the 2-digit SIC code, we apply this same logic to define EXERCISE2_2d and EXERCISE3_2d, alternatively. To measure a firm’s level of diversification (DIVERSIF), we employ the total entropy measure (Jacquemin & Berry, 1979): DIVERSIF =∑ n s=1 Ps×ln(1 Ps ) where ‘P s ’ is the proportion of a firm’s sales in business ‘s’ for a corporation with ‘n’ different 4-digit SIC segments. The higher the DIVERSIF, the greater the diversification. The volatility of a firm’s businesses (VOL) is calculated as the average of the volatility of the different 2-digit SIC code sectors in which the firm operates, weighted by the proportion of the firm’s sales allocated in each sector. The volatility of each 2-digit SIC code sector is measured as the average of the volatility of all unisegment firms operating in that industry. Consistent with prior studies, such as Grullon et al. (2012), we estimate a firm’s annual volatility as the standard deviation of the firm’s daily returns during year t. As far as relatedness is concerned, prior literature considers a multisegment company as related diversified when its divisions belong to the same 2-digit SIC industry. Following Robins & Wiersema (2003), we measure relatedness (RELATED) from total entropy measure, which is split into unrelated and related entropy: RELATED =DIVERSIF −UNRELATED Table 1 Distribution of GO exercises over the sample years. EXERCISE2_3d EXERCISE3_3d Year no. EXERCISE2_3d =1 % EXERCISE2_3d =1 no. EXERCISE3_3d =1 % EXERCISE3_3d =1 1998 0 0.00 0 0.00 1999 125 15.06 47 12.88 2000 120 14.46 75 20.55 2001 110 13.25 48 13.15 2002 45 5.42 18 4.93 2003 16 1.93 7 1.92 2004 43 5.18 14 3.84 2005 38 4.58 13 3.56 2006 53 6.39 27 7.40 2007 83 10.00 35 9.59 2008 79 9.52 31 8.49 2009 18 2.17 10 2.74 2010 27 3.25 11 3.01 2011 44 5.30 16 4.38 2012 29 3.49 13 3.56 TOTAL 830 100 365 100 EXERCISE2_2d EXERCISE3_2d Year no. EXERCISE2_2d ¼1 % EXERCISE2_2d ¼1 no. EXERCISE3_2d ¼1 % EXERCISE3_2d ¼1 1998 0 0.00 0 0.00 1999 127 15.07 48 12.97 2000 123 14.59 76 20.54 2001 114 13.52 50 13.51 2002 46 5.46 19 5.14 2003 16 1.90 7 1.89 2004 43 5.10 14 3.78 2005 38 4.51 13 3.51 2006 55 6.52 27 7.30 2007 83 9.85 35 9.46 2008 79 9.37 31 8.38 2009 18 2.14 10 2.70 2010 27 3.20 11 2.97 2011 45 5.34 16 4.32 2012 29 3.44 13 3.51 TOTAL 843 100 370 100 This table contains GO exercise distribution (number and percentage of firm-year observations displaying a GO exercise). The variables EXERCISE2_3d, EXERCISE 3_3d, EXERCISE 2_2d and EXERCISE 3_2d are the alternative proxies for the occurrence of a GO exercise based on different thresholds (a firm’s ratio of CAPEX to assets greater than two or three times the firm median) and the different SIC industry classification (either 3-digit or 2-digit SIC code levels). EXERCISE2_3d is equal to 1 if the firm’s ratio of capital expenditures to total assets is two times greater than the firm-median and the number of 3-digit SIC code segments remains constant, and 0 otherwise. EXERCISE3_3d, equals 1 if the firm’s ratio of capital expenditures to total assets is three times greater than the firm-median and the number of 3-digit SIC code segments remains constant, and 0 otherwise. EXERCISE2_2d is equal to 1 if the firm’s ratio of capital expenditures to total assets is two times greater than the firm-median and the number of 2-digit SIC code segments remains constant, and 0 otherwise. EXERCISE3_2d, takes the value of 1 if the firm’s ratio of capital expenditures to total assets is three times greater than the firm-median and the number of 2-digit SIC code segments remains constant, and 0 otherwise. In all of them, we select only those spikes leading to a decrease in a firm’s Tobin Q the year following the investment spike, and such a variation in absolute terms must be equal to or above the sample median. P. Andr´ es et al. Journal of Business Research 132 (2021) 571–585 576 Table 2 Correlation matrix between GO exercise and measures of growth opportunities. EXERCISE2_3d EXERCISE3_3d EXERCISE2_2d EXERCISE3_2d MBAR t-2 MBAR t-1 MBAR t MBAR t+1 MBAR t+2 Q t-2 Q t-1 Q t Q t+1 Q t+2 EXERCISE2_3d 1.0000 EXERCISE3_3d 0.6574*** 1.0000 EXERCISE2_2d 0.9908*** 0.6503*** 1.000 EXERCISE3_2d 0.6508*** 0.9904*** 0.6566*** 1.000 MBAR t-2 0.0574*** 0.0329*** 0.0580*** 0.0329*** 1.000 MBAR t-1 0.0734*** 0.0482*** 0.0745*** 0.0486*** 0.8698*** 1.000 MBAR t 0.0648*** 0.0447*** 0.0665*** 0.0455*** 0.7932*** 0.8699*** 1.000 MBAR tþ1 −0.0243*** −0.0130* −0.0233*** −0.0122* 0.7432*** 0.7845*** 0.8611*** 1.000 MBAR tþ2 −0.0290*** −0.0224*** −0.0282*** −0.0224*** 0.6891*** 0.7378*** 0.7983*** 0.8865*** 1.000 Q t-2 0.0500*** 0.0290*** 0.0506*** 0.0289*** 0.9963*** 0.8653*** 0.7893*** 0.7345*** 0.6866*** 1.000 Q t-1 0.0651*** 0.0422*** 0.0663*** 0.0424*** 0.8657*** 0.9960*** 0.8653*** 0.7803*** 0.7345*** 0.8689*** 1.000 Q t 0.0587*** 0.0475*** 0.0605*** 0.0405*** 0.7874*** 0.8651*** 0.9956*** 0.8568*** 0.7930*** 0.7913*** 0.8687*** 1.000 Q tþ1 −0.0318*** −0.0188*** −0.0308*** −0.0182** 0.7374*** 0.7790*** 0.8558*** 0.9944*** 0.8797*** 0.7415*** 0.7826*** 0.8607*** 1.000 Q tþ2 −0.0364*** −0.0265*** −0.0357*** −0.0267*** 0.6845*** 0.7322*** 0.7931*** 0.8813*** 0.9928*** 0.6894*** 0.7370*** 0.7976*** 0.8863*** 1.000 This table presents the pairwise correlations between the (lags, present and lead values) of growth opportunities and large investments. EXERCISE2_3d, EXERCISE3_3d, EXERCISE2_2d and EXERCISE3_2d are the alternative proxies for the occurrence of a GO exercise based on different thresholds (a firm’s ratio of CAPEX to assets greater than two or three times the firm median) and the different SIC industry classification (either 3-digit or 2digit SIC code levels). In all of them, we select only those spikes that lead to a decrease in a firm’s Tobin Q the year following the investment spike, and such a variation in absolute terms must be equal to or above the sample median. MBAR and Q are proxies for growth opportunities. MBAR is the market-to-book assets ratio as defined by Adam and Goyal (2008) and Q is Tobin’s Q as computed in Cao et al. (2008). ****, ** and * denote statistical significance at the 1%, 5%, and 10% level, respectively. Table 3 Descriptive statistics and correlation matrix for explanatory variables. N Mean Median Std. deviat. Min. Max. DIVERSIF VOL RELATED RIVAL SIZE LEVERAGE GDPgrowth INDUSTRYgrowth DIVERSIF 8,524 0.6043 0.6193 0.2952 0.0001 1.3002 1.000 VOL 8,524 0.2476 0.0775 0.6498 0.0198 4.6726 0.0047 1.000 RELATED 8,524 0.2486 0.0000 0.3064 0.0000 1.0979 0.4143*** −0.0143 1.000 RIVAL 8,524 10.397 9.924 1.4647 6.9726 13.099 −0.0549*** 0.1075*** 0.0052 1.000 SIZE 8,524 6.5136 6.5589 1.9712 2.2976 10.495 0.1686*** 0.0231** 0.1265*** −0.2007*** 1.000 LEVERAGE 8,524 0.2430 0.2232 0.2011 0.0000 1.0495 −0.0391*** −0.0086 −0.0302*** −0.0959*** 0.1302*** 1.000 GDPgrowth 8,524 0.0238 0.0250 0.0180 −0.0280 0.0470 0.0054 −0.0171 0.0295** 0.0102 −0.1015*** 0.0304** 1.000 INDUSTRYgrowth 8,524 0.1181 0.0885 0.9909 −0.9651 40.271 −0.0086 −0.0132 −0.0072 0.0190* −0.0200* 0.0079 0.0834*** 1.000 This table provides the main summary descriptive statistics and pairwise correlations for the explanatory variables. VOL is the volatility of a firm’s portfolio of businesses and is computed as the average of the volatility of the different 2-digit SIC code sectors in which the firm operates, weighted by the proportion of the firm’s sales allocated in each sector. RIVAL denotes rivalry in the core business and is calculated as the natural logarithm of the number of firms operating in the same 2-digit SIC code industry as the core business of the corresponding firm. DIVERSIF measures a firm’s level of diversification and is taken as the total entropy index by Jacquemin and Berry (1979). RELATED denotes the degree of relatedness between businesses and is Jacquemin and Berry’s (1979) related entropy index. Control variables: SIZE, calculated as the natural logarithm of book value of total assets; LEVERAGE, measured by the ratio of total debt to total assets; GDPgrowth, calculated as the U.S. gross domestic product annual change based on 2009 dollars; and INDUSTRYgrowth, measured by the annual growth rate of a firm’s 2-digit primary industry. ****, ** and * denote statistical significance at the 1%, 5%, and 10% level, respectively. P. Andr´ es et al. Journal of Business Research 132 (2021) 571–585 577 where UNRELATED =∑m r=1Pr×ln(1 Pr)and ‘P r ’ is the proportion of a firm’s sales in business ‘r’ for a corporation with ‘m’ different 2-digit SIC segments. Following Folta & Miller (2002), we compute the degree of rivalry in a firm’s core business (RIVAL) by the number of rivals actively operating in the same product domain. RIVAL is calculated as the natural logarithm of the number of firms operating in the same 2-digit SIC code industry as the core business of the corresponding firm. We gather yearly data on the total number of U.S. firms by NAICS codes from the U.S. Census Bureau and then match NAICS codes to SIC codes. Finally, we include a set of control variables that are of paramount importance for corporate investment decisions: a firm’s size, a firm’s leverage, gross domestic product growth, and industry growth. We proxy size (SIZE) by the natural logarithm of the book value of total assets. Since large firms hold fewer unexercised options, the larger a firm, the lower the probability of exercising additional GO (Bernardo & Chowdhry, 2002). Leverage (LEVERAGE) is measured by the ratio of total debt to total assets. The higher the leverage, the lower the financial flexibility to exploit further GOs. Moreover, Myers (1977) notices that debt may lead to valuable investment opportunities being passed up, especially when debt matures after GO expire. Finally, gross domestic Table 4 Difference of means tests between GO exercise and non-GO exercise firm-year observations. Panel A: Difference of means test with DIVERSIF EXERCISE2_3d =0 (1) EXERCISE2_3d =1 (2) Difference (1)-(2) EXERCISE3_3d =0 (1) EXERCISE3_3d =1 (2) Difference (1)-(2) 0.1931 0.1502 0.0429*** (3.7174) 0.1925 0.1351 0.0574*** (3.3258) EXERCISE2_2d =0 (1) EXERCISE2_2d =1 (2) Difference (1)-(2) EXERCISE3_2d =0 (1) EXERCISE3_2d =1 (2) Difference (1)-(2) 0.1929 0.1543 0.0386*** (3.3704) 0.1925 0.1386 0.0539*** (3.1469) Panel B: Difference of means test with VOL EXERCISE2_3d =0 (1) EXERCISE2_3d =1 (2) Difference (1)-(2) EXERCISE3_3d =0 (1) EXERCISE3_3d =1 (2) Difference (1)-(2) 0.2419 0.2126 0.0293 (1.2876) 0.2418 0.1801 0.0617* (1.8115) EXERCISE2_2d =0 (1) EXERCISE2_2d =1 (2) Difference (1)-(2) EXERCISE3_2d =0 (1) EXERCISE3_2d =1 (2) Difference (1)-(2) 0.2419 0.2121 0.0298 (1.3174) 0.2418 0.1806 0.0612* (1.8102) Panel C: Difference of means test with RELATED EXERCISE2_3d =0 (1) EXERCISE2_3d =1 (2) Difference (1)-(2) EXERCISE3_3d =0 (1) EXERCISE3_3d =1 (2) Difference (1)-(2) 0.2466 0.2688 −0.0222 (-1.0596) 0.2469 0.2658 −0.0189 (-0.5959) EXERCISE2_2d =0 (1) EXERCISE2_2d =1 (2) Difference (1)-(2) EXERCISE3_2d =0 (1) EXERCISE3_2d =1 (2) Difference (1)-(2) 0.2464 0.2745 −0.0281 (-1.3755) 0.2469 0.2735 −0.0266 (-0.8597) Panel D: Difference of means test with RIVAL EXERCISE2_3d =0 (1) EXERCISE2_3d =1 (2) Difference (1)-(2) EXERCISE3_3d =0 (1) EXERCISE3_3d =1 (2) Difference (1)-(2) 10.5314 10.6576 −0.1262** (-2.3462) 10.5326 10.7266 −0.1940** (-2.4157) EXERCISE2_2d =0 (1) EXERCISE2_2d =1 (2) Difference (1)-(2) EXERCISE3_2d =0 (1) EXERCISE3_2d =1 (2) Difference (1)-(2) 10.5312 10.6611 −0.1299** (-2.4321) 10.5325 10.7333 −0.2008** (-2.517) This table shows the difference of means tests for the different explanatory variables between firm years in which there is a GO exercise and those in which there is not. EXERCISE2_3d, EXERCISE3_3d, EXERCISE2_2d and EXERCISE3_2d are the alternative proxies for the occurrence of a GO exercise based on different thresholds (a firm’s ratio of CAPEX to assets greater than two or three times the firm median) and the different SIC industry classification (either 3-digit or 2-digit SIC code levels). In all of them, we only select those spikes that lead to a decrease in a firm’s Tobin Q the year following the investment spike, and such a variation in absolute terms must be equal to or above the sample median. Panel A contains the difference of means tests for VOL, the volatility of a firm’s portfolio of businesses, which is computed as the average of the volatility of the different 2-digit SIC code sectors in which the firm operates, weighted by the proportion of the firm’s sales allocated in each sector. Panel B shows the difference of means tests for the variable RIVAL, which denotes rivalry at the core business and is calculated as the natural logarithm of the number of firms operating in the same 2-digit SIC code industry as the core business of the corresponding firm. Panel C comprises the difference of mean tests for DIVERSIF, which denotes a firm’s degree of diversification and is taken as the total entropy index by Jacquemin and Berry (1979). Panel D displays the difference of means tests for RELATED, the degree of relatedness between businesses measured by Jacquemin and Berry’s related entropy index. t-statistics are in parentheses. ****, ** and * denote statistical significance at the 1%, 5%, and 10% level, respectively. P. Andr´ es et al. Journal of Business Research 132 (2021) 571–585 578 product growth (GDPgrowth) and industry growth (INDUSTRYgrowth) control for the economic cycle attractiveness and the expansion behaviour of a firm’s industry peers, respectively. These two factors are likely to affect a firm’s investment cycle. GDPgrowth is calculated as the U.S. gross domestic product annual change based on 2009 dollars; and INDUSTRYgrowth is measured by the annual growth rate of sales of a firm’s 2-digit SIC primary industry. 3.3. Descriptive statistics Table 1 shows the distribution of GO exercises in the sample. The alternative industry classification used does not lead to a substantial difference in the number of GO exercises. Most occurred between 1999 and 2001 (about 45% of the sample total). There is also a noticeable peak in 2007–2008, with said years marking the onset of the financial crisis. During the rest of the sample period, the distribution of the number of GO exercises across years remained quite stable. To verify the accuracy of our EXERCISE proxies for identifying large investments implying GO exercise, we compute the pairwise correlations of these proxies with two widely used measures of GO: the marketto-book assets ratio (Adam & Goyal, 2008; Folta & O’Brien, 2004), and Tobin’s Q (Alessandri, Tong, & Reuer, 2012; Cao, Simin, & Zhao, 2008; Wright, Ferris, Sarin, & Awasthi, 1996). We follow Adam and Goyal (2008) to compute the market-to-book assets ratio (MBAR) and we apply Cao et al.’s (2008) definition of Tobin’s Q (Q). We also consider two lag values and two lead values of MBAR and Q. Table 2 reports the correlations. EXERCISE dummies are positively correlated with the lags and present values of MBAR (MBAR t-2 , MBAR t-1 and MBAR), and are statistically significant at the 1% level. This indicates that firms displaying a GO exercise have greater prior GO available. In contrast, EXERCISE variables show a negative correlation with the lead values of MBAR (MBAR t+1 and MBAR t+2 ). These latter correlations are greater in absolute value at lower investment thresholds. Results are robust to the use of Q. Finally, Table 3 summarizes the descriptive statistics of our variables and the correlation between them. The positive correlation between RIVAL and VOL (0.1075, p-value =0.0000) suggests that firms which are more exposed to rivalry in the core business are also subject to greater average volatility in their businesses. SIZE and LEVERAGE display negative significant correlations with RIVAL. Particularly noticeable is the correlation between SIZE and RIVAL (-0.2007, p-value =0.0000), suggesting that larger firms enjoy lower levels of rivalry in their core industry. RIVAL has a positive correlation with INDUSTRYgrowth (0.0190, p-value =0.0787), indicating that competitive rivalry increases in industry expansive cycles. In addition, related diversification (RELATED) is positively associated with SIZE, probably because of Table 5 Baseline model (Eq. (1)). Dependent variable: EXERCISE2_2d Dependent variable: EXERCISE3_2d Variables (1) (2) (3) (4) Coefficient Marginal effect Coefficient Marginal effect Coefficient Marginal effect Coefficient Marginal effect Intercept −1.8957*** (0.2765) −0.9463 (0.8624) −1.7733*** (0.4018) −1.7859 (1.2401) DIVERSIF −0.2311* (0.1242) −0.0121* (0.0065) −0.2334* (0.1300) −0.0107* (0.0060) −0.4851*** (0.1794) −0.0098*** (0.0037) −0.5061*** (0.1818) −0.0086*** (0.0033) VOL −0.1469** (0.0668) −0.0077** (0.0035) −0.1018* (0.0613) −0.0047* (0.0028) −0.0996 (0.0906) −0.0020 (0.0018) −0.0497 (0.0792) −0.0008 (0.0014) RELATED 0.2845*** (0.1099) 0.0149*** (0.0058) 0.2355** (0.1134) 0.0108** (0.0052) 0.3936** (0.1602) 0.0079** (0.0032) 0.3668** (0.1639) 0.0062** (0.0028) RIVAL 0.0618*** (0.0202) 0.0032*** (0.0010) 0.0579*** (0.0211) 0.0027*** (0.0010) 0.0601** (0.0291) 0.0012** (0.0006) 0.0572* (0.0298) 0.0010* (0.0005) Controls SIZE −0.0913*** (0.0173) −0.0048*** (0.0009) −0.0902*** (0.0184) −0.0042*** (0.0008) −0.1571*** (0.0288) −0.0032*** (0.0005) −0.1610*** (0.0300) −0.0027*** (0.0005) LEVERAGE −0.4910*** (0.1814) −0.0258*** (0.0094) −0.5661*** (0.1914) −0.0261*** (0.0086) −0.3966 (0.2430) −0.0080 (0.0050) −0.4233* (0.2495) −0.0072* (0.0044) GDPgrowth 2.5261 (1.6992) 0.1326 (0.0891) −63.7813 (43.4223) −2.9369 (1.9726) 1.9668 (2.3379) 0.0396 (0.0485) −9.0489 (60.3648) −0.1542 (1.0295) INDUSTRYgrowth −0.0160 (0.0103) −0.0008 (0.0005) −0.0228 (0.1400) −0.0010 (0.0064) −0.0197 (0.0156) −0.0004 (0.0003) −0.0853 (0.1719) −0.0014 (0.0030) Year fixed effects NO YES NO YES Clustered standard errors by firm YES YES YES YES N 8524 7935 8524 7935 Pseudo R 2 0.0436 0.0921 0.0813 0.1225 Log likelihood −994.4783 −929.1826 −495.6598 −467.1568 p-value chi2 0.000 0.000 0.000 0.000 This table summarizes the probit estimation results for our baseline model (Eq. (1)). The probit estimated coefficients and the marginal effects for the explanatory variable are displayed (standard error in parentheses). GO exercise occurrence is regressed on the volatility of a firm’s portfolio of businesses, rivalry in the core business, degree of diversification and relatedness. EXERCISE2_2d and EXERCISE3_2d are the alternative proxies for the event of a GO exercise based on different thresholds (a firm’s ratio of CAPEX to assets greater than two or three times the firm median), at the 2-digit SIC code level. In all of them, we only select those investment spikes that lead to a decrease in a firm’s Tobin Q the year following the GO exercise, and such a variation in absolute terms must be equal to or above the sample median. VOL is the volatility of a firm’s portfolio of businesses and is computed as the average of the volatility of the different 2-digit SIC code sectors in which the firm operates, weighted by the proportion of the firm’s sales allocated in each sector. RIVAL denotes rivalry in the core business and is calculated as the natural logarithm of the number of firms operating in the same 2-digit SIC code industry as the core business of the corresponding firm. DIVERSIF measures a firm’s level of diversification and is taken as the total entropy index by Jacquemin and Berry (1979). RELATED denotes the degree of relatedness between businesses and is Jacquemin and Berry’s related entropy index. Control variables: SIZE, calculated as the natural logarithm of book value of total assets; LEVERAGE, measured by the ratio of total debt to total assets; GDPgrowth, calculated as the U.S. gross domestic product annual change based on 2009 dollars; and INDUSTRYgrowth, measured by the annual growth rate of a firm’s 2-digit primary industry. Standard errors (clustered by firm) are in parentheses below each estimated coefficient. ****, ** and * denote statistical significance at the 1%, 5%, and 10% level, respectively. The log-likelihood ratio (LR) test follows a χ 2 distribution. P. Andr´ es et al. Journal of Business Research 132 (2021) 571–585 579 larger companies’ greater ability to exploit advantages such as economies of scope. 3.4. Econometric approach and empirical models To test our hypotheses, we perform a number of empirical analyses. As preliminary tests, we carry out a set of difference of means tests for all the variables involved in our subsequent regressions in order to ascertain whether they differ significantly between firm-year observations that show GO exercise and those that do not. Given the dichotomous nature of our dependent variable EXERCISE, we then use a probit model to estimate the likelihood of pursuing a GO exercise. Our baseline model is: Prob(EXERCISE =1) =β0+β1DIVERSIFit +β2VOLit +β3RELATEDit +β4RIVALit +β5SIZEit +β6LEVERAGEit +β7GDPgrowtht+β8INDUSTRYgrowthit + ε it (1) where i denotes each company, t represents the time period, and ε it is the error term. This equation is estimated alternatively with the four different proxies EXERCISE2_3d, EXERCISE3_3d, EXERCISE2_2d and EXERCISE3_2d, in order to test the consistency of our results. In all regressions, we use firm-clustered standard errors to account for unobserved firm effects (Petersen, 2009). This also enables us to control for residual dependence within a firm 6 . Year fixed effects are Table 6 Baseline model (Eq. (1)). Additional robustness checks using EXERCISE2_3d and EXERCISE3_3d. Dependent variable: EXERCISE2_3d Dependent variable: EXERCISE3_3d Variables (1) (2) (3) (4) Coefficient Marginal effect Coefficient Marginal effect Coefficient Marginal effect Coefficient Marginal effect Intercept −1.9168*** (0.2827) −0.9630 (0.8630) −1.8200*** (0.4058) −1.8174 (1.2319) DIVERSIF −0.2105* (0.1257) −0.0107* (0.0064) −0.2124* (0.1312) −0.0095* (0.0059) −0.4539** (0.1783) −0.0092** (0.0037) −0.4754** (0.1810) −0.0082** (0.0033) VOL −0.1474** (0.0695) −0.0075** (0.0035) −0.1054* (0.0643) −0.0047* (0.0029) −0.0973 (0.0921) −0.0020 (0.0019) −0.0505 (0.0803) −0.0009 (0.0014) RELATED 0.2488** (0.1103) 0.0127** (0.0056) 0.2049* (0.1136) 0.0092* (0.0051) 0.3466** (0.1652) 0.0071** (0.0034) 0.3214* (0.1682) 0.0056* (0.0029) RIVAL 0.0605*** (0.0208) 0.0031*** (0.0010) 0.0561*** (0.0216) 0.0025*** (0.0010) 0.0567* (0.0297) 0.0012* (0.0006) 0.0538* (0.0303) 0.0009* (0.0005) Controls SIZE −0.0881*** (0.0175) −0.0045*** (0.0009) −0.0875*** (0.0186) −0.0039*** (0.0009) −0.1454*** (0.0287) −0.0030*** (0.0005) −0.1494*** (0.0300) −0.0026*** (0.0005) LEVERAGE −0.4945*** (0.1861) −0.0252** (0.0094) −0.5659*** (0.1959) −0.0254*** (0.0087) −0.3874 (0.2507) −0.0079 (0.0053) −0.4105 (0.2562) −0.0071 (0.0046) GDPgrowth 2.3197 (1.7186) 0.1180 (0.0873) −63.2422 (43.3423) −2.8450 (1.9236) 2.0030 (2.3238) 0.0408 (0.0487) −9.3834 (59.8410) −0.1629 (1.0401) INDUSTRYgrowth −0.0127 (0.0087) −0.0006 (0.0004) 0.0261 (0.1337) 0.0012 (0.0060) −0.0187 (0.0149) −0.0004 (0.0003) −0.0889 (0.1716) −0.0015 (0.0030) Year fixed effects NO YES NO YES Clustered standard errors by firm YES YES YES YES N 8524 7935 8524 7935 Pseudo R 2 0.0410 0.0872 0.0723 0.1128 Log likelihood −962.3737 −901.7729 −483.9482 −456.7519 p-value chi2 0.000 0.000 0.000 0.000 This table summarizes the probit estimation results for our baseline model (Eq. (1)). The probit estimated coefficients and the marginal effects for the explanatory variable are displayed (standard error in parentheses). GO exercise occurrence is regressed on the volatility of a firm’s portfolio of businesses, rivalry in the core business, degree of diversification, and relatedness. EXERCISE2_3d and EXERCISE3_3d are the alternative proxies for the event of a GO exercise based on different thresholds (a firm’s ratio of CAPEX to assets greater than two or three times the firm median), at the 3-digit SIC code level. In all of them, we only select those investment spikes that lead to a decrease in a firm’s Tobin Q the year following the GO exercise, and such a variation in absolute terms must be equal to or above the sample median. VOL is the volatility of a firm’s portfolio of businesses and is computed as the average of the volatility of the different 2-digit SIC code sectors in which the firm operates, weighted by the proportion of the firm’s sales allocated in each sector. RIVAL denotes rivalry in the core business and is calculated as the natural logarithm of the number of firms operating in the same 2-digit SIC code industry as the core business of the corresponding firm. DIVERSIF measures a firm’s level of diversification and is taken as the total entropy index by Jacquemin and Berry (1979). RELATED denotes the degree of relatedness between businesses and is Jacquemin and Berry’s related entropy index. Control variables: SIZE, calculated as the natural logarithm of book value of total assets; LEVERAGE, measured by the ratio of total debt to total assets; GDPgrowth, calculated as the U.S. gross domestic product annual change based on 2009 dollars; and INDUSTRYgrowth, measured by the annual growth rate of a firm’s 2-digit primary industry. Standard errors (clustered by firm) are in parentheses below each estimated coefficient. ****, ** and * denote statistical significance at the 1%, 5%, and 10% level, respectively. The log-likelihood ratio (LR) test follows a χ 2 distribution. 6 Additional robustness estimations have been conducted by including firm fixed effects. Results prove to be robust and are available upon request. P. Andr´ es et al.