INTRODUCTION
Firm size is among the most well-documented organizational characteristics affecting competitive processes, and has been identified as a critical antecedent of firm survival along the industry life cycle (Agarwal & Audretsch, Reference Agarwal and Audretsch2001; Agarwal, Sarkar, & Echambadi, Reference Agarwal, Sarkar and Echambadi2002; Baum, Reference Baum1995; Bayus & Putsis, Reference Bayus and Putsis1999; Dobrev & Carroll, Reference Dobrev and Carroll2003; Dowell, Reference Dowell2006; Josefy, Kuban, Ireland, & Hitt, Reference Josefy, Kuban, Ireland and Hitt2015; Klepper, Reference Klepper1997). As competition increases over the industry life cycle, products undergo a commoditization process that decreases average prices and unit margins (Gort & Klepper, Reference Gort and Klepper1982; Klepper & Graddy, Reference Klepper and Graddy1990; Klepper & Thompson, Reference Klepper and Thompson2006). Size helps larger firms cope with increased competition using mechanisms such as scale economies, increased product customization, and brand and product differentiation (Agarwal et al., Reference Agarwal, Sarkar and Echambadi2002; Bhaskarabhatla & Klepper, Reference Bhaskarabhatla and Klepper2014; Dobrev & Carroll, Reference Dobrev and Carroll2003). Firms that fail to achieve sufficient size can follow niche strategies, occupying industry segments that are unattractive to larger competitors (Agarwal et al., Reference Agarwal, Sarkar and Echambadi2002; Baum, Reference Baum1995; Klepper, Reference Klepper1996, Reference Klepper1997). Between these extremes lie mid-sized companies, which suffer from a ‘liability of middleness’ (Agarwal et al., Reference Agarwal, Sarkar and Echambadi2002; Barroso & Giarratana, Reference Barroso and Giarratana2013; Baum, Reference Baum1995; Dobrev & Carroll, Reference Dobrev and Carroll2003); these companies experience the highest competitive pressures and exhibit the lowest survival rates.
However, when certain assumptions behind these evolutionary processes do not hold, exceptions to these dynamics may exist (Klepper, Reference Klepper1997). The underlying mechanisms shaping the evolution of populations in natural resource industries are different from those operating in manufacturing and technological industries. These unique mechanisms alter fit conditions and selection pressures and, consequently, have an impact on size effects. First, natural resource industries are not born from radical technological changes at the product level. These industries do experience technological shocks, but these occur mainly at the process level. Second, since the main product is and remains a commodity, it makes no sense to refer to a commoditization process. This is fundamentally distinct from manufacturing and technological products, which start out highly differentiated and become progressively commoditized. Third, the product is decoupled from the producer when consumers purchase it. Commodities like pulp are perfect substitutes for each other under equivalent specifications. This implies that the market will trade them, regardless of the producer, as long as they meet a specified minimum standard, known as basis grade. Moreover, commodities not only serve as inputs for different industries, but also as underlying assets for financial derivatives, which are traded in financial markets. Finally, price determination depends on factors exogenous to the industry as much as endogenous factors, generating levels of price volatility much higher than those observed in manufacturing and technological industries.
The need for studies of competition in natural resource industries is substantial, given the importance of these industries for emerging economies. The wealth embodied in natural resources constitutes a significant proportion of the wealth of most nations, often exceeding the wealth embodied in produced capital, which makes natural resource management a key aspect of economic development (World Bank, 2006). Natural resources represent between a quarter and a third of global exports (World Trade Organization, UNCTAD, 2013, 2015), and national economic activity in most emerging economies depends heavily on natural resource industries. Natural resources generally form the backbone of rural economies in low- and middle-income countries and, if managed wisely through sound policies and institutions, can be used to generate growth that benefits the most vulnerable segments of these countries’ populations. Moreover, natural resource endowments can compensate for institutional weaknesses in attracting foreign direct investment to emerging markets (Aleksynska & Havrylchyk, Reference Aleksynska and Havrylchyk2013). Natural resource prices are closely related to stock market development and, ultimately, economic growth in resource-abundant countries (Billmeier & Massa, Reference Billmeier and Massa2009). Additionally, resource-rich countries are usually high-price economies that tend to miss out on export-oriented growth (Sachs & Warner, Reference Sachs and Warner2001). Despite the prominent role that natural resource industries play in emerging economies, few studies on these industries exist in the strategic management field (George, Schillebeeckx, & Liak, Reference George, Schillebeeckx and Liak2015). Exploring how size affects competitive evolution in natural resource industries thus addresses a highly relevant topic and fills a significant gap in the literature.
In this article, we define firm size based on two elements: the volume of production of a particular product (i.e., scale) and the number of submarkets in which the firm competes (i.e., breadth). Scale and breadth have absolute and relative values. Theoretically, we focus on the latter values (Dobrev & Carroll, Reference Dobrev and Carroll2003) and explore their effect on firms’ survival in natural resource industries. We focus the analysis on two theoretical mechanisms – sunk costs and price volatility. We assume an industry with no product differentiation, in which the effect of intangible assets is irrelevant. Both mechanisms emerge from endogenous and exogenous sources that, combined, generate unique selection pressures that differ from those observed in similar processes in manufacturing and technological industries. We proceed in two phases, first analyzing size effects at the product-line level within a submarket and next examining the mechanism that relates size with firm survival.
At the product-line level, selection fundamentally depends on cost advantages. We call this supply-side competition, that is, competition based mainly on supply-side isolating mechanisms rather than demand-side isolating mechanisms. Since natural resource industries tend to be capital intensive, cost advantages depend on scale. However, scale exposes firms to higher levels of sunk costs, which – mainly in a context of strong price volatility – generate the risk of excess capacity (Sutton, Reference Sutton1991). We theoretically deduce that small firms live at the extinction threshold due to the need to achieve minimum efficient scale in a context where prices oscillate significantly due to endogenous and exogenous reasons. We also suggest that the largest firms face a misfit with the environment. This occurs because strong price volatility introduces the liability of sunk costs. Therefore, we hypothesize that scale in a particular product line has an inverted U-shaped relationship with the survival of that product line.
At the firm level, selection results from a combination of mechanisms that determine minimum efficient scale, sunk costs/excess capacity, and price volatility in the absence of consumption scope economies. We start by arguing that, in general, size brings benefits stemming from scope economies (Stern & Henderson, Reference Stern and Henderson2004). As firms diversify, they can minimize the total excess capacity from sunk costs since price volatility is not perfectly correlated among submarkets. In natural resource industries, however, production scope economies prevail over consumer scope economies. Because production scope economies offer a much smaller cost advantage than consumption scope economies do (Cottrell & Nault, Reference Cottrell and Nault2004), the effect of scope economies is substantially smaller in natural resource industries than in other industries. Limited scope economies bound the potential gains from breadth competition. Moreover, managing excess capacity has nontrivial coordination costs; firms must decide, for instance, whether to add new production lines in order to take advantage of growing demand on a certain line and whether to upgrade technology in order to maintain production cost advantages on alternative lines. As firms increase their presence in multiple markets, coordination costs to optimize these different investments grow at increasingly higher rates. Therefore, we hypothesize that specialists have the lowest survival rates, followed by the largest generalists. Companies in the middle – those with a moderate level of breadth – have the best combination of resources to maximize their chances of survival. This inverted U-shaped relationship between breadth and survival contrasts with what has been observed in technological or manufacturing industries (e.g., Cottrell & Nault, Reference Cottrell and Nault2004; Sorenson, Reference Sorenson2000; Stern & Henderson, Reference Stern and Henderson2004).
We study the paper and pulp (P&P) industry since it is particularly relevant in emerging economies (FAO, 2015). Table 1 shows the relative magnitude of forest products exports compared to the total exports and the gross domestic product of eight selected developing countries. Since detailed firm-level data is not available for emerging countries, we test our hypotheses using data from the US market, where industry reports provide comprehensive company-level data for the period 1970–2000. These reports provide direct measures of installed capacity as well as the number of submarkets in which each firm competes.
Table 1. Forest products exports as a percentage of total exports and GDP, 2015
Sources: The World Bank and FAO-Forest Products Yearbook (2015).[Footnote 2]
The P&P industry exhibits persistently high heterogeneity in product-line scale and submarket presence, ranging from small, single-product firms to very large, multi-product firms. Therefore, we proxy scale and breadth by exact measures of installed capacity and submarket presence, something that would be very difficult to achieve in a multiple-industry study. We find strong empirical support for our hypotheses, validating the mechanisms described in the previous paragraphs.
Our research advances several related literatures, including evolutionary models, relative size effects, scale and breadth competition, and competition in natural resource industries. Although the manuscript uses a database from the US, the competitive implications are particularly relevant for emerging economies given the importance of natural resource industries in general and P&P industries in particular in these regions. Our findings partially complement recent studies on product proliferation (Barroso & Giarratana, Reference Barroso and Giarratana2013) by exploring a context where product proliferation is absent. We also address some recent concerns regarding the lack of strategic management studies on natural resource industries (George et al., Reference George, Schillebeeckx and Liak2015). The manuscript has a confirmative methodological contribution on U-shape effects, strictly following important recent recommendations on how to theoretically derive and test these effects (Haans, Pieters, & He, Reference Haans, Pieters and He2015). Finally, the manuscript also has important managerial implications from the strategy content perspective.
The manuscript is organized as follows. In the next section, we review the literature relating size and survival in non-commodity industries and stress potential shortcomings. Next, we clarify the particularities of natural resource industries, present theoretical mechanisms, and develop hypotheses. The remaining sections provide the empirical context, empirical tests, discussion, and conclusion.
SIZE COMPETITION: SCALE AND BREADTH PORTFOLIO EFFECTS
The understanding of competitive evolution in differentiated products industries and the effect of size on survival has its roots in different theories, such as the product life cycle (Levitt, Reference Levitt1965; Vernon, Reference Vernon1966), technology-dominant design (Utterback & Abernathy, Reference Utterback and Abernathy1975), and evolutionary economics (Gort & Klepper, Reference Gort and Klepper1982). The evolution of competition starts with a technological innovation followed by a diffusion process (Levitt, Reference Levitt1965; Utterback & Abernathy, Reference Utterback and Abernathy1975; Vernon, Reference Vernon1966). In the initial years of the industry life cycle, competition is characterized by low rivalry levels as the new industry gains legitimacy. After the establishment of a dominant design, rivalry increases as incumbents grow and new competitors enter the market (Agarwal et al., Reference Agarwal, Sarkar and Echambadi2002; Baum, Reference Baum1995; Klepper, Reference Klepper1997; Utterback & Abernathy, Reference Utterback and Abernathy1975). Consequently, average prices and unit margins decrease (Klepper & Graddy, Reference Klepper and Graddy1990). Increasing rivalry, together with increasing product commoditization, leads competitors to shift their focus from product innovation to process innovation (Utterback & Abernathy, Reference Utterback and Abernathy1975). Rivalry peaks when the industry reaches maturity, inducing an endogenous shake-out process (Klepper & Graddy, Reference Klepper and Graddy1990; Klepper & Thompson, Reference Klepper and Thompson2006). The industry enters into an era of strong and persistent price competition (Carree & Thurik, Reference Carree and Thurik2000; Peltoniemi, Reference Peltoniemi2011). Firms that fail and exit the industry are often those that have failed to achieve the optimum size (Agarwal et al., Reference Agarwal, Sarkar and Echambadi2002).
The decrease in unit margins acts as an imperative to grow, so sales increments eventually sustain total profits (Klepper, Reference Klepper1997). This imperative to grow can take at least two different forms: scale competition and breadth competition (Agarwal et al., Reference Agarwal, Sarkar and Echambadi2002; Dobrev & Carroll, Reference Dobrev and Carroll2003; Dowell, Reference Dowell2006; Klepper & Thompson, Reference Klepper and Thompson2006). Scale competition is a decision to grow by focusing on a certain product line. The strategic aim is to reduce a particular product's cost per unit, which is possible when sales increases are larger than fixed-cost increases. Scale economies in a product line allow competitors to increase unit margins by selling more units of the same product, reducing pressures on margins and consequently enhancing survival rates. In industries with differentiated products, this growth strategy can be complemented by pricing initiatives that optimize consumer surplus (Besanko, Dranove, & Shanley, Reference Besanko, Dranove and Shanley2000). However, the opportunity to achieve production scale economies has technical and organizational limits. In fact, scale economies in manufacturing are rarely large enough to explain the observed levels of industry concentration (Gilbert & Harris, Reference Gilbert and Harris1984). This limitation is not as strong for intangible assets like brands, which are not subject to scale diseconomies, allowing larger firms to maintain competitive advantages of scale (Besanko et al., Reference Besanko, Dranove and Shanley2000).
Breadth competition is an alternative size strategy to scale competition (Bhaskarabhatla & Klepper, Reference Bhaskarabhatla and Klepper2014; Buenstorf & Klepper, Reference Buenstorf and Klepper2010; Dowell, Reference Dowell2006). Breadth competition takes advantage of the fact that a market always has multiple submarkets. A submarket is a subgroup of products with homogenous tangible characteristics (Klepper & Thompson, Reference Klepper and Thompson2006). Markets are often composed of different varieties of a product, which reflect the variation in users’ needs and preferences (Buenstorf & Klepper, Reference Buenstorf and Klepper2010). The decision to compete by increasing the number of products in submarkets eventually leads to competitive advantages based on differentiation, scope economies, and deterrence (Barroso & Giarratana, Reference Barroso and Giarratana2013; Sorenson, Reference Sorenson2000). R&D expenditures and brand expenses are often strong sources of scope economies, helping larger companies achieve a superior competitive position and lowering unit costs. In addition, as market size increases, adding product lines generates other advantages, such as addressing a wider range of consumer preferences and, consequently, increasing market share and prices (Bayus & Putis, Reference Bayus and Putsis1999; Dowell, Reference Dowell2006); allowing firms to use multimarket contact to reduce rivalry with competitors (Gimeno & Woo, Reference Gimeno and Woo1999; Schmalensee, Reference Schmalensee1978); and dissuading new competitors from entering. Studies indicate that larger firms show the highest survival rates and grow faster, in part by creating new submarkets (Bhaskarabhatla & Klepper, Reference Bhaskarabhatla and Klepper2014; Buenstorf & Klepper, Reference Buenstorf and Klepper2010).
While larger competitors occupy the center of the competitive landscape, smaller competitors face greater selection pressures. Niche competitors’ survival depends on the existence of spaces or submarkets in the competitive landscape that are not attractive to large competitors (Agarwal et al., Reference Agarwal, Sarkar and Echambadi2002). Usually, niche competitors develop specific market knowledge and reputations based on specialized assets (Baum, Reference Baum1995; Bhaskarabhatla and Klepper, Reference Bhaskarabhatla and Klepper2014; Caves & Porter, Reference Caves and Porter1977). Consequently, small players in a mature industry might not need to grow in order to survive (Agarwal et al., Reference Agarwal, Sarkar and Echambadi2002; Porter, Reference Porter1979).
Particularly difficult is the strategic position of middle-sized competitors, which lack both the cost and bargaining advantages of large competitors and the flexibility of smaller ones (Agarwal et al., Reference Agarwal, Sarkar and Echambadi2002; Klepper & Graddy, Reference Klepper and Graddy1990; Mueller, 1998). Also, while mid-sized competitors have less access to resources than their larger counterparts (Agarwal et al., Reference Agarwal, Sarkar and Echambadi2002; Baum, Reference Baum1995), lower market power (Bain, Reference Bain1956), and smaller scale and scope economies derived from intangible assets, the similarity between their strategic position and that of larger competitors means that they face the highest competitive pressure. In contrast, small firms combat their size disadvantage by occupying strategic niches and fulfilling demands left unmet by larger firms (Agarwal et al., Reference Agarwal, Sarkar and Echambadi2002). Small firms’ size does not pose a threat to their survival ‘simply because their interdependence with larger firms assumes a noncompetitive aura’ (Agarwal et al., Reference Agarwal, Sarkar and Echambadi2002: 990). This protection is not available to mid-sized competitors; as a result, these firms have the lowest survival rates (Agarwal et al., Reference Agarwal, Sarkar and Echambadi2002; Barroso & Giarratana, Reference Barroso and Giarratana2013; Baum, Reference Baum1995; Dobrev & Carroll, Reference Dobrev and Carroll2003; Klepper & Graddy, Reference Klepper and Graddy1990).
Although absolute size plays a major role in firms’ survival, relative size is important as well. In fact, many contemporary theoretical arguments about large organizations imply relative size effects (Dobrev & Carroll, Reference Dobrev and Carroll2003). At first glance, the definition of scale economies seems to assume absolute effects on survival. However, the theoretical mechanisms relating scale economies with survival imply that firms with relatively worse cost structures (smaller firms) cannot compete with firms with relatively better cost structures (larger firms), determining selection pressures. If the larger firms did not exist, then the outcomes of the smaller firms would be different (Dobrev & Carroll, Reference Dobrev and Carroll2003). Similar arguments can be made regarding breadth competition.
Empirical studies on industry structure tend to support these arguments, showing increasing consolidation as an industry evolves. Statistical distributions closely resemble the upper tail of the Log-Normal, the Pareto, and the Yule distributions, with the bulk of firms concentrated at smaller sizes together with a number of larger firms (Hannan & Carroll, Reference Hannan and Carroll1992; Klepper & Thompson, Reference Klepper and Thompson2006). Most of the papers do not report effect sizes regarding unexplained variance. Baum (Reference Baum1995) and Dobrev and Carroll (Reference Dobrev and Carroll2003), for example, report effect sizes by comparing the relative magnitude of the estimated coefficients and provide numerical evidence, based on data from a few individual firms, to illustrate what the coefficients would imply on those specific cases. Unfortunately, from the reported data, it is difficult to precisely state effect sizes for middle size competitors relatively to larger and smaller ones.
We conclude that the reported empirical patterns hold for a wide range of industries, including manufacturing (Gort & Klepper, Reference Gort and Klepper1982; Jovanovic & MacDonald, Reference Jovanovic and MacDonald1994), services (Baum, Reference Baum1995), and technology products (Agarwal et al., Reference Agarwal, Sarkar and Echambadi2002). However, these findings might not necessarily apply when some assumptions about the industry life cycle do not hold (Klepper, Reference Klepper1997). We point to a specific case: natural resource industries.
NATURAL RESOURCE INDUSTRIES IN PERSPECTIVE
We start by defining natural resource industries as industries whose main product is a commodity. Commodities remain substantially unaltered for decades and their price has no direct connection to the producer but rather depends on their intrinsic specifications. In fact, numerous ‘Commodity Exchanges’ around the world trade different commodities or their financial derivatives, offering a price for specific commodity products regardless of their producer. For example, the Chicago Board of Trade (CME) trades forest products such as lumber and pulp as well as agriculture products such as wheat, corn, soybeans, oats, and livestock. Another example is the London Metal Exchange (LME), which trades ferrous metals like aluminum, copper, and gold.
Some natural resource companies opt to produce specialties, which are commodities with certain characteristics that allow producers to charge a higher price per unit. For example, in the P&P industry, products such as coated freesheet paper, coated groundwood paper, or tissue paper are considered specialties. Even in these cases, the product value depends mostly on its intrinsic characteristics, not on the particular producer. In addition, all products with equivalent specifications in each submarket, irrespective of whether they are commodities or specialties, are perfect substitutes. Therefore, marketing expenditures are extremely low, since there is no differentiation to communicate to customers. Innovation also follows a unique path in commodity industries. Since the product undergoes minimal changes over time, most innovation happens at the level of production processes. Furthermore, in these industries, suppliers drive process innovation (Pavitt, Reference Pavitt1984). For example, the numerical control systems introduced into the P&P industry emerged from suppliers rather than industry incumbents. Therefore, technological change is a selection force that influences commodity industries very differently than it does differentiated-products industries. Firms competing in renewable natural resource industries face supply-side competition. Competitive advantages lie in supply-side isolating mechanisms, while demand-side isolating mechanisms are irrelevant from a selection perspective.
Not surprisingly, the long-term decline of prices, a fundamental evolutionary pattern for differentiated products, does not hold for natural resource industries. Rather, natural resource prices do not seem to follow a clear trend, even over a hundred-year time frame, and are subject to high price volatility (World Trade Report, 2010). Price evolution results from a combination of endogenous and exogenous factors, with patterns that differ significantly from those observed for manufacturing or technology products. Prices depend on endogenous supply decisions based on prior investments in capacity, but also on exogenous factors such as the business cycle (Pindyck & Rotemberg, Reference Pindyck and Rotemberg1990; World Trade Report, 2010).
Moreover, since commodities are homogenous assets, they can be used as the underlying assets for sophisticated financial derivatives. For instance, futures are standardized, exchange-traded contracts to buy or sell a specific quantity of a commodity at a predetermined price and at a specified time in the future. These futures are used by consumers and producers to hedge price risk. Producers can short contracts to lock in a selling price for their future production, while firms that require the commodity can take long positions on these contracts to secure a future purchase price. However, these derivatives are also traded by market speculators. Speculators buy or sell futures when they believe that commodity prices will rise or fall, respectively, assuming the price risk that hedgers try to avoid in return for a potential profit. This trading activity introduces an important exogenous component to the process of price determination. Speculators build their portfolios based on the relationship between commodity prices and the prices of other assets that depend on the US dollar, US interest rate, or other countries’ exchange rates, among other factors.
Given that competitive evolution at the product level looks very different for natural resource industries than for manufacturing or technology industries, we also expect to observe substantive differences in how size affects survival. Indeed, for differentiated products industries, brand names are some of the most valuable assets companies can hold (Batra, Lehmann, & Singh, Reference Batra, Lehmann and Singh1993; Hall, Reference Hall1993), and since these assets can be plausibly extended to different products at a marginal cost, brand leverage is a highly valuable source of consumer scope economies (Hall, Reference Hall1993; Maoz & Tybout, Reference Maoz and Tybout2002). However, in natural resource industries, in which a product's specifications determine its value, the identity of the producer becomes less relevant and opportunities to leverage brand goodwill and create consumer scope economies are almost nonexistent. R&D activities, another important source of scope economies for differentiated products, are also very limited in industries where product changes are marginal and where process innovation comes primarily from suppliers (Pavitt, Reference Pavitt1984).
TOWARDS A THEORY OF SIZE COMPETITION IN NATURAL RESOURCE INDUSTRIES
Having established the most important differences between industries that adjust to the industry life cycle and natural resource industries, we turn to the effect of size on survival. We focus on two specific dimensions that are particularly relevant to natural resource industries: scale and portfolio breadth. We define scale as the volume produced of a certain product (Mueller, Reference Mueller1997) and we define portfolio breadth as the number of different submarkets in which the company competes (Barroso & Giarratana, Reference Barroso and Giarratana2013). Our goal is to develop hypotheses regarding the effect of scale and portfolio breadth on the survival rates of companies competing in natural resource industries in order to evaluate theoretical mechanisms that act against the ‘liability of middleness’. In order to achieve this goal, we complement antecedents with two different mechanisms borrowed from the industrial organization field.
The first theoretical mechanism is sunk costs (Schmalensee, Reference Schmalensee1992; Sutton, Reference Sutton1991). Natural resource industries face exogenous and endogenous sunk costs. Exogenous sunk costs are those investments that firms realize at the moment of entering a submarket and refer mainly to setup costs and fixed outlays (Sutton, Reference Sutton1991). These sunk costs are determined by the current technological conditions of production. Endogenous sunk costs, in contrast, are adjustments that firms perform once they are already competing in a submarket. In manufacturing, consumer products, services, and technology, endogenous sunk costs are mainly related to investments in intangible assets (advertising and R&D, among others) that expand consumers’ willingness to pay (Schmalensee, Reference Schmalensee1992; Sutton, Reference Sutton1991). In natural resource industries, endogenous sunk costs are generally related to plant expansions and technological upgrades to production processes. A key concept for understanding selection forces is the idea that endogenous sunk costs eventually vary with market size, while exogenous sunk costs have a fixed magnitude irrespective of changes in market size, though both are fixed with respect to output (Dick, Reference Dick2004). Sunk costs, either exogenous or endogenous, allow for cost efficiencies but generate liabilities if price reductions bring prices below average costs.
Our second theoretical mechanism, price volatility, closely interacts with the first. In natural resource industries, price formation depends simultaneously on product demand and supply and on financial derivatives trading activity. Since commodities are also financial assets, their prices have an important exogenous component, as explained in the previous section. In the face of considerable sunk costs, the exogenous component of prices adds a layer of uncertainty to competitive moves.
These two mechanisms operate both at the product-line and at the firm level. Given the absence of studies addressing evolutionary patterns at the product-line level in natural resource industries, we first set selection conditions at this level and next explore the effects of portfolio breadth on firm survival.
Scale and Product-line Survival
We claim that the effect of scale on survival is the combination of a latent cost function and a latent benefit function. We argue that benefits that promote firm survival grow at a decreasing rate with product-line scale, while costs grow at an increasing rate with product-line scale.
The latent benefit function
Scale economies increment the viability of the firm at the product level: as a firm increases the output of a certain product, it can eventually benefit from cost advantages (Clark, Reference Clark1988; Josefy et al., Reference Josefy, Kuban, Ireland and Hitt2015; Mueller, Reference Mueller1997). For scale economies to exist, technological conditions must create an inverse relationship between the number of units produced and unit cost. This happens when firms require fixed assets that are independent of their output level. It is expected that, as the proportion of fixed assets increases, selection forces will favor larger firms, simultaneously determining competitors’ minimum efficient scale. In natural resource industries, scale economies related to initial setup costs and plant investments determine exogenous sunk costs for every product. As a submarket grows, the relative magnitude of initial investments decreases. Therefore, the effect of exogenous sunk costs on selection is mainly that of an entry barrier (Schamlensee, Reference Schmalensee1992).
Certain additional effects favor the survival of larger competitors beyond the minimum efficient scale threshold. First is the possibility of incrementing investments to follow market growth. Also, production facilities need regular technological upgrades to remain competitive at the cost level. Both types of investments lead to the existence of endogenous sunk costs that reinforce larger competitors’ advantages.
Scale also confers several other advantages at the product-line level by making learning economies possible. Larger product lines reinforce their cost advantages through process repetition (Argote & Epple, Reference Argote and Epple1990). Early entrance also provides an advantage for learning economies if the firm scales up at a rapid pace (Sutton, Reference Sutton1991). Scale promotes further gains due to in-depth employee specialization based on the division of labor, together with the use of specialized manufacturing equipment and reduced overhead costs (Dobrev & Carroll, Reference Dobrev and Carroll2003).
Therefore, the viability of a product line in a submarket depends on the latent benefit function that relates scale with survival, increasing the likelihood of survival for larger firms. The existence of an industry with scale economies imposes competitive pressures on small product lines that fail to reach the minimum efficient scale of the submarket. Small product lines operating at a level of output below the minimum efficient scale are sub-optimal, since their average costs exceed those of larger, more efficient, product lines (Audretsch, Houweling, & Thurik, Reference Audretsch, Houweling and Thurik2000). The term ‘sub-optimal’ describes a condition in which some plants are too small to be efficient (Weiss, Reference Weiss, Audretsch and Yamawaki1991). These small product lines are also in a particularly vulnerable position when prices fall considerably and, consequently, cost pressures peak. In natural resource industries, where prices oscillate more than in other industries, the existence of a minimum efficient scale becomes a strong liability for smaller firms. In that sense, these lines (or companies, if they operate in just one line) operate near the extinction boundary, where a random shock that barely affects large product lines increases the likelihood of failure for their smaller counterparts.
The latent cost function
However, scale has been proved insufficient to explain the levels of concentration at the submarket level (Sutton, Reference Sutton1991). In every submarket, scale economies operate in tandem with the liability of sunk costs, given the risk of excess capacity (Sutton, Reference Sutton1991). Price volatility in natural resource industries makes it harder for the largest product lines to survive. This occurs because, in industries with increasing returns to scale, firms compete over both the amount and the timing of new capital construction (Gilbert & Harris, Reference Gilbert and Harris1984). Once a plant has reached its full capacity, adding further increments usually requires a large investment based on estimates of future demand. Firms that build capacity ahead of demand risk carrying excess capacity when prices drop (Besanko et al., Reference Besanko, Dranove and Shanley2000). Fluctuating prices are the norm in natural resource industries (Cuddington & Jerret, Reference Cuddington and Jerrett2008; Erten & Ocampo, Reference Erten and Ocampo2012; Jacks, Reference Jacks2013), and their negative effects on survival are greater for competitors that make larger investments (Gilbert & Harris, Reference Gilbert and Harris1984).
Selection pressures
In other industries, intangible assets generate the opportunity to compete through endogenous sunk costs that expand consumers’ willingness to pay (Schmalensee, Reference Schmalensee1992; Sutton, Reference Sutton1991). Sunk costs, either exogenous or endogenous, allow for cost efficiencies but generate liabilities if price reductions bring prices below average costs. However, endogenous sunk costs generate much higher levels of concentration than exogenous costs given their relationship with market growth (Schmalensee, Reference Schmalensee1992; Sutton, Reference Sutton1991). In natural resource industries, which are dominated by exogenous sunk costs, this imposes a ceiling on the cost advantage that the largest competitors can develop and an important cost liability when prices drop at the product-line level. Therefore, we posit:
Hypothesis 1:
In natural resource industries, scale and product-line survival have an inverted U-shaped relationship.
Breadth and Firm Survival
We next shift from the conditions that determine the viability of a particular product line to the conditions determining the survival of entire firms. Following a similar structure to that in the previous section, we claim that the effect of portfolio breadth on survival is the combination of a latent cost function and a latent benefit function. We argue that benefits that promote firm survival grow at a decreasing rate with portfolio breadth, while costs grow at an increasing rate. Table 2 summarizes the mechanisms that shape the two latent functions. The table lists the mechanisms, their linear effects (first order effects, i.e., how breadth affects benefits or costs through each mechanism), and their quadratic effects (second order effects, i.e., the rates of change of the linear effects).
Table 2. Linear and quadratic effects of the mechanisms affecting the latent benefit and cost functions
The latent benefit function
Firms reap several benefits from incrementing portfolio breadth. First, the shared use of intangible fixed assets allows for scope economies (Hashai, Reference Hashai2015; Tanriverdi & Lee, Reference Tanriverdi and Lee2008). Scope economies are similar to scale economies in the sense that they arise as a consequence of excess capacity in a context of asset indivisibility, when diverse product lines allow firms to better utilize sunk investments (Bailey & Friedlaender, Reference Bailey and Friedlaender1982). For scope economies to exist, excess asset capacity must be accessible to different product lines that serve different submarkets. However, the benefits of scope economies are limited for two reasons. First, the impact of production scope economies in natural resource industries is low because of the low importance of intangible assets. Second, in any industry, as the number of product lines increases, it becomes increasingly difficult to share resources across lines (Stern & Henderson, Reference Stern and Henderson2004). Therefore, scope economies add value at low levels of portfolio breadth but become less relevant as breadth increases.
Since competing in natural resource industries requires nontrivial sunk costs and prices oscillate with imperfect correlation across product lines (Jacks, Reference Jacks2013), portfolio breadth creates particularly valuable advantages related to cash flow management. A broader portfolio improves the relationship between revenues and costs (Nicholson & Stephenson, Reference Nicholson and Stephenson2015), since competing in multiple submarkets enables firms to compensate for losses in one submarket with gains in other submarkets, stabilizing cash flows. However, as a firm's portfolio grows more diversified, stabilizing cash flows becomes more difficult. For cash flow stabilization to occur, a newly added product line must have a different cash flow pattern from existing product lines. The broader the product portfolio, the more difficult it is to find a new product line that is differentiated from existing ones and use it to manage cash flows.
In the long term, a broader portfolio provides additional benefits stemming from operations management. The existence of multiple products can help mitigate inefficiencies (Arango & Moxnes, Reference Arango and Moxnes2012). Finally, entering different submarkets might decrease industry rivalry in cases where multimarket contact creates a credible threat of retaliation, dissuading potential competitors from entering the submarket (Gimeno & Woo, Reference Gimeno and Woo1999). Larger incumbents can deter new entrants with the threat of incrementing production capacity, further diminishing selection pressures. Therefore, the existence of a credible threat of investment might increase larger competitors’ probability of survival. However, beyond a certain level of multimarket contact, these advantages tend to decrease (Anand, Mesquita, & Vassolo, Reference Anand, Mesquita and Vassolo2009).
In addition, large companies with a presence in different locations diversify location-specific risks. In agriculture, for example, having a presence in different locations helps to minimize weather risk. In mining, it helps to deal with environmental and social issues that might temporarily constrain production. However, these advantages also decrease as a firm's presence in different locations increases beyond a certain point. Finally, larger natural resource companies, like large companies in any industry, usually have more bargaining power over suppliers, allowing them to negotiate lower prices on critical inputs and, in turn, lower their minimum efficient scale. In addition, larger firms often receive favorable treatment from regulators. Overall, all of these mechanisms reduce selection pressures on larger competitors while negatively affecting smaller ones.
The latent cost function
The process of increasing portfolio breadth generates important coordination costs. Coordination costs are related to the creation and sharing of resources among different product lines (Hashai, Reference Hashai2015) as well as to managing excess capacity. To share knowledge and resources across multiple plants, all plants must perform similar tasks (Audia, Sorenson, & Hage, Reference Audia, Sorenson, Hage, Baum and Greve2001), which is not the case for plants operating in different submarkets in natural resource industries. Moreover, the relatively low positive effect of production scope economies in natural resource industries limits the benefits of resource sharing.
In natural resource industries, coordination costs emerge mainly from managerial optimization of two sources of endogenous sunk costs: plant expansions and technological upgrades. As markets grow, firms need to make further investments in plant expansions. Due to the magnitude of these endogenous sunk costs, they require significant managerial attention. For example, in the P&P industry, plant expansion costs account for up to 30% of the initial plant cost. As companies increase their number of product lines in different submarkets, they face increasing organizational coordination costs to avoid excess capacity in some of those lines. Additionally, cost competition depends on production efficiency, which in turn depends on regular technological upgrades. These upgrades are the result of suppliers’ R&D activities on process equipment (Pavitt, Reference Pavitt1984) and also imply substantial investments.
To coordinate the operations of multiple units and maintain consistency across locations, multiunit organizations add layers of managerial staff for coordination and control (Audia et al., Reference Audia, Sorenson, Hage, Baum and Greve2001). These complex bureaucratic structures allow multiunit firms to operate effectively, but they can also inhibit the organization's ability to adapt to shifts in environmental conditions (Hannan & Freeman, Reference Hannan and Freeman1984), such as a sudden shift in price levels. Particularly important is the need to create and maintain effective communication, information processing, and decision-making mechanisms to manage joint planning and scheduling (Hashai, Reference Hashai2015) and coordinate expansion investments and technological upgrades. Larger organizations competing in multiple submarkets face higher organizational difficulties in anticipating shifts in environmental conditions due to more complicated long-term planning processes. For instance, they take longer to approve budgets and plan for expansions. Also, time-consuming compromises between departments simultaneously requesting increased capacity and technological investments introduce political interests into the decision-making process of highly diversified companies (March & Olsen, Reference March, Olsen, Christensen and Cohen1976). Finally, competing in multiple submarkets also has a hidden cost: losing the focus required to adequately meet the needs of each specific submarket (Kekre & Srinivasan, Reference Kekre and Srinivasan1990; Ocasio, Reference Ocasio1997).
In sum, as the number of product lines increases, coordination costs grow at an increasing rate. This deduction holds even for industries with intangible assets, where it has been argued that coordination costs across multiple related product categories lead to diseconomies in managing operations and impose ineffective control and governance mechanisms (Hashai, Reference Hashai2015). In natural resource industries, which face supply-side competition and have few intangible assets, the net effect of these costs relative to the benefits is expected to be even higher.
Selection pressures
The combination of the latent cost and benefit functions described above imposes non-linear selection pressures on firm survival, depending on the level of portfolio breadth. We conclude that the latent benefit function increases at a decreasing rate, while the latent cost function increases at an increasing rate. Combining these two latent functions allow us to hypothesize that:
Hypothesis 2:
In natural resource industries, portfolio breadth and firm survival have an inverted U-shaped relationship.
METHODS
The Pulp and Paper Industry
We select the P&P industry in the United States during the period 1970–2000 to test our hypotheses. Pulping is the process by which cellulose fiber is extracted from wood and papermaking is the process that transforms pulp into paper. Broadly speaking, papermaking involves removing cellulose from the lignin matrix and forming its fibers into a web of paper. Several characteristics favor the selection of this empirical setting. The P&P industry is one of the largest and most important capital-intensive sectors in the world, as measured by average investment per employee, with global annual revenue that exceeded USD500 billion on more than 300 million tons of product in the year 2000 (USEPA, 2000). In addition, the P&P industry is a technology absorber rather than a developer. This provides an ideal setting to examine the effect of sunk costs. Our sample focuses on the United States because a third of worldwide P&P revenues can be attributed to this country.
Our chosen setting is also adequate to address scale and breadth effects since it exhibits persistently high heterogeneity in product-line scale and submarket presence, ranging from small, single-product firms with a production capacity of less than 2,000 tons/year to very large, diversified firms with a capacity of more than 12 million tons/year (more than 6,000 times larger than the smallest firm).
Firms compete in 13 different submarkets that remain active throughout our period of interest, although their production levels vary over time. These submarkets are defined such that products with similar specifications are perfect substitutes within each submarket but not across different submarkets. Moreover, the conversion processes and technical specifications of individual product lines are similar within submarkets but different between them. These 13 submarkets are: pulp, newsprint, tissue paper, special and industrial papers, linerboard, corrugated medium, solid bleached board, recovered board, coated freesheet paper, coated grownwood paper, kraft paper, uncoated freesheet paper, and uncoated grownwood paper. It is worth noting that pulp is the primary raw material for paper production. P&P production can be either vertically integrated in a single mill or separated across two or more mills.
Sample and Sources
We gathered firm-level panel data for the US P&P industry in 1970–2000 from two main sources. Our primary source was the FPL-UW database housed at the USDA (U.S. Department of Agriculture) in Madison, Wisconsin, in collaboration with the University of Wisconsin-Madison. This dataset contains estimates of the annual production capacity for all mill locations in the US. We also relied on company reports from the 100 largest US P&P firms, published annually by Scandinavian Pulp & Paper Reports. These documents provide key historical data for the firms.
We used several sources to complement our primary data. US P&P industry-level data were collected from various resources, including Pulp & Paper Magazine, which contains detailed information on the main grade categories of US P&P production. We also used the American Fact & Price Book published annually by American Paper and Pulp International, which contains detailed information at the industry and country level. The FAOSTAT database was also used to gather complementary information on the US P&P industry from 1970 onward.
We tracked detailed data on entry, exit, size, and product portfolios each year for our entire sample period. During that period, the number of P&P firms in the US decreased from 300 in 1970 to 234 in 2000, while average production capacity increased from 187,000 to 434,000 tons per firm. Overall industry capacity increased by 82% during the study, from 56 million to 102 million tons; however, at the submarket level, the 13 principal commodities show substantially different patterns of growth.
Table 3 presents, for each product line: (i) the total number of firms present at any point during the sample period, (ii) the average number of firms per year, (iii) the mean annual number of new entrants, (iv) the mean annual number of exits, (v) the average product-line capacity per year, (vi) its standard deviation, and (vii) its total growth during the sample period. While the capacity of some submarkets, such as coated freesheet paper, increased more than 155% over the period of interest, others, such as kraft paper, exhibited capacity decreases of more than 35%. These data allow us to control for submarket growth, among other variables, in our regression models.
Table 3. Submarket descriptive statistics
Note: All firms with five or more full observations, 1970–2000.
Dependent Variables
We measure survival at the product-line and firm levels. Therefore, we have two different dependent variables. Product-line survival relates to the decision of a particular firm to continue competing in a given submarket. If the firm exits that submarket, this is considered a failure, even if the firm continues to operate in other submarkets. Firm survival relates to the survival of the overall firm. In this case, an exit occurs if the whole firm ceases to exist and stops producing in all submarkets.
In both cases, we focus on the survival function S(t) = Pr{T > t} and the hazard function λ(t) = f(t)/S(t), where T is a continuous non-negative random variable denoting a survival time with probability density function f(t). Therefore, the survival function S(t) represents the probability of being in the submarket or industry at time t and the hazard function λ(t) denotes the instantaneous rate of occurrence of the exit event.
Main Covariates
The two constructs that we measure are product-line scale at the submarket level and portfolio breadth at the firm level. Since we have precise information on the production levels for each submarket and firm, we define the variable product-line scale at the submarket level as the natural logarithm of the production capacity a firm has in a given submarket during a year. Formally, for each firm i at year t operating in submarket s:
$$Product{\hyphen} line\; Scale_{i,t,s} = Ln\lpar {Cap.\; of\; Product \; Line\; s_{i,t}} \rpar \; \; \; \; \; \; \; \forall i,t,s$$The definition of portfolio breadth can be challenging for differentiated products industries since companies sell different products in the same submarket (e.g., Barroso & Giarratana, Reference Barroso and Giarratana2013). For the P&P industry, however, this definition is straightforward, since, by definition, each submarket is associated with only one product line. We measure portfolio breadth as the inverse of the sum of the squares of the capacity shares of each submarket:
$$Portfolio\; Breadth_{i,t} = 1/\mathop \sum \limits_{s = 1}^{13} \left( {\displaystyle{{Product{\hyphen}line\; Scale_{i,t,s}} \over {Firm\; Scale_{i,t}}}} \right)^2\; \; \; \; \; \; \forall i,t$$We use the inverse of this sum so that the variable portfolio breadth takes high values for larger firms competing in many submarkets, and low values for smaller firms competing in few submarkets.
Control Variables
We have several control variables. Some controls are directly related to the firm, others are specific to each product line, and others are related to macroeconomic and environmental conditions. Some of these controls are used in the firm-level model, some are used in the submarket specification, and some appear in both.
Firm age is an important control variable for survival at both the firm and submarket levels, controlling for the liability of newness (Freeman, Carroll, & Hannan, Reference Freeman, Carroll and Hannan1983) and for first-mover advantages in different submarkets (Sutton, Reference Sutton1991). We define it in two different ways. For firms that appear in the database for the first time after 1970 (approximately half of the firms in our sample), we can define age as the difference between current and entry year. For companies that existed in 1970, which correspond to a mix of new entrants from that year plus survivors from earlier years, we use the difference between current and startup year, which was hand collected.
Previous studies find that firm growth affects survival and is usually negatively correlated with the hazard rate (Dunne & Hughes, Reference Dunne and Hughes1994; Evans, Reference Evans1987). In the P&P industry, changes in scale could be the result of lumpy investments. Therefore, we include at the firm level the control variable firm growth, computed as the 1-year annual difference in firm scale. Similarly, at the submarket level we include the control variable product-line growth, defined as the 1-year annual difference in product-line scale.
At the firm level, the degree of vertical integration also affects survival. Pulp is a common input for all other submarkets and 15.3% of firms are vertically integrated in the pulp submarket during one or more periods within the sample timeframe. Vertical integration might have positive or negative effects on survival at the firm level, depending on the existence of transaction costs. We measure it using an indicator variable that takes the value 1 for firm-periods in which a firm has positive product-line scale in the pulp submarket and at least one additional submarket, and zero otherwise.
Finally, at the submarket level we also control for firm exit, which is a binary variable that takes the value 1 when a firm ceases to exist. We expect firm exit to be positively related to exit in the survival model at the submarket level.
We also use several variables to control for environmental conditions at the firm and submarket levels. Rivalry levels affect firms’ survival since they increase competition for resources (Baum, Reference Baum1995). To capture the possible impact of industry concentration on survival, we control for market size, computed as the sum of the firm-level capacity of all firms during any given year, in millions. Additionally, we use a normalized Herfindahl index, computed at the submarket and industry level depending on the survival model being tested. For the submarket level:
$$\eqalign{& Norm.Herfindal_{\; s,t} = \left[ {\mathop \sum \limits_{i = 1}^{NFirms_{s,t}} {\lpar {Mkt.Share_{i,s,t}} \rpar }^2-\displaystyle{1 \over {Num.Firms_{s,t}}}} \right] \cr&\qquad\qquad\qquad\qquad{\hskip-3pt} / \left[ {1-\displaystyle{1 \over {Num.Firms_{s,t}}}} \right]\; \; \; \; \; \forall s,t$$where mkt. share is computed as the capacity of the firm in submarket s divided by the capacity of all firms in that submarket at period t. Num. firms is the total number of companies competing in submarket s at period t. Normalized Herfindahl at the industry level is defined analogously but using the markethare and the total number of firms at the industry level. Our results are robust to the use of a regular Herfindahl index rather than a normalized one.
The technological conditions of production alter competitive pressures, enabling new technology adopters to develop cost advantages. Therefore, we include the variable machine speed change at the firm and submarket levels to capture the possible impact of exogenous technological change on the rivalry level of the P&P industry. Improvements in production technologies decrease average costs but increase endogenous sunk costs, eventually increasing competitive pressures and diminishing survival rates. This variable is computed as the biannual change in paper machine speed over the period 1970–2000.[Footnote 1]
Macroeconomic conditions also affect firm- and submarket-level survival since they alter resource munificence. Expansionary periods augment resource munificence, diminishing selection pressures, while recessions have the opposite effect. Therefore, we include the variable gross domestic product growth (GDP growth) as an additional control variable. We compute GDP growth as the annual growth rate of GDP per capita based on purchasing power parity in the United States. We obtain this data from the Maddison Project dataset described in Bolt and van Zanden (Reference Bolt and van Zanden2014). We expect this variable to be positively associated with survival.
Finally, fluctuating prices are the norm in natural resource industries (Cuddington & Jerret, Reference Cuddington and Jerrett2008; Erten & Ocampo, Reference Erten and Ocampo2012; Jacks, Reference Jacks2013), and we posit that their negative effects on survival are greater for smaller and larger competitors. Therefore, we include the variable price change as an additional control. We proxy for this variable using the annual changes in the Producer Price Index by Commodity for Pulp, Paper, and Allied Products, retrieved from the Federal Reserve Economic Data database maintained by the Federal Reserve Bank of St. Louis.
Estimation Technique
We test our hypotheses using a parametric survival model. A fundamental question concerns assumptions regarding error terms. Following the Akaike information criterion (AIC), we let T, the continuous non-negative random variable representing firm survival time, follow a Weibull distribution with parameters λ and p. Under this distribution, the hazard rate is λ(t) = λ p t p−1; thus, the survival function is S(t) = exp{−λ t p} and the density is f(t) = exp{−λ t p} λ p t p−1. We perform a maximum likelihood estimation with robust standard errors for the accelerated failure-time model and let the natural logarithm of the parameter λ depend on different subsets of the covariates while simultaneously estimating the natural logarithm of the parameter p. Results, not reported here, are robust for different specifications of survival time distribution.
RESULTS
Summary Statistics
The left-hand side of Table 4 shows descriptive statistics for our main variables at the firm level. This table includes panel information for the 386 firms with 5 or more full observations during the sample period (1970–2000). Of these firms, 103 are incumbents that remain in the industry throughout the sample period, while 120 are incumbents that exit the industry before 2000. 86 are new entrants that remain in the industry through 2000 and 77 are companies that enter and exit the industry during the sample period. On average, we have 17.25 annual observations per firm.
Table 4. Firm descriptive statistics and pairwise correlation matrix
Notes: *** p < 0.001, ** p < 0.01, * p < 0.05, † p < 0.1.
All firms with five or more full observations, 1970–2000. The number of observations for each variable is 6,659.
The right-hand side of the table shows pairwise correlations for the main variables. Most correlations have low magnitudes and the expected signs. However, breadth and the control vertical integration have a significant correlation of 64.7%. This is unsurprising given that firms operating in more submarkets are more likely to be vertically integrated. Additionally, market size and normalized Herfindahl have a significant correlation of 88.1%, reinforcing the fact that they are both measures of industry concentration. Finally, market size and machine speed change are also significantly correlated at 39.6%. Our results in the following sections are robust to the removal of vertical integration and market size from the control variables.
Regression Results
Table 5 shows the results of a maximum likelihood parameter estimation of the survival model at the submarket level. A positive sign on an estimated coefficient relates to a higher probability of survival.
Table 5. Survival model at the submarket level
Notes: Robust standard errors in parentheses.
*** p < 0.001, ** p < 0.01, * p < 0.05, † p < 0.1.
Time-varying covariates. All firms with five or more full observations, 1971–2000.
We test H1 by introducing the main covariate product-line scale. Results show that product-line scale has a positive and significant coefficient in the quadratic model specification (model 3). Its regression coefficient is 0.871 and is significant at the 0.1% level. The product-line scale squared term is also significant at the 0.1% level, but has a different sign and magnitude, suggesting an inverted U-shaped relationship with survival. The likelihood ratio test shows that the quadratic model (model 3) fits the data significantly better than the linear model (model 2), providing further support for the inverted U-shaped relationship. We also test whether the slope at each end of the data range is sufficiently steep and whether the turning point of the inverted U-shaped relationship is located within the range of the scale variable (Haans et al., Reference Haans, Pieters and He2015; Lind & Mehlum, Reference Lind and Mehlum2010). Untabulated results confirm that the slopes at the low and the high ends of the data range for the scale variable are significantly positive and negative, respectively, at the 1% level. Additionally, the 99% confidence interval around the turning point is within the data range for scale. These findings provide robust evidence of an inverted U-shaped relationship between scale and survival at the submarket level, strongly supporting H1.
Figure 1A illustrates the previously described inverted U-shaped relationship. This panel plots the log of the product-line survival time as a quadratic function of product-line scale in the relevant range, which goes from 0.59 to 8.26. The equation coefficients are derived from model 3 in Table 5. As the curve illustrates, our results show that the relationship between scale and survival rate is non-linear and favors mid-sized lines over their smaller and larger counterparts.
Figure 1. Implied relationship between survival, scale, and breadth
Figure 2A shows how the product-line survival distribution evolves through time for different values of product-line scale, considering the mean values of the control variables. The blue and green lines depict the survival distributions for the largest and smallest product lines in the sample, respectively. The red line shows the survival distribution for a mid-sized product-line, measured as the average of the smallest and largest lines.
Figure 2. Survival distribution, scale, and breadth
Besides, we obtain the effect size of the relationship between product-line scale and product-line survival rate through time. For example, after 10 years in the sample, the survival rates for the largest, smallest, and mid-size product-line scale are 69.4%, 30.2%, and 88.2%, respectively. After 20 years, the survival rates are 46.9%, 8.4%, and 77.1%, respectively. As the curves illustrate, these results are substantial and further emphasize that the relationship between product-line scale and survival rate favors mid-sized companies over their smaller and larger competitors.
Finally, we examine a quadratic model that interacts the main covariate product-line scale and its squared term with the control variable price change (model 4). Results show that the coefficients on product-line scale and its squared term are statistically significant and have the same signs as they do in the quadratic model without interactions (model 3), evidencing an inverted U-shaped relationship between scale and survival. Moreover, the coefficients on the interaction terms product-line scale × price change and product-line scale squared × price change are also significant and have the opposite signs of the non-interacted terms. Thus, when prices decrease (increase), the inverted U-shaped relationship between scale and survival becomes more (less) pronounced, enhancing (lessening) the probability of survival of mid-sized firms compared to their smaller and larger counterparts. These findings provide further support for H1.
Table 6 shows the results of a maximum likelihood parameter estimation of the survival model at the firm level. As at the submarket level, a positive sign on an estimated coefficient relates to a higher probability of survival.
Table 6. Survival model at the firm level
Notes: Robust standard errors in parentheses.
*** p < 0.001, ** p < 0.01, * p < 0.05, † p < 0.1.
Time-varying covariates. All firms with five or more full observations, 1971–2000.
We test H2 by introducing the main covariate portfolio breadth. This covariate has a positive and significant coefficient in both the linear and quadratic specifications presented in Table 6 (models 2 and 3). In the quadratic model (model 3), we also include portfolio breadth squared. A likelihood ratio test shows that this specification fits the data significantly better than the linear one (model 2). Portfolio breadth and its squared term have significant coefficients with different signs. We also measure the effect size in terms of explained variance using Royston and Sauerbrei (Reference Royston and Sauerbrei2004)’s R-squared. This measure is 52.7% for the model that only has controls (model 1) and 64% for the quadratic model (model 3); which gives an increase of 11.3% in explained variance.
Further, the slopes at the low and the high ends of the data range for the variable are significantly positive and negative, respectively, at the 0.1% level for the low end and 8.4% level for the high end (untabulated). Finally, the 99% confidence interval around the turning point is within the data range for portfolio breadth (untabulated). These findings provide robust evidence of an inverted U-shaped relationship between portfolio breadth and survival at the firm level, strongly supporting H2.
Figure 1B illustrates the relationship between portfolio breadth and survival. The figure outlines the log of the firm survival time as a quadratic function of portfolio breadth in the relevant range, which goes from 1 to 8.45. The equation coefficients are derived from column 3 of Table 6. The figure illustrates our results, which depict an inverted U-shaped relationship between portfolio breadth and survival, with moderately-diversified companies (within their industry) having higher survival rates than their smaller and larger competitors.
Additionally, Figure 2B shows how the firm survival distribution evolves through time for different values of portfolio breadth, considering the mean values of the control variables. The blue and green lines depict the survival distributions for the firms with the smallest and largest portfolio breadth in the sample, respectively. The red line shows the survival distribution for a firm with a medium portfolio breadth, measured as the average of the smallest and largest breadths.
We can use these curves to illustrate the effect size of the relationship between portfolio breadth and firm survival rate through time. For example, after 10 years in the sample, the survival rates for the largest, smallest, and mid-size portfolio breadth are 88.9%, 82.2%, and 95.9%, respectively. After 20 years, the survival rates are 71.7%, 57.4%, and 88.9%, respectively. As the curves illustrate, these results are substantial and further emphasize that the relationship between portfolio breadth and firm survival rate favors mid-sized companies over their smaller and larger counterparts.
Finally, we examine a quadratic model that interacts the control variable price change with both the main covariate portfolio breadth and its squared term (model 4). Results show that the coefficients on portfolio breadth and portfolio breadth squared are significant and have the same signs as they do in the quadratic model (model 3). Moreover, the coefficients on the interaction terms are also significant and have the opposite signs of the non-interacted terms. Hence, when prices decrease (increase), the U-shaped relationship between breadth and survival becomes more (less) prominent, enhancing (lessening) the probability of survival of mid-sized firms compared to their smaller and larger competitors. These findings provide further support for H2.
Robustness Checks
There are at least three potential sources of bias in this study: the existence of outliers, the process of mergers and acquisitions (M&A), and the measure that we use for within-industry diversification. Overall, our main results remain unchanged after accounting for each of these potential biases.
A first concern is that the curvilinear effect observed in our main results may be generated by the existence of outliers. To address this potential problem, in untabulated results we perform a 90% winsorization of our main covariates: product-line scale and portfolio breadth. This process sets the lowest and highest tail values to the values corresponding to the 5th and 95th percentiles, respectively. Our main conclusions remain unchanged; coefficients continue to be significant and have the correct signs, although their magnitudes are different.
Second, in untabulated results we also address potential biases that might emerge from M&A activity. Several adjustments have been suggested for these cases, none of them completely free of additional problems (Dunne & Hughes, Reference Dunne and Hughes1994). We opt to test the robustness of our findings by rerunning the econometric analysis after deleting acquired firms from the sample. To detect M&A activity, we gather M&A data from the SDC Platinum database. We analyze all available transactions in which the target's SIC code belongs to the P&P industry and the percentage of shares owned by the acquirer after the transaction is at least 50%. Then, we match each of the 197 non-survivors in our sample to the M&A database to identify which companies exited the sample due to an acquisition. We identify 26 out of 197 companies as acquisition targets. Our main findings remain unaltered: the U-shaped relationships persist after dropping acquired firms from the sample. Coefficients have the correct signs, although they are different in magnitude and slightly less significant in some specifications.
Finally, we address the potential limitations of our measure of portfolio breadth by rerunning the analysis with an alternative definition: we simply count the number of submarkets in which the company competes (out of the 13 total submarkets in the industry). Formally:
$$Portfolio\; Breadth_{i,t} = Number\; of\; Submarkets\; with\; Positive\; Product{\hyphen}line\; Scale_{i,t}\; \; \; \; \forall i,t$$Our main results remain unaltered; there are changes in the magnitudes of the coefficients, but they remain significant and have the same signs (results not reported here but available upon request).
DISCUSSION AND CONCLUSION
Firm size is a critical dimension to consider when analyzing survival. A large body of literature on differentiated products concludes that large competitors occupy the center of an industry while small competitors gravitate toward niche positions (Agarwal et al., 1995, Reference Agarwal, Sarkar and Echambadi2002; Hannan, Pólos, & Carroll, Reference Hannan, Pólos and Carroll2007; Klepper, Reference Klepper1997). Thus, such industries are comprised primarily of large competitors and niche players. Most of these studies find mid-sized companies to be the worst performers, exhibiting the highest likelihood of abandoning the industry. Traditional theoretical and empirical work has referred to this phenomenon as the ‘liability of middleness’.
In this study, we theoretically question the ‘liability of middleness’ for firms competing in natural resource industries; our empirical findings provide strong support for our theoretical claim. We observe lower mortality rates for mid-sized companies compared to their larger and smaller competitors. These results are consistent when analyzing scale at the submarket level and breadth at the firm level. The effect size in terms of explained variance at the firm level indicates that our quadratic model explains 11.3% more of the variance than the base model that only includes controls. We also show that the effect size of the relationship between scale and product-line survival as well as portfolio breadth and firm survival indicates substantial advantages for middle size competitors. For example, after 20 years in the sample, the survival rate of a middle size company is 31.5% larger than that of the smallest competitor and 17.2% larger than the largest competitor. We conclude that size has a different effect on firms competing in natural resource industries than it does on firms in industries with differentiated products. Specifically, mid-sized companies occupy a superior competitive position in natural resource industries.
Our results are not totally unexpected, given that various studies have predicted the possibility of finding different patterns of competitive dynamics in different industries. Antecedents refer to service industries, complex product and systems industries, and cultural industries as examples that seem to violate the predictions of the industry life cycle (Klepper, Reference Klepper1997; Peltoniemi, Reference Peltoniemi2011). Our contribution lies in the fact that, while all of the previous exceptions arguably fit into the differentiated products category, we focus on industries that lack product differentiation and are subject to what we define as supply-side competition.
To run the empirical tests, we must overcome an important theoretical challenge, since the constructs of scale and breadth have historically been used to derive the opposite relationships from those we predict and empirically test. This forces us to specify the role of certain mechanisms of differentiation as a source of competitive advantage. In particular, we stress the importance of intangible assets such as brands or R&D as selecting mechanisms that explain the traditional results. This allows our manuscript to make an interesting indirect contribution: it not only specifies selection mechanisms in natural resource industries, but also helps to contextualize the traditional analysis of differentiated product industries that follow the standard industry life cycle.
Scale and breadth, our specific constructs for size, generate alternative mechanisms of competitive evolution. Overall, selection results from the combination of mechanisms that determine minimum efficient scale, sunk costs/excess capacity, and price volatility in the absence of consumption scope economies. Scale directly affects unit costs, providing a competitive advantage for larger competitors that focus on a single product. In an industry where competitors are price takers, as in the P&P industry, cost advantages are fundamental for survival. However, scale has been proved insufficient to explain the levels of concentration in different industries (Gilbert & Harris, Reference Gilbert and Harris1984), mainly because most organizations also face important scale diseconomies. In fact, scale diseconomies act as a threshold in any industry. This growth ceiling selects out larger competitors and favors mid-sized ones. Our econometric analysis on the evolution of the P&P industry shows an inverse U-shaped relationship between scale and survival, supporting this claim.
Multiple factors can explain the inverse U-shaped relationship between scale and survival. For one, firms competing in renewable natural resource industries face supply-side competition. That is, they compete mostly through supply-side isolating mechanisms, while demand-side isolating mechanisms are almost absent. This has implications for coordination, capabilities, and learning. Most coordination costs faced by natural resource industries are similar to those observed in industries with product differentiation (Barroso & Giarratana, Reference Barroso and Giarratana2013; Hashai, Reference Hashai2015). However, coordination costs related to managing excess capacity are particularly salient in these industries. Additionally, in industries with differentiated products, product proliferation generates learning curves and positive synergies between a brand and a submarket niche (Barroso & Giarratana, Reference Barroso and Giarratana2013). These effects are absent in natural resource industries. Finally, in natural resource industries, the main source of competitive advantages at the capability level relates to cost advantages. The absence of intangible assets, which are prevalent in industries with product differentiation, makes certain critical capabilities irrelevant in natural resource industries. These capabilities (related to brand differentiation and R&D) are among the most important for supporting large size advantages in industries with differentiated products, but do not play a role in natural resources.
Another explanation lies in the selecting mechanism that emerges from the interaction of the intense capital investments required in these types of industries with price volatility, which generates systematic cycles of excess capacity. In particular, endogenous sunk costs have been reported to have much more concentration power than exogenous sunk costs (Dick, Reference Dick2004; Sutton, Reference Sutton1991). In fact, in industries where endogenous sunk costs dominate, it is possible to observe the co-existence of larger generalists and smaller specialists (Dick, Reference Dick2004). In our case, where endogenous sunk costs are relatively less important, it is reasonable to expect survival rates to be highest for midsize competitors.
This study has important implications for emerging economies given the greater importance of natural resource industries in their economic development. Moreover, this study has implications at the firm, industry, and country levels. The above discussion considers survival from a firm-level perspective; when more of a country's companies belong to natural resource industries, these findings also become relevant from a macroeconomic point of view. In addition, we observe significant competitive entry and exit activity in the P&P industry and expect to find similar forces at work in other natural resource industries. Conceptual frameworks at the firm level that improve the understanding of this pattern of competition will eventually reduce firm mortality rates and consequently improve countries’ economic performance. In contrast, failing to adequately address growth strategies at the firm level may result in unnecessary mortality rates at the national level. This might imply higher rates of unemployment as well as lower exploitation of resources, consequently leading to lower long-term GDP growth.
The managerial implications of our study will be useful for guiding the strategies of firms in natural resource industries when considering optimal size. Our results question both the sustainability of a niche strategy and the convenience of being large. For small companies, the imperative is to grow. However, large companies should maintain a careful balance to avoid over-investments in scale and unnecessarily broadening product portfolios. This adds complexity to managerial decisions around optimal size and creates tension for management teams when stakeholders pressure them to grow. Striking a balance between scale and breadth strategies is also nontrivial and has a relative component – firms should take competitors’ growth into account. Our study suggests that a key managerial skill is navigating this balance – weighing gains from scale in a few product lines, gains from breadth, and competitors’ decisions in both of these dimensions.
Boundary Conditions and Limitations
We set clear boundary conditions on our theoretical development. Our work determines selection mechanisms for industries that mainly produce commodities and are capital-intensive. Our theory also applies to industries that have relatively high exogenous price volatility. As we relax these assumptions and move to industries that offer differentiated products and develop intangible assets, selection pressures change and the expected inverted U-shaped relationship between portfolio breadth and survival does not necessarily hold. It is also unclear whether the effect of portfolio breadth on survival holds for industries that produce commodities but are less capital-intensive. For example, the levels of required investment to compete in the P&P industry are significantly larger than those observed in agriculture. In fact, in several countries, agriculture is a highly atomized industry despite producing commodities with similar endogenous and exogenous price volatility.
In addition, competitive evolution might look different for non-renewable natural resource industries such as oil or mining, which require substantial investments in exploration and are more sensitive to political intervention. The need for investments in exploration activities might favor the survival of larger competitors. Similarly, the need for lobbying activities in non-renewable resource industries might favor the survival of the largest companies, which usually have stronger capabilities in this regard. Additional limitations on broadly applying our findings follow from recognizing other barriers to the growth of companies in various commodity sectors, such as regulatory regimes, locations of natural resources, and the need for proximity and location benefits that limit size and impact areas.
In the P&P industry, the number of submarkets has remained largely stable during the period under analysis, which may not be the case for other natural resource industries. For example, nickel, graphite and lithium have recently experienced a proliferation of new submarkets due to demand for batteries for different products. We believe that the creation of new submarkets tends to reinforce our hypotheses, since production scope economies continue to predominate, and price volatility also remains imperfectly correlated between submarkets. As the number of submarkets in an industry grows, all of the mechanisms described in the derivation of H2 act in the same direction and with similar intensity. However, if the magnitudes of exogenous and endogenous sunk costs change, expectations about the effect of firm size on survival might change and H2 might need to be revisited.
Another particularity of the P&P industry is that the pulp submarket is different from the others; although pulp can be bought and sold like any other commodity, it is also a critical input in the production processes for the remaining submarkets. We tested for vertical integration effects and observed that they negatively affect survival. We conjecture that this result provides additional evidence of the increasing coordination costs that result from entering different submarkets.
A fundamental aspect of our theoretical derivation is the fact that firms’ sunk costs are tangible assets. However, firms also report expenditures on intangible assets (although in significantly lower amounts). As the amount of intangible assets – such as R&D for improving production processes – increases, this might favor the survival of larger firms. These types of investments are endogenous sunk costs that have proven to be a source of industry consolidation (Sutton, Reference Sutton1991).
Empirically, the decision to focus on one industry improves accuracy at the variable-building level and construct consistency but decreases generalizability. Our empirical setting allows us to obtain good proxies for exogenous and endogenous sunk costs. In addition, we have a precise identification of submarkets. However, the measures for scope economies and coordination costs are indirectly assumed in the variable that measures the degree of portfolio breadth. Therefore, the way we empirically address the different variables explaining the latent functions that relate portfolio breadth and survival results in some limitations. Future research will gain from obtaining information regarding the organizational structure and decision processes of firms competing in natural resource industries.
In spite of these limitations, our study is to our knowledge the first to address the mechanisms of competitive evolution and size in natural resource industries. Given the importance of these industries for the world economy and for the development of a large list of emerging countries, this study opens a relevant avenue for future theoretical advances in a largely unexplored setting.
