The dynamics of superior performance among the largest firms in the global oil industry, 1954–2008

1. Introduction

The measurement and explanation of superior corporate economic performance, particularly the distribution and the persistence of superior performance, is of great interest to scholars from different fields, including industrial economics, strategy, and technological change (Malerba and Orsenigo, 1996), as well as to practitioners, among which corporate executives, investors, and strategy consultants. The existing empirical evidence about the persistence of superior performance (e.g. Mueller, 1986; McNamara et al., 2003; Wiggins and Ruefli, 2005) is mixed. To contribute to the understanding of the dynamics of superior performance, this research aims at identifying and explaining three important aspects of patterns of superior performance over time. First, how does the rate of performance convergence develop over time? Second, how does the occurrence and persistence of superior performance evolve over time? Do we observe less incidences of persistent superior performance and do superior performance periods shorten over time? Third, how does the migration of firms across performance strata unfold over time? Is strata membership relatively stable or does it fluctuate?

Most “persistence of profits” studies covering the period 1974–2003 find persistence of superior performance (e.g., Mueller, 1986; Goddard and Wilson, 1996; McGahan and Porter, 1999). These studies use autoregression techniques, which largely ignore the dynamics of persistence over time. However, two more recent studies do examine the effects of time on persistence. McNamara et al. (2003), also using autoregression techniques, find cyclicality of performance in a US population over the 1978–1997 period. Wiggins and Ruefli (2005) apply non-parametric stratification techniques and find a shortening of superior performance periods in a US population¹ during the same period.

In this present study, we build on previous research (see review by Wiggins and Ruefli, 2005) and study performance dynamics of large firms in the global oil industry. We measure economic performance of firms in terms of the return on assets. We identify performance patterns over time by applying various methodologies, including autoregression tests and non-parametric stratification. In particular, this study combines different levels of analysis of the dynamics of an industry (e.g., Malerba and Orsenigo, 1996). By taking a closer look at the history of the industry and its constituent firms, we aim to present causal explanations of the observed performance patterns (Dosi et al., 1997). These empirically oriented explanations could be characterized as an appreciative theorizing attempt to contribute to the growing body of stylized facts about performance patterns (Malerba et al., 1999; Helfat, 2007).

This research investigates the global oil industry, both upstream and downstream activities from 1954 until 2008.² Various considerations motivate the choice of this industry. First, the global oil industry is unparalleled in terms of economic and geopolitical impact. The worldwide turnover of crude oil amounted to 2.9 trillion dollars in 2008 (BP, 2009). Second, the extensive research period, 1954–2008, has witnessed considerable changes of the industry. Third, notwithstanding these changes, the nature of activities has remained fairly comparable throughout the research period, making the oil industry an attractive subject for longitudinal analysis. Fourth, because of the global scope of competition in the oil industry, our study complements existing single-country performance studies, which mostly focus on the United States (see Chacar and Vissa, 2005; Wiggins and Ruefli, 2005). Finally, the oil industry is well-documented, which facilitates the causal explanations of the observed performance dynamics.

Our study contributes to prior performance research in several respects. First, the combination of parametric and non-parametric research methods applied to the same set of firms enhances our understanding of method effects. Previous performance studies largely confine their methodologies to one particular method of analysis, which makes it difficult to disentangle method effects from observed phenomena. Second, our study covers a time span of 55 years. Such a long time frame is necessary to capture the institutional change that is characteristic for the global oil industry. Moreover, a long time frame is particularly useful to identify patterns of superior performance over time, especially when these patterns are of a cyclical nature, as McNamara et al. (2003) assert. Third, we analyze an international sample of firms, while prior research focused on US samples. Because the industry has become increasingly global over the research period, we need to broaden the scope of our research from largely US-based firms to worldwide competition.

2. Background: theoretical perspectives and the recent history of the global oil industry

As a background, this section provides an overview of theoretical perspectives applied in prior empirical performance research. To facilitate the causal explanation of the performance dynamics, this section also presents a historical account of the main dynamics of the global oil industry during the research period.

2.1 Theoretical perspectives on the dynamics of superior performance

A rich literature has been developed to explain superior performance. In this section, we discuss these theoretical perspectives that have been applied in the main empirical studies on performance dynamics. Mueller (1977, 1986) adopted the economic perspective of competitive markets, wherein equilibrium firms cannot sustain systematic rents above the industry norm (e.g., Stigler, 1963). Mueller has initiated a series of empirical analyses in which the speed of transition toward such a competitive equilibrium is measured. Although Mueller’s approach is based on efficient product markets, it allows for the measurement of convergence dynamics in the industry.

Wiggins and Ruefli (2002) used the concept of sustainable competitive advantage (Porter, 1985) and found the persistence of superior performance to be most consonant with the resource-based view of the firm (e.g. Barney, 1991; Conner, 1991). In a later study, McNamara, Vaaler, and Devers (2003) tested the “hypercompetitive perspective” (D’ Aveni, 1994, 1995) but did not find support. Instead, the findings of McNamara et al. (2003) reflected a punctuated equilibria process (Tushman and Romanelli, 1985). In another study, Wiggins and Ruefli (2005) found the shortening of periods of superior performance to support Schumpeter’s theory of competitive behavior (Schumpeter, 1939). Kato and Honjo (2009) compared the Chandlerian view with the Schumpeterian view (Sutton, 2007) to explain persistence of market leadership, and concluded that persistence varies according to industry-specific characteristics.

We acknowledge the challenge of the interpretation of statistical data in the wealth of stylized facts (Dosi et al., 1997). Following Louca and Mendoca (2002), we assume that historical and appreciative theorizing based on concrete statistical information is the preferred tool for explaining the patterns of long-run performance.

2.2 Main dynamics of the global oil industry

For this historical account of the global oil industry, we combine desk and field research. Regarding the former, we build upon the rich oil industry literature. We examined scientific journals (e.g., Cibin and Grant, 1996; Grant, 2003; Woiceshyn and Daellenbach, 2005; Kilian, 2006), financial newspapers, and professional journals (e.g., Financial Times and Petroleum Intelligence Weekly). We also studied books and case studies on the industry (e.g., Hawdon, 1985; Sampson, 1988; Yergin, 1993; Van der Linde, 2000) and on individual companies (e.g., Van Zanden et al., 2007). We investigated company data (i.e., annual reports and web sites and other data, such as the BP Statistical Review) and also publications and databases provided by industry institutions (e.g., International Energy Agency, US Energy Information Administration, Chatham House, and the Baker Institute for Public Policy). To complement the desk research, we have verified our interpretations with industry experts.

In order to facilitate a causal explanation of the performance dynamics, we a priori investigate the changes of various facets of firms and the industry over the research period. By analyzing the changes of various facets over time, we distinguish phases in the development of the global oil industry with distinct characteristics. To identify these phases, we put forward the concept of competitive regimes.³ We focus on the following properties of competitive regimes: the main source of competitive advantages, firms’ strategies, and the rates of entry and exit of firms. We have identified four competitive regimes in the global oil industry during the research period (Table 1).

2.2.1 The “concession-based” regime

The concession-based regime is named after the oil production concessions granted by “crude oil reserves holding countries”, i.e., Saudi Arabia to “The Seven Sisters”, an oligopoly of seven major international oil companies (IOCs).⁴ This cartel controlled production and prices of crude oil through “horizontal integration”—almost all concessions were shared through cross-ownership by the Sisters—and through “vertical integration”—all Sisters integrated upstream and downstream activities (Stevens, 1985). Initially, the crude oil reserves holding countries were not involved in the exploration and production of their natural resources but only levied taxation on production of crude oil. The Seven Sisters kept posted prices low to reduce taxation on production (Figure 1).

In 1960, five crude oil reserves holding countries, Venezuela, Saudi Arabia, Kuwait, Iraq, and Iran, established the Organization of Petroleum Exporting Countries (OPEC) as a countervailing power to the Seven Sisters. In the 1960s, OPEC expanded and member states increased taxation. In the 1970s, the OPEC members began to unilaterally participate in the production of crude oil by nationalization or expropriation of the IOCs’ upstream assets in their countries. For instance, in 1971 Libya nationalized its oil industry, and in 1972 Iraq and Algeria followed Libya’s example.

2.2.2 The “reserves access” regime

The “reserves access” regime started in 1973⁵ when OPEC countries put an embargo on crude oil supplies to specific Western countries. The world’s first oil crisis was the result, and the (nominal) oil price soared. In the 1970s, the OPEC members further raised the oil price, they increased taxation rates, and an increasing number of OPEC countries nationalized exploration and production.⁶ The ending of concessions and the nationalization of the oil companies’ upstream assets in the OPEC countries changed the role of IOCs, in particular the Seven Sisters, from owners of the oil they found to contractors (Sampson, 1988). The Seven Sisters lost their control over OPEC supplies and over the price of crude oil (Cibin and Grant, 1996).

During the 1970s, OPEC countries coerced the Seven Sisters to produce more crude oil than their refineries could process, thereby forcing the Seven Sisters to sell surpluses of crude oil on the open market. This benefited the so-called “Independents”, i.e., IOCs not belonging to the group of Sisters. Moreover, OPEC countries increasingly bypassed the Seven Sisters by directly selling crude oil on the open markets to the Independents. The increased use of markets, as indicated by the rise of spot and future markets for crude oil, undermined the benefits of vertical integration for IOCs. After the second oil crisis (1978), the IOCs began to change their vertical integration strategies by moving toward financial integration and a greater use of markets (Stevens, 2005).

Having lost their concessions, access to crude oil reserves became a major issue for the Seven Sisters (Cibin and Grant, 1996). The sharp price rise (fueled by two oil crises) led to large financial surpluses for the IOCs. These firms massively invested their surpluses in exploration activities outside OPEC-territory as well as in diversification—including office automation and hotels—to reduce their dependence, respectively, on OPEC resources and on oil in general. However, exploration and diversification investments failed and the negative returns on these investments caused the stock market values of many oil companies to decline until below their intrinsic values. Induced by the stock market response, a merger and acquisition (M&A) wave took place among IOCs between 1979 and 1984 (e.g. Stonham, 2000).⁷ All Sisters were involved in this M&A wave, except Exxon that refrained from acquisitions because of anti-trust regulations.

As a result of the nationalizations of the upstream assets, the 1970s witnessed the rise of a new group of companies: national oil companies (NOCs) from the crude oil reserves holding countries. These state-owned and state-controlled companies had governance models that radically differed from the investor-owned IOCs. NOCs had different objectives and were—in some instances—political tools of the governments (e.g. Van der Linde, 2000). In addition to NOCs from oil reserve holding countries, the industry faced the entry of NOCs from oil importing countries. These companies typically ran a national downstream monopoly and had the objective to secure supply to their home country. Although the NOCs benefited from national monopolies on upstream and/or downstream activities, they were generally less efficient, and had to fulfill political objectives of their country’s government, such as subsidizing domestic sales of oil product.

The high oil price of the 1970s allowed the development of new non-OPEC fields in the North Sea, Alaska, and Mexico. These fields had been discovered in the 1960s, but were economically unattractive at the then low prices. The high oil price not only stimulated supply, in particular from non-OPEC sources, but also negatively influenced demand for oil products, resulting in growing oversupply. The Iranian revolution led to a second oil crisis in 1978. Following the declining crude oil prices since 1981—note that early 1981 the United States ended the oil price regulation—OPEC introduced in 1983 a quota system to stem the tide. However, OPEC never agreed on the price, and lacked a mechanism for punishing cartel members for deviation. Saudi ARAMCO—the NOC of Saudi Arabia, which was the only country not assigned a quota as it would be the “swing producer”—caused a price war as a means to regain market share and reduce the oil revenues of rival Iran, illustrating the mixture of economic and political roles a NOC may fulfill.

2.2.3 The “efficiency focus” regime

The failure of the OPEC quota system contributed to the plummeting of the oil price in 1986. This price collapse marked the start of a new competitive regime, termed “efficiency focus”. In the first and the second regime, respectively, cartels of IOCs (Seven Sisters) and crude oil reserves holding countries (OPEC) had controlled the price and volume of crude oil. In the third regime, the price and volume of crude oil trading was strongly influenced by the spot and futures markets. The oversupply of crude oil and the resulting low price discouraged exploration of new crude oil reserves and forced an efficiency focus upon the (international) oil companies (Grant, 2003). Moreover, these IOCs no longer had access to most reserves holding countries that conceded upstream monopolies to their NOCs. Without growth opportunities and without the possibility of raising prices, efficiency improvement was the only path to profitability left.

In the 1980s, shareholder value management started its rise to prominence (Copeland et al., 1990). Exxon was among the first IOC to adopt this management philosophy. Shareholder value management dictated a focus on efficiency and caused IOCs to divest assets with relatively low returns: mainly unrelated assets (result of diversification in the 1970s), and parts of downstream (e.g., Stevens, 2005, 2009; Moss, 2008). As it became clear that OPEC’s control over the industry was weaker than thought, IOCs divested unrelated businesses and refocused on their oil business. The divested downstream assets of the IOCs were often bought by NOCs from exporting countries that wanted to integrate forward into refining, marketing, and distribution to secure the demand for oil (Stonham, 2000). Several NOCs from other oil reserves holding countries (e.g., Saudi ARAMCO, PDVSA, Kuwait Petroleum Corporation) began to expand internationally. International competition further increased as Russian Federation’s (national) oil companies started to internationalize, after the ending of the Soviet Union in 1989. In the 1980s, some countries with NOCs started to deregulate their domestic oil markets and to (partly) privatize their NOCs. This regime witnessed a transformation of some NOCs that makes them more resemble the IOCs, with investor-ownership, vertical integration, and international presence.

In 1998, oil prices declined below the level of 1986 (in nominal terms), further increasing the pressure on oil companies to improve efficiency. The IOCs suffered from low returns on their investments due to the low oil price. Rather than investing, these companies bought back their shares and paid out high dividends. Having exhausted the potential to improve efficiency and confronted with the stock market crash of 1998, IOCs once again embraced the strategy of M&A to achieve further cost savings, triggering the M&A wave of 1998–2001 (Reinhardt et al., 2006). This wave led to the formation of six “Super Majors”.⁸ In 1998, BP bought Amoco and in 1999 Atlantic acquired Richfield. In 1998, following relaxation of anti-trust regulation, Exxon acquired Mobil and Chevron acquired Texaco in 2001. Royal Dutch Shell was the only Sister not to participate in this M&A wave. Examples of large acquisitions by other IOCs, not being Sisters, were Phillips Petroleum’s acquisition of Conoco, and Total’s acquisition of Elf Aquitaine and Petrofina. As a result of the M&A wave and the increased efficiency, the industry became more concentrated, less competitive, and supply became tight, due to reduced exploration activities and reduced investment in production infrastructure (Stevens, 2005).

2.2.4 The “production focus” regime

Rising demand for crude, in particular from Newly Industrialized Countries, such as China and India, in combination with tight supply at the end of the efficiency regime caused the spare capacity of OPEC and in particular Saudi Arabia, to be increasingly absorbed. The result was growing undersupply, manifested by a sharply rising price of crude oil starting around 2002. The increasing tightness of the market indicated the beginning of the “production focus” regime. To meet the growth in demand, oil companies focused in this regime on production expansion. Access to developed crude oil reserves—that could produce in short term—became the most important source of competitive advantage.

Unlike the 1970s, when the crude oil reserves of the North Sea, Alaska, and Mexico could be developed, there were no reserves ready to be developed. In the 2000s, the proven oil reserves were increasingly concentrated in the Middle East, which did not allow access to IOCs. Moreover, in the Russian Federation the oil industry was increasingly brought under government control. The high oil price led to financial surpluses for the oil companies. But, rather than investing in exploring new reserves, IOCs returned substantial amounts of capital to shareholders via dividends and share buy backs (Stevens, 2009).

At the same time, IOCs faced increasing competition for reserves from state-controlled NOCs from importing countries, such as China and India. They became increasingly active in the international arena to secure supply for their home countries, and bypassed the market through bilateral transactions between countries (e.g., Chinese NOCs in Africa). Being denied access to “easy” oil, IOCs were only invited by crude reserves holding countries to execute more difficult exploration projects (e.g., deepwater and unconventional oil) requiring advanced technology (Jesse and Van der Linde, 2008), for which NOCs lacked the knowledge and the capabilities.

In all, the past five decades have seen periods of low and intense competition, as well as expansion and consolidation of the industry, vertical integration and divestitures, and varying degrees of political influence. The question how these dynamics affected the ability of firms to acquire and sustain superior performance positions will be the topic of the next sections.

3. Methods and data

To identify and assess the performance patterns over time, we combine various methodologies, i.e., autoregression modeling and strata-based analyses of persistent performance. In this section, we describe our empirical techniques, performance metric, and data source.

3.1. Persistence of performance: autoregression modeling

All models are estimated by maximum likelihood, rather than OLS, to preserve the initial observations of each firm’s return on assets (ROA)-series and to obtain efficient estimates of λ_j and π_j. The model is estimated on the pooled set of all firms and years in the sample period 1954–2008, as well as on rolling 5-year windows. Firms with less than 5 years of ROA-information are excluded from the analysis in order to have sufficient information to estimate firm λ_j’s. Specification tests include a Dickey-Fuller (1979) single mean test of stationarity (λ_j = 1), Likelihood-Ratio (LR₀) tests of the appropriateness of the autoregressive assumption (λ_j = 0, ε_tj = η_tj, the so-called null model), LR-tests of other restrictions on the persistence rate (needed, for instance, when evaluating the assumption that all firms in the industry share a common rate of convergence), and multiple F-tests to evaluate the contribution of fixed firm effects. Denominator degrees of freedom of the F- and t-statistics are based on a correction proposed by Satterthwaite (1946).⁹

3.2 Persistence of superior performance: non-parametric stratification

The dynamics of superior economic performance are analyzed using the stratification method developed by Ruefli and Wiggins (2000). Their method analyzes performance rankings of firms within industries over five subsequent years to arrive at a stratification of firms into superior, modal, and sub modal classes in the 6th year.¹⁰ Firms that maintain a position in the superior stratum for at least six consecutive windows are declared persistently superior performing. Using this stratification method, we analyze the patterns of persistent and transient superiority over time. In addition to the fraction of persistently superior performing firms, we study the duration of superior performance periods for all firms with superior positions for at least 1 year, using non-parametric life tables and parametric failure time (hazard) models (Neumann, 1999). If competition becomes more intense, then the estimated expected duration of persistent superior performance is expected to decrease. Accelerated failure time models are applied to obtain parametric estimates of the distribution of superior performance duration assuming duration log-normally distributed. The log-normal distribution gave a better fit than the often-used Weibull, and behaved slightly better than the more general gamma distribution. Differences with respect to the log-logistic distribution were small, and the consequences for the results were negligible. All models are estimated by maximum likelihood. Censored observations due to unfinished periods of superior performance at the closure of the observation period have been taken into account.

The discussion of performance stratification tends to focus on the persistent behavior of the superior stratum. However, firms migrate from one stratum to another marking performance patterns over time, which go unobserved by the isolated analysis of superior performance periods. This migration process can be conveniently monitored when interpreting annual performance stratifications as realized state distributions of a Markov-chain process. The literature contains but few analyses of this type, despite their early support by Nelson and Winter (1982) and Fiegenbaum, Thomas, and Tang (2001). The stability of performance strata is analyzed with mobility indices based on the properties of strata transition matrices. Let x_t = (x_1t, … , x_s,t)’ summarize the distribution of the number of firms over s performance strata in any year t, then the next period’s performance stratification x_t+1 is obtained as x'_t+1 = x'_tP. The s × s transition matrix P summarizes the conditional probabilities p_ij of moving from stratum i to stratum j; Σ_j p_ij = 1 for all originating strata i. Future behavior of the process is completely determined by this one-period transition matrix. Assuming convergence, the long-run equilibrium stratification follows as π' = lim _n→∞x'_t+n = x'_t lim _n→∞Pⁿ= x'_t Π, with Π= ι π', a transition matrix with rows equal to the equilibrium strata distribution.¹¹ In equilibrium, π' = π'P, implying perfect mobility. The transition matrix P has been estimated using the observed transitions for the entire observation period and for 5-year rolling windows. Moreover, it has been defined for performance stratifications based on three states (superior, modal, and sub-modal). In the robustness analysis, we include a fourth state to cope with entering and exiting firms.

Various mobility indices can be derived from the estimated transition matrix and the associated equilibrium stratification. We define four indices. First, the sojourn time in the superior stratum S₁, which is defined as 1/(1–p₁₁). The sojourn time S₁ measures the expected length of stay in the superior performance stratum (Prais, 1955). The larger S₁, the higher the expected persistence of superior performance positions. If competition intensifies, the migration of firms between strata increases and sojourn times become shorter, particularly for the superior performance stratum.

Second, the estimated equilibrium probability to migrate to the superior stratum in equilibrium, π₁. The higher this probability, the more competitive the superior stratum. Low equilibrium probabilities π₁ are expected to go along with longer sojourn times for incumbent firms in the superior stratum.

Third, the reciprocal of the harmonic mean of sojourn times in all strata, M_P = (s–Σ_ip_ii)/(s–1). We use M_P to indicate overall mobility (Shorrocks, 1978): the larger the probability that firms stay in their performance stratum, the closer M_P will be to zero. Low M_P values indicate stable performance stratifications, whereas high M_P values reflect intensive transition patterns, as typically found in hypercompetitive contexts.

Fourth, the second-eigenvalue index M₂ = 1 – |λ₂|, where λ₂ is the second eigenvalue of P, indicates the speed of convergence of P to its equilibrium state. We use M₂ to indicate the changeability of firms across performance strata. If M₂ is low, then the transition behavior converges slowly toward an equilibrium state, and persistent stratifications are more likely to occur. Alternatively, higher index values are indicative of more dynamic performance stratifications, in which superior position are less likely to persist.

The properties of all indices have been explored by Geweke et al., (1986), Prais (1955), and Shorrocks (1978).¹²

The transition matrix summarizing the migration of firms from one stratification to another has been estimated for the entire observation period as well as for 5-year rolling windows. Moreover, the transition matrix will be defined for performance stratifications based on three states (superior, modal, and sub-modal) as well as on four states, including the entering and exiting firms. The latter transition matrix typically has zeros for the fourth diagonal element.

3.3 Data sources

Our study focuses on the economic performance of the largest firms in the global oil industry, i.e., the “core of the industry” (Dosi et al., 1995). Size is relevant in the oil industry as it is a capital intensive sector where the economies of scale are large. As a result, the main competitors in this industry are large and globally-operating firms. We use Fortune Directories to obtain performance data, because most large oil firms are present in these listings. The use of Fortune data is in accordance with other studies that have focused on the largest and most visible firms (e.g., Lazonick, 1992; Fortanier and Van Tulder, 2009). We select the Global Fortune 500 as opposed to the US 500 to better explore the global characteristics of the oil industry.¹³ Our analysis is concerned with the oil industry, corresponding to the corresponding Fortune Directories “crude-oil production” and “petroleum refining”. Usage of the Fortune data enables us to analyze a global sample for a period of 55 years.^14,15

The data set contains 131 firms. The length of stay of firms in the sample varies from 1 to 55 years, while on average firms have 16.98 yearly observations. The total number of firm-year observations is 2225.

3.4 Performance metric

In line with previous performance research (Mueller, 1986; Cubbin and Geroski, 1987; Waring, 1996; McGahan and Porter, 2002, 2003; Wiggins and Ruefli, 2002), we measure corporate economic performance by ROA, i.e. annual accounting profits after taxes divided by total assets. The use of ROA enhances the comparability of our analysis with previous studies, but also facilitates the compilation of a global sample over a relatively long time frame. Clearly, ROA, being based on accounting profitability, has been criticized as a measure of firm performance (Harcourt, 1965; Fisher and McGowan, 1983; Hawawini et al., 2003). But at the same time, many studies report a remarkable consistency in outcomes regardless the use of ROA or market-based indicators, such as Tobin’s Q (Jacobsen, 1987; McCahan, 1999). Besides, stock-market based performance measures, like Tobin’s Q, are also not without dispute, as they may confound the actual performance of the firm with investor expectations (Thomas and D’Aveni, 2006). More specifically, McGahan and Porter note that “accounting biases are likely to influence levels of effects to a greater extent, however, than the persistence in effects”, which further supports the use of ROA in our analysis of long-term performance dynamics (McGahan and Porter, 1999: 145).

In our analysis, we compare firm performance over time as well as over countries. It is noted that we implicitly assume that the reported profits and assets are based on similar accounting standards and auditing practices. In the earlier years of our analysis, United States firms dominate the sample and these firms report under the same regime. In the most recent years, reporting standards have converged and improved, for example through the efforts of the IASB (Barth et al. 2008). We acknowledge that in the intermediate years of our sample our results may be influenced by differences in reporting standards.

3.5 Robustness analysis

A key issue in longitudinal panel analysis is survivor bias (e.g., Gschwandtner, 2005). The Fortune Global 500 also suffers from this bias. To investigate the consequences of this survivor bias for the conclusions of our study, we conducted an analysis of all exits from the Fortune Global 500 during the research period. Subsequently, we performed a robustness analysis for a sample of exitors.

We have investigated the reasons for firms to drop out of our sample. In the Fortune sample, the inclusion criterion is relative size, measured in revenues. We distinguish between temporary exits—that is departure from the Fortune directories and subsequent return in later years—and permanent exits. In total, 31 firms leave the Fortune sample to return at a later instance. These firms leave our sample because they fail to meet the revenue criterion of the Fortune Global 500 Directories. In the sample of permanent exits, we find 74 firms that leave the sample, including 11 firms that had a prior temporary exit. Permanent exits occur because of an acquisition (23), bankruptcy (1), new industry classification following diversification (5), failing to meet the size criterion (38), or for reasons that could not be retrieved (7). Clearly, exits—temporary and permanent—based on the size criterion are relevant for our study. In particular, the height of the oil price directly influences company revenues and thus the probability of a company’s presence in the Fortune directories. Increasing oil prices cause oil companies to displace companies from other (Fortune) industries, while declining prices have the opposite effect.¹⁶

We analyze the robustness of our results for the size criterion as follows. For all temporary exits of firms we collect ROA information from the Compustat North America (for US and Canadian firms) and Compustat Global (for the other countries) databases, which are also used by Fortune. In total, we add 122 observations to our data (this is 58.3% of all observations concerning temporary exits). In an unreported analysis, we show that our results remain largely unaltered.¹⁷ This robustness analysis indicates that the size-based inclusion criterion does not influence our conclusions.

4. Results: identifying performance dynamics

In this section, we describe the results of the statistical analysis of the performance patterns, preceded by the descriptive statistics of our sample.

4.1 Descriptive statistics

Table 2 shows for 3-year intervals¹⁸ the descriptive statistics, including the return on assets of the firms in our sample, and the price and the production volume of crude oil.

For example, in the 1954–1956 interval, the sample had on average 26 firms per year in these 3 years, with an average ROA of 8.42 and a median of 8.09. By presenting the median, minimum, maximum, 25th and 75th percentiles, we provide a detailed description of the distribution of the data in each of the intervals. Table 2 also reports the fraction of firms with a negative ROA, i.e., loss-making firms. The table shows that the average return fluctuates over time. From the start of the research period until 1977, average return decreased. After a recovery in the late 1970s, average performance plummeted and remained low until the mid 1990s, after which the average performance increased. In the last interval, the average return even surpassed the high level of the first interval of the sample period (Figure 2).

4.2 Persistence of performance: autoregression results

The results of the autoregression analysis are presented in Table 3 and Figure 2. We estimate the model over the 5-year rolling windows. In Table 3, we document each third window as of 1954, while Figure 2 shows all estimations.¹⁹

The analysis demonstrates, for example, that the estimated equilibrium performance level π in the 1954–1958 window is 6.94, which is significantly different from zero at the 1% level; the convergence rate λ is 0.85 and also significant at the 1% level. These results demonstrate that in the 1950s, the equilibrium return on assets was almost 7%. Convergence to equilibrium levels was relatively slow, as the λ is close to one.

Over the entire period, the estimations for the equilibrium performance (π) show that the industry has experienced a period of high performance in 1954–1970 (between 5.32% and 6.94%). Then a steadily declining trend is visible until the 1990–1994 window, after which performance increases again. Toward the end of our sample period, equilibrium performance again reaches high levels, around 8%.

In relation to the research question concerning performance convergence, stated in the introduction, we find in the 1954–1970 periods low convergence levels (i.e., a relatively high λ of at least 0.76). Then, we observe an increase in convergence levels, as in the 1980s and 1990s the λ never reaches the levels of the 1950s and 1960s again. By the end of our sample period, convergence again decreases, as λ rises to levels between 0.65 and 0.79.

The Akaike information criterion (AIC) is an indication of the overall fit of the model, where lower values indicate a better fit. The criterion first increases over time and decreases in later periods, which reflects that the autoregressions model fits the data best in the 1950s and 1960s.

4.3 Persistence of superior performance results: non-parametric stratification

The results of the stratum analysis are presented in Tables 4–7 and Figures 3 and 4. Table 4 presents the fractions of firms in the three strata. ²⁰

In the 1954–1958 window, 4% of the sample firms are in the submodal stratum, 68% is in the modal stratum and 28% is in the superior performing stratum. A further investigation of the superior stratum shows that 12%—less than half of the superior sample—of all firms are persistently²¹ superior performers. The other 16% of the firms are temporary superior performers.²² In Table 4, we also present the average ROA in each of the subsamples. For example, in the 1954–-1958 window the average ROA in the modal sample is 7.05%. These averages emphasize the stratification approach, which is based on relative performance; in the 1990–1994 window the average performance in the superior stratum is 5.73%, which is below the average in the model sample in the first five windows presented.

A clear time trend in the fraction of superiorly performing firms emerges. We distinguish between persistent superior performance—at least six concatenated windows of superior performance—and transitory superior performance, that is, up to five concatenated superior windows. Figure 2 demonstrates that the fraction of persistently superior performing firms shows much less fluctuations than the fraction of transitory superior performers. This difference may (partly) be explained by the difference in duration of superior performance of the two fractions. The fraction of persistent superior performers displays a downward trend until window 1979–1983 when the fraction benefits from a strong recovery until a peak in the 1986–1990 window. After this peak, the fraction faces a downward trend until it disappeared in the window 1993–1997.

The fraction of transitory superior performing firms exhibits a downward trend until the window 1977–1981, which is slightly before the fraction of persistent superior performers interrupted its downward slide. The fraction of transitory superior performance shows a strong recovery to a peak in the 1983–1987 window, after which a new downward trend emerges. After bottoming out in the window 1994–1998, the fraction displays wide fluctuations around an upward trend.

Table 5 displays all firms that achieved persistent superior performance. Five out of the nine persistently superior performing firms belong to the Seven Sisters. This table also shows that the duration of persistent superior performance has steadily declined from Texaco‘s record length of 21 years to the minimum periods of 10 years (Royal Dutch Shell, Repsol, and Chinese Petroleum Corporation). Table 6 presents the analyses of the migration of firms across performance strata over the research period.

In ‘Transitions excluding sample exit’ column of Table 6, the results are shown when exits from the sample are not taken into account. When a firm is in the highest stratum, there is a 76.7% probability of staying in this group, a 23.3% probability of migration to the middle stratum and a zero probability of migration to the lowest performance group. Although in all three groups, the probability to stay in that group is the highest, we find that most migrations are possible. When we include in ‘Transitions including sample exit’ column, sample exits our results do not change materially. The estimated probabilities to migrate to a stratum (π_i) are 0.113 for the superior stratum, which implies that there is an 11.3% chance to enter this group. For the middle and low strata the probabilities are larger, 68.4% and 20.3%, respectively. The sojourn time (S_i), the expected length of stay in a stratum, is 4.3 years for the superior stratum, whereas in the two other strata we find sojourn times of 13.5 years and 8.3 years. Finally, the overall mobility (M_p) and the convergence speed (M₂) are 0.214 and 0.145. An interpretation of these estimates becomes meaningful in comparisons over time. In Table 7 and Figure 4, we present estimates using a rolling window approach. In Table 7, we present every third window, while Figure 4 contains estimates of all windows.²³

The estimated probabilities to migrate to a stratum (π_i) are highest in the first four windows, at least 11.8%. The sojourn times (S_i) are also relatively high, at least 4.8 years. In other words, in the period 1954–1967 the probability of entering the superior stratum was relatively high and the length of stay was relatively long. During the 1970s the π_i’s are lower, from 9.4% in the 1966–1970 window to 5.9% in 1975–1979. Also, the sojourn times drop from 3.4 years to only 2.6 years. In the 1970s, achieving and maintaining superior performance is clearly more difficult. Then the 1978–1982 period yields a high π_i of 0.421 and S_i of 13 years. The probability to migrate to the superior stratum goes from 18.7% to zero in the 1993–1997 window and recovers in the 1999–2003 window (to 0.9%). The sojourn times decrease from 6.8 years to 3 years (1990–1994) and cannot be estimated in the next two windows due to the absence of observations in this stratum. Whereas the previous statistics focus on the top segment, the overall mobility (M_p) and the convergence speed (M₂) capture the effect for all firms. The convergence speed appears relatively stable over the sample period.

5. Discussion: causal explanation of the performance dynamics

In this section we present our causal explanations of the observed patterns of performance as outlined in the previous section. The analyses of the main dynamics of the global oil industry (see Section 2.2, in particular Table 1) are the basis of our interpretation of the performance dynamics. In this discussion, it should be noted that the stratification analysis is based on rolling 5-year windows while the historical analysis is based on years. When interpreting windows we need to take into account that the performance of a single year will influence five subsequent windows.

5.1 “Concession-based regime” (1954–1972): The Seven Sisters rule

The initial slow convergence of performance during this period (Figure 2) is consistent with the concession system, the main source of sustainable competitive advantage in that phase. The foundation of OPEC in 1960 coincided with an increase of the pace of convergence around that time. As a countervailing power to the Seven Sisters, OPEC contributed to a leveling of the playing field for the so-called “Independents” and thereby induced faster convergence. Moreover, IOCs as well as the first NOCs entered the sample. Because most entrants were IOCs, like the incumbents, the inter-firm differences remain relatively small, compared to subsequent competitive regimes.

A scenario that explains the rise and subsequent peak in migration to the superior stratum around 1960 (Figure 4) is that the Independents benefitted from the leveling of the playing field by OPEC. In this regime, the first NOC—Petróleo Brasileiro (Petrobras) from Brazil—entered the superior performance stratum.

The large fraction of persistently superior performing firms at the beginning of the research period (Figure 3) is entirely accounted for by the Seven Sisters (i.e., Chevron, Gulf Oil, and Texaco). Their performance may be explained by their concessions and their strategy of vertical and horizontal integration. Although all Seven Sisters achieved superior performance during this era, not all Sisters achieved the same duration of superior performance. Apparently, membership of the Seven Sisters cannot fully explain performance. The subsequent decrease of the persistent superior fraction may be explained in terms of both an absolute decrease of the occurrence of persistent superior performance and an increase of the sample size due to new entry. Increased taxation and nationalization (in the 1970s) may explain the ending of periods of persistent superior performance of the Sisters. In 1965, Petrobras initiated a period of persistent superior performance (Table 5). This persistent superior performance may be ascribed to the company’s monopoly on oil exploration, production, and downstream activities in its home country Brazil.

The initially high level of the fraction of transitory superior performing firms and its subsequent downward trend (Figure 3) could reflect the increasing taxation and nationalization. The fraction of transitory superior firms is equally split between Independents, such as Amerada and Marathon, and the remaining, that is not persistently superior performing, Sisters, among which Exxon and Royal Dutch Shell. No NOCs entered the superior stratum on a transitory basis.

5.2 “Reserves access regime” (1973–1985): the rise of the “Independents” and the NOCs

The continued trend of the increasing pace of convergence from the second half of the 1970s till the end of the regime may be explained by various scenarios. During this period, nationalizations and tax raises by an increasing number of OPEC members continued. Furthermore, the already high entry rate increased further, expanding the number of competitors in the sample. Competition for access to crude oil reserves outside OPEC-domain—the main source of sustainable competitive advantage for the IOCs in the new regime—increased. In contrast to the prior regime, NOCs entered on large scale, contributing to increasing inter-firm differences in the sample. In addition to entry by NOCs from crude oil exporting countries, the second regime witnessed the entry of NOCs from importing countries, which further contributed to inter-firm differences in the sample.

The merger and acquisition wave among IOCs during 1979–1984 may be due to a strategic response to the increasing intensity of competition in the industry and the related performance convergence. One of the Seven Sisters, that had been persistently superior performing in the past was acquired, i.e. Chevron acquired Gulf. During the wave, the downward trend of the sample’s average performance reversed for a transitory recovery. The upward trend of the pace of convergence continued, with wide fluctuations, during the wave. Apparently, this consolidation could not stem the tide of increasing convergence.

A new peak in the migration to the superior stratum in the second half of the 1970s, after a period of low migration rates, may primarily be explained by Independents and NOCs. Independents benefitted from OPEC’s policy of nationalization of the Seven Sisters’ concessions, and OPEC’s bypassing the Sisters. The superior performance of the NOCs may be accounted for by their domestic monopolies.

The fraction of persistently superior performing firms continued its downward slide until the early 1980s. The subsequent recovery of this fraction may be explained by NOCs and two Sisters, Exxon and Royal Dutch Shell. All Sisters suffered from the loss of their concessions—their prior source of sustainable competitive advantage—and from failed diversification strategies. A possible scenario for explaining the persistent superior performance of these two largest Sisters based on their scale (Dobrev and Carroll, 2003) and technology, which allowed them to produce from offshore fields outside OPEC domain (North Sea, and the Gulf of Mexico). In the case of Exxon, the firm’s early adaptation of shareholder value management and subsequent strategy change may contribute to the explanation. For Royal Dutch Shell, its pioneering with scenario planning may have helped the company to formulate effective strategies for anticipating changes in the oil price. NOC Petrobras’ persistent superior performance ended in 1979. At that time the company’s capital expenditures in exploration and production took off in an effort to make Brazil less dependent on oil import. This example of Petrobras illustrates how political objectives may influence the strategy and performance of NOCs. In 1976, another NOC from an exporting country, Petróleus de Venezuela (PDVSA) of Venezuela, began its period of persistent superior performance.

In 1982, the first NOC of a crude oil importing country achieved persistent superior performance: Chinese Petroleum Corporation of Taiwan (in 2007 renamed CPC Corporation), which operated a downstream monopoly in its home country.

The widely fluctuating but rising fraction of transitory superior performing firms in the second half of the 1970s and the first half of the 1980s seem predominantly caused by Independents, and to some extent, NOCs. Besides Exxon and Royal Dutch Shell, none of the Sisters achieved superior performance, not even on a transitory basis.

5.3 “Efficiency focus regime” (1986–2001): the fight for survival

The upward trend of the convergence speed was reversed in the third regime. The slow down of performance convergence after a peak in the early 1990s, may be due to the low rate of entry combined with a high exit rate. The industry shakeout further contributed to the slowing down of the convergence among the “survivors”. The convergence lessened during the new merger and acquisition wave between 1998 and 2001. Both during and after this wave, the pace of convergence continued a downward trend whereas average performance of the sample prolonged an upward trend. Again, one of the Seven Sisters, that had been persistently superior performing in the past was acquired, i.e., Mobil was taken over by Exxon. Migration to the superior stratum was low and declined further in this regime and bottomed out in the 1990s as control over crude oil reserves was no longer enough for obtaining a sustainable competitive advantage. Efficiency became the main source of sustainable competitive advantage in this regime. Moreover, several NOCs lost their national monopolies as their home country governments started to liberalize their markets and (semi) privatized their NOCs.

The continued rise of the fraction of persistent superior performing firms in the first stage of this regime may have to do with the declining sample size. After the peak in the second half of the 1980s the fraction sharply declines until it becomes zero in the mid-1990s. For each case we briefly outline possible scenarios for explaining the ending of their persistent superior performance. The two largest Sisters faced in the early 1990s the end of their periods of persistent superior performance but continued transitory superior performance until the mid 1990s. In the case of Royal Dutch Shell, the firm’s strategy of expanding into less remunerative specialty chemicals contributed to its loss of superior performance. Exxon’s exit from the superior stratum may be due to this company’s struggles with flat reserves and declining production. Venezuela’s NOC, PDVSA, continued its persistent superior performance until 1990. In 1988, this company integrated forward, and expanded abroad by building up a large refining and marketing system in the United States and in Western Europe. The Chinese Petroleum Corporation continued its persistent superior performance until 1991. In 1989, this state-owned company faced the liberalization of Taiwan’s domestic oil industry, and subsequently the company diversified. Repsol lost its monopoly at home in 1992, after the liberalization of the Spanish oil industry. The company was privatized in 1996, and expanded in downstream activities in Latin America. In the same year, Repsol’s persistency of superior performance ended. Loss of domestic monopolies and changes in firm strategies—diversification, vertical integration, and internationalization—could explain the loss of persistent superior performance by NOCs.

The rising trend of the fraction of transitory superior performers is reversed to a declining trend, possibly as a result of increased competition.

5.4 “Production focus regime” (2002-end of research): Exxon again

The reversal of the downward trend of the pace of convergence, after reaching the lowest level of the research period in the window 2003–2007, may be accounted for by the increasing rate of entry while the exit rate becomes low. The increasing number of competitors adds to the competitive intensity of the industry.

Migration to the superior stratum is equally split by NOCs and IOCs. All entry is on a transitory basis. The sharp rise of demand and the crude oil price since 2002 (continued price increase until mid-2008) benefited in particular firms with strong proved and developed, but non-producing, crude oil reserves. These companies could increase production without massive investments.

The fraction of persistent superior performing firms remained zero. As a result, the fourth regime is the first regime without persistent superior performance. A scenario for explaining the absence of persistent superior performance is that this regime—only five windows—is relatively short. Because of the relative newness of this regime at the time of this study, it is too early to determine whether the superior performance be sustain sufficiently long to qualify as persistent.

The fraction of transitory superior performing firms continued its rise. Three NOCs and three IOCs, among which two Sisters, achieved transitory superior performance. The three NOCs are all from Asia: Oil and Natural Gas Commission of India, Petronas of Malaysia, and PTT of Thailand. A relative newcomer in the superior stratum is Lukoil, the first IOC from the Russian Federation. Lukoil is not a NOC. In 1993, it was formed as the first joint stock oil company of Russia. In 1999, it joined the Fortune Global 500, and directly entered the superior stratum. Lukoil achieved the longest period of superior performance in the fourth regime. Its superior performance may be attributed to its crude oil reserves.

Finally, we find two Sisters to achieve transitory superior performance: Chevron and ExxonMobil. After absence during the second and the third regime, Chevron returned in 2004 to the superior stratum. Chevron acquired two other Sisters: Gulf Oil in 1984 during the first merger and acquisition wave and Texaco in 2001 during the second wave. Chevron’s size is likely to have contributed—through economies of scale—to its superior performance. Remarkable is the achievement of Exxon. This is the only company within the sample that managed to achieve superior performance during all four regimes, including persistent superior performance in the second and third regimes. In total, Exxon showed superior performance in 23 windows, which is the highest number in the sample. We already discussed several scenarios for explaining Exxon’s performance in previous regimes. The acquisition of Mobil in 1998 adds to the explanation of ExxonMobil’s superior performance during the fourth regime. As a result of this acquisition, the company had, at the start of the new regime, the reserves (e.g., the gas fields of Quatar) ready for production. Moreover, Exxon’s cost control and the cost synergies of the Mobil-acquisition possibly contribute to its superior performance.

6. Conclusions

In this study, we have used both statistical and historical analyses to measure and explain the dynamics of superior performance in the global oil industry during 1954–2008. Our study contributes to prior research on performance in three respects. First, by combining parametric and non-parametric research methods to the same set of data we enhance our understanding of the method effects on performance research. Second, with a long time frame of 55 years our study of the oil industry captures the long run institutional changes that are inherent to this geo-political industry. Third, by analyzing a global sample of oil firms we are able to include all relevant players in this industry regardless of their home country.

6.1 Combining methods

We first use autoregression modeling to measure the persistence of performance and find cyclical patterns of both the equilibrium performance rate and the rate of performance convergence. We then measure the persistence of superior performance using non-parametric stratification, which also shows cyclical patterns of the fractions of persistent and transitory superior performing firms and of the probability of firms to migrate to the superior performance stratum. Although both methods generate cyclical patterns, these patterns are not identical, which may be explained by the difference in aspects measured. By combining the two methods we were able to provide a more complete measurement of the performance dynamics.

6.2 Long-run dynamics

The global oil industry is unparalleled in its geopolitical impact. In particular, in the global oil industry, institutions influence performance dynamics, next to firm and industry factors (North, 1990; Peng et al., 2009). For a strategic natural resource like crude oil, control over (crude oil) reserves plays a key role in explaining performance dynamics. Institutions, in particular national governments, play a large role in the control over, and access to, crude oil reserves through mechanisms such as their national policies, national companies, and supra-national cartels such as OPEC. Three out of four regimes point to control over and access to reserves as the main source of sustainable competitive advantage. The influence of institutions on the global oil industry seems related to the scarcity of oil reserves. Institution-based competitive advantage (Peng et al., 2009) plays an important role in the global oil industry, as illustrated by the concessions of the Seven Sisters and the national monopolies of NOCs. The increasing scarcity of oil in the fourth regime suggests an increasing influence of institutions on performance dynamics.

Institutional change is typically slower than corporate and industrial change. By adopting a longer timeframe, our study captures all relevant changes, not only corporate and industrial change but also institutional change. Our analysis covered the main institution-based changes as well as firm-level and industry-level changes.

6.3 Global scope

For decades and continuing until the first years of our research period, the concession-based regime prevailed, in which the oil industry consisted of a relatively small and homogeneous group of Western-based international oil firms. However, during the research period, the oil industry has become increasingly international and diverse. Institutional change led to the entry of national oil firms in the industry and into the superior performance stratum. These national oil firms came from different geographies and operated under different governance regimes than the traditional international oil firms. First, the industry witnessed the entry of national oil firms from exporting countries. Subsequently, national oil firms from importing countries entered the industry as well. These two groups of national oil firms differed in terms of goals and governance from each other and from the privately owned international oil firms.

By adopting a global scope, our research could take these new oil firms into account and capture the increasing variety in corporate goals and governance within the global oil industry (Hall and Soskice, 2001). The global oil industry shows the emergence of competition between international oil firms and national oil firms. National oil firms’ competitive advantage is typically institution-based. National monopolies or other privileges contribute to their performance. National oil firms’ performance suffers from the obligation to pursue political goals, and from inefficiencies, at least compared to the international oil firms. Some national oil firms have been (persistent) superior performers. At the end of the research period, half of the superior stratum consisted of national oil firms from importing countries. The other half consists of international oil firms. We acknowledge that the performance definition used in our research is aligned to the goals of the privately owned oil firms. To completely assess the performance of national oil firms we need a stakeholder view instead of a shareholder perspective.

6.4 Stylized facts

Our study intends to contribute to the understanding of the dynamics of superior performance among the world’s largest oil companies by adding to the growing body of stylized facts about performance patterns (Malerba et al., 1999). In this concluding section, we emphasize the six most important stylized facts. First, over the whole research period (1954–2008), we find the convergence of performance to have increased. Second, we observe that the length of periods of superior performance of firms follows a declining trend. Third, we find the fraction of persistent superior performing firms to disappear in the third regime. Fourth, even in the fourth regime we find transitory superior performing firms. The fourth regime at the end of the research period is still too recent to conclude that persistent superior performance is not possible in this regime. Fifth, the composition of both the industry and the superior performance stratum has changed through the research period. Both have become more heterogeneous in terms of firm characteristics. Finally, the development of the different aspects of performance patterns (convergence rate, performance strata, and strata migration) is cyclical. These patterns vary through time. Appreciative theorizing enables us to causally explain the patterns in terms of an interplay of firm, industry, and institutional factors.

Acknowledgements

The authors gratefully acknowledge the use of IKS software developed by Tim Ruefli. They also benefitted from comments by participants at the 2008 European Business History Association Meeting (Bergen), Pursey Heugens, Keetie Sluyterman, two anonymous referees and oil industry experts, in particular Jan-Hein Jesse (International Energy Agency). They would like to thank Paul Plaatsman for his data search. They are grateful to Sandra Prenger, Dennis Wageveld, and especially Rob van Dale, for their research of the global oil industry.

References