«Industrial policy and the creation of new industries: evidence from Brazil’s bioethanol industry Santiago Mingo*,y and Tarun Khanna** Downloaded ...»
In 1986, the price of oil plummeted. This price reduction—in addition to a shortage of ethanol fuel—brought into question the use of ethanol as a substitute for gasoline and led the Brazilian government to start terminating the Pro-alcohol program. The industrial policy program was coming to an end. During the period following the end of Pro-alcohol, the government stopped offering soft loans for the construction of new bioethanol plants and support for the bioethanol program from state trading companies was eliminated (Weidenmier et al., 2008). During the early Downloaded from http://icc.oxfordjournals.org/ at Harvard Library on July 10, 2014 ´ ¸´ 1990s, the Instituto do Acucar e do Alcool —the main government agency in charge of regulating the sugarcane sector—was dismantled. The sugarcane sector underwent a period of significant restructuring.
During the 1990s, there was an active period of acquisitions in the industry (Mingo, 2013b). The sugarcane company Cosan8 exemplifies what was occurring in the Brazilian sugarcane sector after Pro-alcohol ended. The company, with a sugarcane tradition dating back to the 1930s, started an aggressive acquisition program in 1986—until that year the company owned only one plant. Taking advantage of a fragmented and troubled industry, Cosan acquired six new plants during the period 1986–2000. The growth and success of the company has allowed it to become a global giant in ethanol and sugar production. During the aftermath of Pro-alcohol, the business environment in Brazil facilitated the occurrence of acquisitions in many sectors of the economy (KPMG, 2001). This “acquisition boom” was a consequence of the opening and liberalization of the Brazilian economy in the early 1990s. Figure 1 provides a timeline of the evolution of the Pro-alcohol program in Brazil.
Even though our focus is on Pro-alcohol and the creation of the bioethanol industry in Brazil, it is important to note that different forms of government intervention have continued to exist.9 For instance, according to federal law, gasoline sold at the pump must be blended with anhydrous ethanol. During the past 10 years, the percentage of ethanol content in a liter of gasoline has fluctuated between 20% and 25%. The government adjusts this percentage depending on fuel prices and the supply of ethanol available in the market. Additionally, the government has incentivized the adoption of flex-fuel technology in motor vehicles since its introduction in ´ The sugar and ethanol business of Cosan is now part of Raızen, a separate company that was ´ created in 2010. Raızen—the most important ethanol producer in Brazil—is a joint venture between Shell and Cosan that focuses on ethanol and sugar production, cogeneration of electricity, and fuel distribution.
We want to thank two anonymous reviewers for suggesting the discussion of Post-alcohol government intervention.
Industrial policy and new industries 11 of 32
Figure 1 The Pro-alcohol program in Brazil. The stages of Pro-alcohol are based on historical accounts written by Brazilian scholars that are experts in the industry (Shikida and Bacha, 1999; Walter and Cortez, 1999; Moraes, 2000; Mathews, 2006).
4. Data and methods To explore our theoretical arguments empirically, we put together a novel data set using a sample of bioethanol and sugar plants in the Brazilian Center-South Zone11 (Figure 2). First, we collected information about the yearly production and operational performance of these plants during the period 1999–2005. Therefore, our operational performance data covers a period that occurred approximately 15 years after the end of Pro-alcohol. We define three different subgroups: plants founded before the Pro-alcohol period; plants founded during the Pro-alcohol period; and plants founded after the Pro-alcohol period. We define the Pro-alcohol period as the period between the years 1975 and 1985, which is when the major government policies and incentives intended to create and develop the bioethanol industry were in place (Shikida and Bacha, 1999; Walter and Cortez, 1999; Moraes, 2000;
Mathews, 2006). All the data used to estimate the operational performance of the plants come from UNICA.
Second, we collected detailed historical information about the plants and the entrepreneurs that owned them. Our focus was on getting data about the origins and ownership history of each plant, including the current owners. In this case, by ´ ˆ
¸˜ ¸˜Contribuicao de Intervencao do Domınio Economico.
More than 85% of the sugarcane produced in Brazil comes from the Center-South Zone. The ´ ´ states in the Center-South Zone are: Espırito Santo (ES), Goias (GO), Minas Gerais (MG), Mato ´ Grosso (MT), Mato Grosso do Sul (MS), Parana (PR), Rio de Janeiro (RJ), Rio Grande do Sul (RS), ˜ Santa Catarina (SC), and Sao Paulo (SP).
S. Mingo and T. Khanna 12 of 32 Downloaded from http://icc.oxfordjournals.org/ at Harvard Library on July 10, 2014 Figure 2 Map of Brazil’s Center-South Zone.
current owners we mean owners during the period 1999–2005—approximately 15 years after the end of Pro-alcohol. We identified the year when every plant founder entered the industry and whether the owner of a plant is an independent entrepreneur, family-owned business, or another type of organization—for example, a cooperative. Several sources were used to reconstruct the ownership history of these plants: industry associations, data coming directly from the firms, master and doctoral dissertations published in Brazil, historical accounts of the towns where some of the plants were located, and local and international news databases. As an example, Appendix A shows the case of Generalco, a plant founded during Pro-alcohol.
The data gathering process allowed us to build a sample of 193 plants. These plants represent more than 80% of all the plants located in the Center-South Zone at the end of 2006.12 Last but not least, we complement these data through fieldwork that included interviews with experts in the history of the sugarcane industry and
Pro-alcohol, and executives of some of the most important sugar–ethanol companies in Brazil.
4.1 Dependent variable The main dependent variable (Operational performance) is the average of the yearly operational performance of a plant for the period 1999–2005. The specific measure we use as a proxy for plant performance is the total amount of kilograms of sucrose produced divided by the total number of tons of sugarcane crushed during a harvest season. The dependent variable behaves reasonably well, with a histogram that is close to a normal distribution. Figure 3 shows a simplified diagram of the production process in an integrated sugar–ethanol plant—note that each “plant” is comprised of Downloaded from http://icc.oxfordjournals.org/ at Harvard Library on July 10, 2014 the plantation and the processing facility. A higher capability to produce sucrose per ton of sugarcane is not only the result of a more efficient transformation of sugarcane into final products, but also the result of a higher level of sucrose content in the sugarcane farmed. Small variations in operational performance can have a considerable impact on the quantity of product that can be produced from a ton of sugarcane. Since the production of bioethanol and sugar is a low-margin commodity business, even small changes in operational performance can have a significant impact on the income statement of these companies.
4.2 Independent variables The first independent variable (Industrial policy plant) is a dummy that is equal to one if the plant was founded during the industrial policy period and zero otherwise.
The second independent variable (Industrial policy entrepreneur) is a dummy variable that is equal to one if a plant is currently owned by an entrepreneur who entered the industry during the Pro-alcohol period. As previously mentioned, we call these entrepreneurs industrial policy entrepreneurs. To define current ownership we consider the period after 1999. Note that plants currently owned by industrial policy entrepreneurs are not necessarily industrial policy plants—some of these entrepreneurs have acquired plants founded in other periods. Also, there are many industrial policy plants managed by entrepreneurs that did not enter the industry during the policy period (Table 1). Another independent variable (Acquired plant) is a dummy indicating if the plant was acquired at some point after the end of Pro-alcohol, that is, after 1985.
Additionally, to take a more detailed look at the Pro-alcohol period and take into account the fact that Pro-alcohol might have ended more gradually, we break the industrial policy period into different subperiods. Since the Pro-alcohol period lasted about a decade, it is important to analyze more closely the behavior of the models by using subperiods. These additional analyses try to isolate even more the effects of Pro-alcohol. In one of the estimations, we divide Pro-alcohol into two subperiods— we use the dummy variables Industrial policy plant founded in 1975–1979 and S. Mingo and T. Khanna 14 of 32
Figure 3 Production of sugar and bioethanol in an integrated plant. Both the plantation and the processing facility are considered part of what we call a plant.
Industrial policy plant founded in 1980–1985. In another case, we use three subperiods: 1975–1977, 1978–1981, and 1982–1985. Models with more than three subperiods did not yield stable results because there is not enough statistical power to run these analyses.
Every model also includes two interaction terms that are designed to explore how post-industrial policy acquisitions affect the performance of firms depending on their date of founding and the origins of the owner of the company. The first one is the interaction between the variables Industrial policy plant and Acquired plant. The second one is the interaction between the variables Industrial policy entrepreneur and Acquired plant—this last interaction term was included to tease out the impact of industrial policy entrepreneurs that made acquisitions during the aftermath of Pro-alcohol. Finally, we also include the indicator Post-industrial policy Industrial policy and new industries 15 of 32 plant. This dummy variable is equal to one if a plant was founded after the industrial policy period.
4.3 Control variables We control for different factors that could be leading to omitted variable bias or other spurious results. An important control is the average amount of sugarcane crushed (Sugarcane processed). Naturally, the amount of sugarcane that a plant processes is highly correlated with its production capacity and plantation size. We also include the quadratic term of this control variable to account for scale economies in the production process.
The average proportion of sugarcane used to produce sugar is another important Downloaded from http://icc.oxfordjournals.org/ at Harvard Library on July 10, 2014 determinant of the operational performance of a plant.13 We control for this by including the variables Proportion of sugar production and its quadratic term (Proportion of sugar production)2.
We also use a set of ownership indicators. The type of ownership of a plant can have an impact on the organizational structure and the quality of management, affecting operational performance. We include four ownership indicators: (i) Founded and owned by a multi-plant company indicator; (ii) Family-owned indicator;
(iii) Owned by a cooperative (of farmers) indicator; and (iv) Owned by foreign capital indicator. It is important to note that only (ii), (iii), and (iv) are mutually exclusive.
Finally, state indicators—to control for geography—are also included. State indicators are important because operational performance is highly dependent on ˜ geographic location. Some states, such as Sao Paulo, are known for their excellent conditions to grow sugarcane.
To analyze our cross-section of 193 plants, we use ordinary least squares regressions with heteroskedasticity-robust standard errors. Problems of multicollinearity were not observed. Descriptive statistics are reported in Table 2.
5. Results and discussion We start with a simple analysis of how the performance of a plant depends on the period when it was founded. The 193 plants in the sample were classified according to their year of founding: 85 plants were founded before the Pro-alcohol period (Prealcohol plants); 85 plants were founded during the Pro-alcohol period (Pro-alcohol plants); and 23 plants were founded after the Pro-alcohol period (Post-alcohol plants). Table 3 shows the average operational performance for each of these three categories. According to the t-tests for equality of means, the mean performance of Note that sugarcane can be transformed either into bioethanol or sugar. Therefore, if the proportion of sugarcane dedicated to sugar production is r, then the proportion of sugarcane dedicated to bioethanol production is (1 À r).
S. Mingo and T. Khanna 16 of 32 Table 2 Descriptive statisticsa
Pro-alcohol plants is significantly higher than that of Pre-alcohol plants (P50.10).
The difference between the mean performance of Pro-alcohol plants and Post-alcohol plants is not significantly different from zero (P ¼ 0.415).
The correlation coefficients are reported in Table 4. There is a weakly positive correlation between Operational performance and Industrial policy plant (0.081;
P ¼ 0.266). Also, there is a weakly negative correlation between Operational performance and Pre-industrial policy plant (À0.114; P ¼ 0.116). The correlation between Table 4 Correlationsa