«Industrial policy and the creation of new industries: evidence from Brazil’s bioethanol industry Santiago Mingo*,y and Tarun Khanna** Downloaded ...»
Operational performance and Acquired plant, and Operational performance and Industrial policy entrepreneur, are not significantly different from zero. The correlation between Industrial policy plant and Acquired plant is positive and significant (0.153; P ¼ 0.034), indicating that plants founded during the industrial policy period tend to be acquired at a higher rate than those that were not built during the policy program. As it would be expected, there is also a positive correlation between Industrial policy plant and Industrial policy entrepreneur (0.422; P50.01), indicating that existing entrepreneurs that entered the industry during Pro-alcohol tend to manage plants founded during that period.
Regarding ownership indicators, the dummy variable Founded and owned by a multi-plant company is positively correlated with performance (0.216; P50.01). On Downloaded from http://icc.oxfordjournals.org/ at Harvard Library on July 10, 2014 the other hand, the variable Owned by a cooperative is negatively correlated with performance (À0.145; P ¼ 0.044). Also, as expected, we observe a positive correlation between Operational performance and the amount of sugarcane processed (0.314;
P50.01). Therefore, we can infer that the operational performance of a plant is positively correlated with its size. This is consistent with the presence of scale economies. The correlation between Operational performance and the proportion of sugarcane used to produce sugar is significant and negative (À0.206; P50.01), finding that is also consistent with our previous discussion.
The results for the regressions analyzing the impact of industrial policy on operational performance are reported in Tables 5, 6, and 7. In Table 5, Model (1), the coefficient on Industrial policy plant is positive and significant, that is, the current performance of plants founded during Pro-alcohol is higher than the performance of those built before Pro-alcohol. On the other hand, the performance of plants founded after Pro-alcohol is not significantly higher than the performance of Prealcohol plants. Based on our fieldwork, a possible explanation for this result is that, in the long run, the industry ended up retaining the most promising plants and entrepreneurs from the Pro-alcohol period. Pro-alcohol probably generated a large pool to select from. Regarding the size of the impact, the prediction is that a Proalcohol plant produces 3.62 kg of sucrose per ton of sugarcane more than in the case of Pre-alcohol plants. This is roughly equivalent to a 4% increase in operational efficiency. In the Brazilian sugarcane sector, changes of this magnitude can have a significant impact on profits, especially during tight economic cycles characterized by low sugar and energy prices. As in many commodity businesses, operational efficiency is crucial.
Similar results are observed if we break up Pro-alcohol into subperiods (see Models (5) and (9) in Tables 6 and 7, respectively). Industrial policy plants founded during the periods 1975–1979 and 1980–1985 have an operational performance that is significantly higher than the performance of Pre-alcohol plants. Again, we find that Post-alcohol plants do not have a significantly higher performance than Pre-alcohol plants (Table 6). When we divide the industrial policy period into three subperiods (Table 7), we observe that the coefficients for each of the three subperiods are also Industrial policy and new industries 19 of 32 Table 5 Impact of industrial policy on performance
a Below the value of each coefficient are the heteroskedasticity-robust standard errors, shown in parentheses. y P50.10; *P50.05; **P50.01.
S. Mingo and T. Khanna 20 of 32 Table 6 Impact of industrial policy on performance
a Below the value of each coefficient are the heteroskedasticity-robust standard errors, shown in parentheses. y P50.10; *P50.05; **P50.01.
S. Mingo and T. Khanna 22 of 32 positive, though the coefficient for 1975–1977 is not significant. It is interesting to note that, despite its lack of significance, the coefficient for the subperiod 1975–1977 is larger than the coefficients for the other two industrial policy subperiods. Finally, the coefficient on Post-industrial policy plant continues to be not significant, that is, the performance of Post-alcohol plants is not significantly better than the performance of Pre-alcohol plants (Table 7).
In Models (2), (6), and (10), we include the variable Industrial policy entrepreneur, which indicates if the current owner of a plant is an entrepreneur who entered the industry for the first time during the period of policies. The coefficient on Industrial policy entrepreneur is always negative and significant. Plants managed by industrial policy entrepreneurs are, on average, less efficient than those plants that are managed Downloaded from http://icc.oxfordjournals.org/ at Harvard Library on July 10, 2014 by other types of entrepreneurs. Note that industrial units currently managed by industrial policy entrepreneurs are not necessarily Pro-alcohol plants—some industrial policy entrepreneurs have acquired and founded non-Pro-alcohol plants.
Existing industrial policy entrepreneurs seem to be, on average, of a lower ability than other types of entrepreneurs. Interestingly, in Model (2), the variable Industrial policy plant is still significant but with higher coefficients than in the case of Model (1). Thus, the coefficient on Industrial policy plant in Model (1) appears to be picking up the negative effect of the variable Industrial policy entrepreneur. A similar result is observed in Models (6) and (10).
As already discussed, several new ethanol and sugar companies were created during Pro-alcohol. According to our records, a total of approximately 100 plants were built during this period, almost doubling the total number of units. This led to a substantial increase in the level of fragmentation in the industry. Local farmers and business owners residing in small towns established many of the Pro-alcohol units— the government encouraged the creation of the plants in some of these localities.
Interestingly, and despite the wave of acquisitions and consolidation after the end of the program, the high level of fragmentation remained. Several underperforming local ethanol producers became entrenched in the industry given their importance in the economy of some of these small towns. For instance, during 2005—the last year in our data set—the four largest players in the sugarcane sector processed slightly more than 15% of the total sugarcane harvested during the year. Based on the statistical results in Model (2), and the nature of this persistent fragmentation, we conjecture that a considerable amount of low-ability industrial policy entrepreneurs were still part of the industry even after two decades since the end of the program.
In Models (3), (7), and (11), we include the variables Acquired plant and the interaction (Industrial policy entrepreneur)*(Acquired plant). In these three regressions, the coefficients on (Industrial policy entrepreneur)*(Acquired plant) are greater than zero with values ranging from 6.7 to 7.8. In the case of Models (7) and (11), the coefficients are significantly greater than zero with P50.10. Plants that were acquired by industrial policy entrepreneurs after the end of the program tend to have a higher performance than those plants owned by this type of entrepreneur but Industrial policy and new industries 23 of 32 that have never been acquired. Additionally, the coefficient on Acquired plant is not significantly different from zero and the coefficient on Industrial policy entrepreneur is still negative and significant. It is also interesting to note that in Model (11) the difference between the coefficient for 1975–1977 and the other two industrial policy subperiods gets even larger than in Models (9) and (10).
There are several interesting cases of successful sugar and ethanol producers that entered the industry during Pro-alcohol and later grew their companies through acquisitions. For example, Unialco was founded in 1980 with the support of Pro˜ alcohol in Guararapes, Sao Paulo. During the aftermath of Pro-alcohol, Unialco acquired Alcoolvale, an industrial policy plant located in the state of Mato Grosso ´ do Sul. Another example is Sabaralcool. This company—also created with ProDownloaded from http://icc.oxfordjournals.org/ at Harvard Library on July 10, 2014 alcohol support—acquired Cooperbal in the early 1990s. Cooperbal was originally established in the 1980s by a cooperative of farmers in the municipality of Perobal, ´ Parana. Based on our fieldwork and statistical results, we conjecture that (i) existing industrial policy entrepreneurs are, on average, of a lower ability than entrepreneurs that did not enter during the period of policies, and (ii) within the group of existing industrial policy entrepreneurs there is a subgroup of a higher ability level that were the ones making acquisitions after the end of the policy program.
In Models (4), (8), and (12), we include the variable Acquired plant and the interaction (Industrial policy plant)*(Acquired plant). In Model (4), the coefficients on Industrial policy plant and Acquired plant are not significantly different from zero.
However, the coefficient on the interaction between both variables is positive and significant with a value of 11.41. Similar results are obtained for the models that use subperiod dummies for Pro-alcohol (Tables 6 and 7). Plants founded during the Proalcohol period that were acquired during the aftermath of industrial policy have a higher level of operational performance than Pro-alcohol plants that were never acquired. Many of the biggest sugar and ethanol producers operating in 2005 participated actively during the post-industrial policy acquisition wave. Typically, these were family businesses that started professionalizing their companies during the aftermath of Pro-alcohol. Some of these companies, such as Cosan, followed the strategy of growing through acquisitions instead of greenfield investments because they saw a greater opportunity to create value by buying undermanaged assets.
Generally, these acquisitions were quite effective in terms of improving operational performance. Agricultural best practices were transferred to the new acquired unit relatively quickly, such as more effective use of fertilizers, improved monitoring of the crops, more efficient harvesting methods, and use of new sugarcane varieties and agricultural equipment. Based on these insights learned through fieldwork and Models (4), (8), and (12), we conjecture that successful entrepreneurs acquired Pro-alcohol plants that were owned by less skilled entrepreneurs. Through this process, a fair number of low-ability entrepreneurs should have been selected out. It is important to highlight that the coefficients on Industrial policy plant in Model (4) and the industrial policy subperiods in Models (8) and (12) get close to zero. This S. Mingo and T. Khanna 24 of 32 would be consistent with our argument that post-industrial policy acquisitions are a mechanism that could explain the superiority of industrial policy plants.
Certainly, other mechanisms might explain our results. For example, the longterm effects of industrial policy might be confounded with the imprinting effects of the technologies available at the time of founding of industrial policy plants. In other words, newer plants could be imprinted with better technologies. Even though we cannot completely rule out the possibility of a “positive” technological imprinting, it is reassuring that the operational performance of Post-alcohol plants in all of our regressions is not significantly higher than in the case of Pro-alcohol plants (Tables 5, 6, and 7). According to this result and those presented in Table 3, newness does not necessarily translate into better performance.
Downloaded from http://icc.oxfordjournals.org/ at Harvard Library on July 10, 2014 During our fieldwork, we also discussed directly with managers how technology at the time of founding could affect the future performance of plants. The main conclusion that emerged during these conversations was that investments in agricultural equipment and technologies do not have a significant imprinting or long-term effect on the overall operational performance of sugarcane units.14 Two arguments stood out. First, operational performance is mostly affected by agricultural practices and technologies that determine the quality of the sugarcane grown and subsequently harvested—the impact of the industrial phase of production is not as crucial (Martines-Filho et al., 2006). The fact that the agricultural phase represents 75– 80% of the costs of production of ethanol shows the importance of this stage compared to the industrial phase. Sugarcane companies focus most of their R&D investments on agricultural improvements, such as the development or purchase of new sugarcane varieties.15 Second, the development of new agricultural equipment and technologies in this industry are frequent, incremental, and dynamic. For example, companies and government agencies are frequently developing improved sugarcane varieties using biotechnology and genetic research (Mingo, 2013a). Also, companies regularly invest to upgrade their agricultural machinery, make land improvements, and install more technologically advanced harvesting systems.
As we discussed previously, government intervention has continued to be present in the industry. Even though current government intervention is of a different nature to that of the Pro-alcohol industrial policy program, this is another issue that could affect the interpretation of our results. Certainly, continued government intervention It is important to remember that, although the Pro-alcohol period goes from 1975 until 1985, we measure operational performance between 1999 and 2005. Therefore, the newest plant founded inside the Pro-alcohol period would be approximately 15 years old in 1999.
The sustained capacity to improve sugarcane productivity is one of the most important factors underlying the success and growth of Brazil’s sugar/ethanol industry. Sugarcane productivity has risen steadily at a 2.3% growth rate between 1975 and 2004. This growth rate is the result of new variety development, biological pest control, improved agricultural management, and greater soil selectively (Martines-Filho et al., 2006).