Economics of Microbial Inoculants as an Integrated Pest Management Practice in Apple Production by Holcer Chavez A thesis submitted to the Graduate Faculty of Auburn University in partial fulfillment of the requirements for the Degree of Master of Science Auburn, Alabama May 7, 2012 Keywords: Production Economics, Stochastic Frontier Analysis, Apples, Biological control, Microbial Inoculants Copyright 2012 by Holcer Chavez Approved by Denis Nadolnyak, Chair, Professor of Agricultural Economics and Rural Sociology Norbert Wilson, Professor of Agricultural Economics and Rural Sociology Valentina Hartarska, Professor of Agricultural Economics and Rural Sociology Joseph Kloepper, Professor of Entomology and Plant Pathology ii Abstract Demographic growth, stricter environmental concerns and regulations, and rising prices of chemical pesticides emphasize agricultural production alternatives that substitute chemical input use and increase yields. In recent years, Biological control technologies have emerged as a viable control strategy for plant disease. This thesis analyzes the impact of Microbial Inoculants (MI) technology on pesticide use and yields in apple production using 2007 farm-level data. The analysis employs a pesticide use function and different types of production functions including stochastic production frontier. The results show that pesticide use is not reduced by MI applications. However, the technology has a significant positive impact on the outputs. Adopters of the MI technology have 2.5% higher efficiency rates compared to non-adopters. iii Acknowledgments I wish to express my deepest gratitude to all of the people who have helped me during my time at Auburn. My sincerest thank to my advisor Dr. Denis Nadolnyak, for guiding me through this process and giving me the freedom and confidence to realize my capabilities. My genuine appreciation goes to Dr. Joseph Kloepper, who afforded me the opportunity to come to Auburn, giving me the chance to grow as a professional and human being. I also wish to thank m y other committee members, Drs. Norbert Wilson, and Valentina Hartarksa, for their patience and wise advices that helped to achieve my academic success. Special thanks filled with the utmost appreciation go to my family ? Holcer, Martha, Maby, Agustin and Luciana - in Lima and also to my partner in life Pamela, who was always there to support me. iv Table of Contents Abstract ......................................................................................................................................... ii Acknowledgments ....................................................................................................................... iii List of Tables ................................................................................................................................ v List of Figures .............................................................................................................................. vi List of Abbreviations .................................................................................................................. vii Chapter 1: Introduction ............................................................................................................... 1 Chapter 2: Background .............................................................................................................. 4 2.1 Apple Production ...................................................................................................... 4 2.2 Biological Control: Microbial Inoculants ................................................................ 11 Chapter 3: Literature Review .................................................................................................... 16 3.1 Microbial Inoculants ............................................................................................... 16 3.2 Economics of biological control .............................................................................. 18 Chapter 4: Methodology and Data ............................................................................................ 24 4.1 Data ......................................................................................................................... 24 4.2 Pesticide Use Function ............................................................................................ 27 4.3 Production Function and Stochastic Frontier .......................................................... 29 Chapter 5: Results ..................................................................................................................... 39 Chapter 6: Conclusions ............................................................................................................. 50 Bibliography ............................................................................................................................. 52 v List of Tables Examples of MI registered for use as control agents ................................................................ 13 Variable description .................................................................................................................... 39 Summary statistics for apple production ..................................................................................... 40 Pesticide use function ................................................................................................................. 42 Production function and stochastic production frontier ............................................................. 44 Average efficiency by adoption and state ................................................................................... 48 vi List of Figures Percentage of 2007 apple production by state .............................................................................. 5 2007 U.S. apple production by variety ......................................................................................... 7 Apples marketing-year average grower price and total utilized production, 1990-2008 ........... 11 Pesticide market share for biological control agents .................................................................. 14 Percentage and type of biological control used. ......................................................................... 26 MI, pesticides, and damage control relationship. ....................................................................... 46 vii List of Abbreviations MI Microbial Inoculants BCAs Biological Control Agents GMOs Genetically Modified Organisms U.S. United States ERS Economic Research Service USDA United States Department of Agriculture IPM Integrated Pest Management ICM Integrated Crop Management BT Bacillus Thuringiensis ARMS Agricultural Resource Management Survey EPA Environmental Protection Agency 3SLS Three Stages Least Square SPF Stochastic Production Frontier 1 Chapter 1: Introduction Continually enhancing crop production is essential to supply sufficient food for the increasing human population, satisfy energy demands, and provide essential industrial inputs. However, some current production methods used in agriculture create economic, environmental and health problems. Therefore, a key challenge for agriculture in the twenty-first century is to develop and implement agricultural production systems that maintain or enhance yields while also reducing negative side-effects. Such production systems have been referred to as ?environmentally friendly? or ?sustainable.? Disease management in crops worldwide is heavily dependent upon the application of synthetic (chemical) pesticides for pathogen and insect control. However, their excess application can enhance the development of pest resistance thus requiring more chemicals to control possible losses. Also, stricter regulations concerning the application of agrichemicals, in the United States are based almost entirely on the direct impacts on health and environment (White 1998). Moreover, the price of chemical pesticides have been increasing because of fuel price trends, uncertainty, and also because of increasing concentration of market power in the hands of a few big transnational producers who are becoming the only suppliers (Marcoux and Urpelainen 2011; Fernandez-Cornejo and Just 2007). All of this works against farmer?s profit- maximizing objectives and makes them look for alternatives that can result in higher yields. 2 In the last years, global demand for more environmentally friendly products and sustainable production systems has been increasing. In this context, biological control products offer an attractive alternative to synthetic pesticides. Biological control agents, by the broadest definition, are living organisms or natural products derived from them that can be used against plant damaging agents. Over the last two decades, biological control of plant pathogens has emerged as a viable pest and disease control str a t e g y (Harman et al. 2010; Singh, Pandey, and Singh 2011). Microbial inoculants (MI) are biological control (or often called ?biocontrol?) agents that include virus, bacteria and fungi. MI represents an environmentally friendly approach to reduce losses due to pests and diseases thus representing a potential alternative to chemical pesticides (Lugtenberg, Chin-A-Woeng, and Bloemberg 2002). Impact assessments of biological control are measured by cost-benefit analysis in an ex- ante situation but, for ex-post analysis, a production function is a standard procedure in agricultural production economics. The chosen crop for this study is apples as some MI products currently are in use and because, according to the United States-based Environmental Working Group (EWG), apples rank as the most contaminated fruit and vegetable produce (Lloyd 2011; Bagnato 2011). The general objective of this thesis is to evaluate the impact of the adoption of the MI technology on the U.S. apple industry. The first objective is to estimate the impact of MI use on pesticide usage. The second objective is to quantify the contribution of MI and other production factors and control variables to the U.S. apple yields and estimate production efficiency. The hypotheses of this study are as follow: First, as MI and synthetic pesticides control damaging agents, it is expected that the amount of synthetic pesticides used will be reduced only 3 in a small portion (as they are not perfect substitutes) by the adoption of this technology. Second, the impact over apple output is positive and significant, as is the impact of some other production factors and control variables. Production efficiencies are expected to be in the 50% to 80% interval having better efficiencies in producers applying the technology. 4 Chapter 2: Background The analysis of the impact of M I technology on the U.S. apple production starts with examining the industry?s production patterns. In this section, the essential components of production, like disease management components, are described. Then, the area of Biological control, including Microbial Inoculants, and its potential for future environmental regulations is explored. 2.1 Apple Production The apple is a pomeceous fruit of the apple tree, species Malus domestica. It is one of the most widely cultivated tree fruits around the world for human consumption. Apples grow on small, deciduous trees. There are more than 7,500 known cultivars of apples, resulting in a range of desired characteristics. Different cultivars are bred for various tastes and uses, including in processed food, fresh eating food and drinks. Domestic apples are generally propagated by grafting, although wild apples grow readily from seed. Trees are prone to a number of fungal, bacterial and pest problems, which can be controlled by a number of organic and non-organic means. 5 In 2007, there were more than 4.8 million acres of apple trees producing nearly 69 million metric tons. China is the first apple producer in the world, representing more than 42% of world production. The United States follows with 6.5% of world production. Poland is third, followed by Iran, Turkey, I t a l y , India, France, Russia, Chile, Argentina, Brazil, Germany, Japan and Spain. These top 15 producing nations accounted for more than 80 percent of total world production (ERS 2012). Apples are grown commercially in 35 states, y e t n early 92 percent of 2007 production came from only in eight states. Figure 1 shows the following rates: Washington with 59.4 percent, New York with 12.8 percent, Michigan with 6.2 percent, Pennsylvania with 4.5 percent, California with 3.8 percent, Virginia with 2.4 percent, North Carolina with 1.7 percent and Oregon with 1.3 percent (ERS 2012). According to the 2007 Census of Agriculture, there are 25,591 apple growers in the U.S. In 2007, more than 9.5 million pounds of apples were utilized for fresh and processed consumption. These apples were produced on more than 398,000 acres (2012). Source: Compiled by author. USDA?s Economic Research Service, 2012 Figure1. Percentage of 2007 apple production by state 6 Commercial growers typically use asexual reproduction methods of budding and grafting to grow stock for their orchards. These processes enable the growth of plants identical to the parents, which allows growers to ensure the type and underlying quality of the product. Grafting is where ?the upper part (scion) of one plant grows on the rootstock of another,? while budding uses the bud of one plant to grow on another (North Carolina Cooperative Extension 2012). These reproduction methods can be intensive and costly but ensure that the apple contains the exact traits that producers demand. Apples are self-incompatible and must be cross pollinated. Only few are described as "self-fertile" and are capable of self-pollination but they tend to carry larger crops when pollinated. Apples that can pollinate one another are grouped by the time they usually flower so cross-pollinators are in bloom at the same time. Pollination management is an important component of apple culture. Before planting, it is important to arrange for pollenizers - varieties of apple or crabapple that provide plentiful, viable and compatible pollen. Orchard blocks may alternate rows of compatible varieties, or may plant crabapple trees, or graft on limbs of crabapple. Apple pollen is heavy and is not carried readily by the wind as is the pollen of some tree species, such as conifers and nuts. The pollen is transferred primarily by insects, especially honey and bumble bees (Ferree and Washington 2003). Fruit growers rent honey bees from apiculturists during the bloom period, a minimum of four or five strong colonies per hectare being recommended in mature orchards. Apple production can be challenging for growers as it is a very perennial crop. This does not allow growers to have planting flexibility as happens with other crops. Growers make decisions based on different climate, biologic and economic conditions every year. Climatic conditions during bloom are critical for fruit set (Ferree and Washington 2003). 7 Apple trees vary by the number of nonbearing years after initial establishment: a standard apple tree takes six to ten years, a semi-dwarf takes four to six years and the most commercially common dwarf trees bear apples in two to three years of age (University of Arizona Extension 2012). These differences in the length of nonbearing years increase the difficulty of orchard establishments with heavy initial costs and no 5 revenues from those trees. The nonbearing years occur at the beginning and the end of the life of an orchard. The life expectancy also varies by size as the standard apple tree ranges from 35 to 45 years, the semi-dwarf tree ranges from 20 to 25 years and the dwarf tree ranges from 15 to 20 years (University of Arizona Extension 2012). These time frames for bearing years and life expectancy can also vary by variety. The decision of tree size and variety defines an orchard. Figure 2 shows the most common varieties grown in 2007 in the United States. The most common variety was Red Delicious followed by Gala, Golden Delicious, Granny Smith, Fuji and McIntosh making up more than 72 percent of U.S. total production. Source: Compiled by author USDA?s Economic Research Service, 2012 Figure2. 2007 U.S. apple production by variety Red Delicious Gala Golden Delicious Granny Smith Fuji McIntosh Other varieties 8 Cultivars differ widely by the time of ripening, the average being between 60 and 180 days after full bloom. Several methods are available to determine and/or predict optimum time of harvest such as temperature, the ethylene content of fruit and others. In Washington, which is the state with the highest production, apple harvest begins in late August and continues into October. Fruits are extracted manually and then they are transported from the field in large bins to warehouses where they are placed into standard cold storage or controlled atmosphere storage. Fruit is held for marketing in March through August of the following year. Before apples are packed, they are examined and those that have poor color or damaged by pests are removed and diverted to processing. Apples are washed, brushed and waxed prior to packing in boxes for shipment to market (Washington State University 2012). Apple production is very labor intensive and relies on labor for many crucial tasks. As previously mentioned, labor is used in harvesting, packing and also used for maintenance activities such as pruning and thinning. Water is very important to the function of the apple tree as water is the greatest component of the tree by mass and almost all critical processes can be limited by inappropriate water status. Insufficient water can induce excessive vegetative vigor and compromise fruit development; however, the excessive moisture of the soil can generate problems such as slow roots growth and leaching leading to nutrient deficiencies (Utah State University Cooperative Extension 2008). Irrigation is used primarily to provide supplemental water not provided by rainfall or soil water reserves. Consequently, efficient irrigation management requires knowledge of the water loss of the apple tree, the soil water reserves, and rainfall. There are several practical approaches used to estimate water status with experience being a very common and important one. 9 As many other plants, the mineral elements required by the plant for proper growth and fruit development include nitrogen, phosphorus, sulphur, potassium, calcium and magnesium. Other minerals such as iron, manganese, copper, zinc, boron, molybdenum and chlorine are required in lesser amounts. It is difficult to calculate the total nutrient requirements for apple trees since it is necessary to account for nutrients contained in the soil. Balance is the key for proper fertilization. For example, insufficient nitrogen results in symptoms including less vigor, light green to yellow-green leaves, less vegetative growth and low yields. However, excessive nitrogen can be equally bad causing too much vegetative growth, reduced bloom & fruit set, reduced quality of fruit, and diseases such as fire blight, brown rot and powdery mildew (Ranch 2012). Insects and pathogens that attack apple trees or their fruit are controlled primarily through the use of pesticides. However, biological control of pests (insects and mites) and diseases is achieved in a majority of orchards where selective chemicals and reduced pesticide rates are used. The major pest in apples is codling moth. Management decisions for codling moth have a big impact on many other pests in the orchard. The type of material used to control codling moth determines which of the other pests may develop to the point where additional treatment is required. It is important to mention that in orchards where organically accepted materials are used to control codling moth, problems with secondary pests are less frequent (University of California 1999). Apples are host to over 70 infectious diseases, the vast majority of which are caused by pathogenic fungi. Scab and powdery mildew are the major diseases in apple production. However, the disease that has a major concern is fire blight, which is caused b y bacteria. When this disease is epidemic, it can cause serious tree loss in nurseries and orchards even leading to orchard removal (University of California 1999). 10 Successful damage management usually involves integration of several methods of pest and disease control. This is called integrated pest management (IPM). The use of resistant rootstocks and scions, fungicides, bactericides, biological control agents, environmental modification and site selection are some of the means used to control apple damage factors. From now on for the easiness of the study, the word ?pesticide (s)? will be referring to the control of pests and diseases, not taking into account herbicides. This brief background of apple production has covered much of the most important components affecting apple production. All of these factors - trees management, bees pollination, fertilizers, water management and pest management - are directly included in the analysis. There are also other factors like Ph of the soil, temperature, and amount of light that influence apple productivity. These and other non-controllable factors (such as rainfall) are captured by state dummy variables. Average annual grower prices have an obvious correlation with total utilized production. Figure 3 shows the values from 1990 to 2008. It can be seen that prices are very influenced by the amount supplied each year. Overall, national production volume has been steady for the last years; however, visual examination of the data suggests that prices are becoming more sensitive to the changes in quantity produced. 11 Source: Compiled by author. USDA?s Economic Research Service, 2012 Figure3. Apples marketing-year average grower price and total utilized production, 1990-2008 2.2 Biological Control: Microbial Inoculants The definition of biological control has been evolving. The definition given b y the National Academy of Sciences in 1987 was: "the use of natural or modified organisms, genes, or gene products to reduce the effects of undesirable organisms (pests), and to favor desirable organisms such as crops, trees, animals, and beneficial insects and microorganisms" (Gabriel et al. 1990). Wilson?s definition (1997) is broader: ?The control of a plant disease with a natural biological process or the product of a natural biological process.? A current definition which is particularly useful is the one proposed by Pal and Gardener (2006) which says: ?Biological control refers to the purposeful utilization of introduced or resident living organisms, other than disease resistant host plants, to suppress the activities and populations of one or more plant pathogens?. Biological control agents (BCAs), or commonly called biopesticides, include predators, parasites, pathogenic microorganism, and competitors. According to the International Biocontrol Manufacturers? Association there are three categories: 0.000 0.050 0.100 0.150 0.200 0.250 0.300 0.350 - 2,000.0 4,000.0 6,000.0 8,000.0 10,000.0 12,000.0 1 9 9 0 1 9 9 1 1 9 9 2 1 9 9 3 1 9 9 4 1 9 9 5 1 9 9 6 1 9 9 7 1 9 9 8 1 9 9 9 2 0 0 0 2 0 0 1 2 0 0 2 2 0 0 3 2 0 0 4 2 0 0 5 2 0 0 6 2 0 0 7 2 0 0 8 Total utilized production (Million lbs.) Avg. grower price ($ per lbs.) 12 ? Macrobial: Insects, mites, nematodes, other non-microbial organisms. ? Microbial: virus, fungi, bacteria. ? Bio-rational: Natural products (plant extracts with insecticide or fungicide effects) and Semi-chemicals (behavior modifying agents for control of pest populations). However, according to several pest management researchers (Chandler et al. 2008; Copping and Menn 2000), a new category exclusively including genetically modified organisms (GMOs) is recognized in some countries, such as the United States. These GMOs are basically genetically modified plants that express introduced genes that confer protection against pests or diseases. BCAs are used in two types of agriculture. The first one is Organic farming where no chemical inputs are permitted. The second type, which is the focus of this study, is integrated crop production programs. This type of agriculture includes IPM strategies focusing on a reduction in pesticide use, resulting in improved conservation of the environment and better quality food (less pesticide residues). Biological control is considered in many ways to be the ideal pest-management tactic, because it tends to be environmentally innocuous, self-sustaining and low cost. Also, biocontrol agents can be applied together with chemicals, either in rotation to reduce the possible development of pathogen resistance or in an integrated pest management strategy with the goal of minimizing the use of synthetic pesticides. After reaching a volume of 34 billion USD in 1995, the synthetic pesticide market is declining slowly and continuously. For 2005, the volume of synthetic pesticide sales was 26.7 billion USD (Thakore 2006). This is due to the reduction of pesticide use (IPM) and the introduction of GM crop development. Although more than 1,000 different products or technologies are available through more than 350 manufacturers in the world, the use of BCAs is still marginal: in 2005 they accounted for only around 2.5% of total of plant protection inputs 13 market at end user prices with around 588 million USD (Guillon 2008). However, the use of biopesticides has been growing at an annual rate of 10% representing 4.25% of total pesticide market in 2010 (Ongena and Jacques 2008; Bailey, Boyetchko, and L?ngle 2010). Microbial Inoculants (MI) are control agents of agricultural pests developed from microbial natural enemies in the bacteria, protozoa, fungi and viruses. Of the known potential microbial control agents, only a very small fraction has been investigated for practical use (Chandler et al. 2008). While many technical and ecological challenges remain to the exploitation of microbial control agents, they can form valuable components of Integrated Crop Management (ICM). Table 1 lists some representative species used as commercial control agents. Table1. Examples of MI registered for use as control agents of agricultural pests Source: Chandler et al, 2008 According to Bailey et al. (2010), there are approximately 225 microbial biopesticides being manufactured in the 30 members countries of the Organization for the Economic Development and Cooperation (OECD). In the U.S., there are 53 microbial biopesticides registered. 14 Figure 4 shows the market share for microbial inoculants. MI represented 30% of total sales of biocontrol pesticides in 2006, 75% of which was represented by bacteria. The total value of sales for MI was valued at $205 million. Most of the bacterial strains exploited as biopesticides belong to the genera Agrobacterium, Bacillus and Pseudomonas (Fravel 2005). Bacillus thuringiensis (Bt), specifically devoted to insect pest control, accounts for more than 70% of total biocontrol sales (Bailey, Boyetchko, and L?ngle 2010; Ongena and Jacques 2008; Thakore 2006). Source: Ongena and Jacques, 2008 Figure4. Pesticide market share for biological control agents The U.S. Environmental Protection Agency?s (EPA) is the organism in charge of supervising and regulating the use of pesticides in the U.S. EPA does this under two major federal statutes. First, under the Federal Insecticide, Fungicide, and Rodenticide Act (FIFRA), EPA registers pesticides for use in the United States and prescribes labeling and other regulatory requirements 15 to prevent unreasonable adverse effects on human health or the environment. Second, under the Federal Food, Drug, and Cosmetic Act (FFDCA), EPA establishes tolerances (maximum legally permissible levels) for pesticide residues in food. However, there were always inconsistencies in the two major pesticide statutes. (EPA 2012) In 1996, the U.S. Congress unanimously passed a landmark pesticide food safety legislation supported by the government administration and a broad coalition of environmental, public health, agricultural and industry groups. President Bill Clinton later signed the Food Quality Protection Act (FQPA). The FQPA represented a major breakthrough, amending both major pesticide laws to establish a more consistent, protective regulatory scheme and grounded in sound science (EPA 2012). As the U.S. apple industry is a highly pesticide intensive industry, a big hit seemed to be coming (Roosen 2001). In 2006, EPA declared that the pesticide azinphos-methyl (AZM) cannot be used in apple production after September 30, 2012. While AZM provides important pest control benefits to growers of apples and other crops, it also has potential risks to farm workers, pesticide applicators, and aquatic ecosystems (Cassey, Galinato, and Taylor 2010). This regulation will bring big economic consequences and changes in apple production practices. For example, AZM has been the pesticide most used by Washington State apple growers since the late 1960s; and in 2008, 80% of Washington apple growers used AZM primarily to control codling moth (Cassey, Galinato, and Taylor 2010). In addition, in 2011, by the National Organic Standard Board (NOSB) voted to phased out by October 2014 the antibiotics streptomycin and oxytetracycline, which are the primary tools used by conventional and organic growers to prevent fire blight (Washington State University 2012). This may be the niche opportunity that MI technology needs for a fully development in this crop. 16 Chapter 3: Literature Review This chapter is divided in two sections. The first section compiles some of the many publications about application of microbial inoculants as biopesticides in greenhouse and controlled fields for apple production. The second section reviews economic studies related to this specific topic. Until the day this thesis was written, there were no economic studies addressing the impact of this specific BCA in ex-post situations. However, several studies about the most popular type of BCAs (GMOs) are reviewed as they will become useful in developing the methodology. In these studies, authors conduct impact analysis of GMO adoption. In theory, both types of BCAs, GMOs and MI, should have similar impacts on crops (reduction in synthetic pesticides and increase in yields). 3.1 Microbial Inoculants The main microbial insecticide registered and available for use in apple orchards is Bacillus thuringiensis (Bt). With multiple applications of this material, farmers have achieved some degree of control or suppression of Leafrollers, moths and fruit worms. Also important, but not as commercially available, is the granulosis or granulo virus. This virus is a highly selected targeted microbial insecticide that attacks codling moth (University of California 1999). 17 There are several studies about the efficacy of Bacillus for controlling many pests in apples. For example, Peighami-Ashnaei (2009) investigated fifteen strains of identified Pseudomonas fluorescens and Bacillus subtilis for biological control activity against Blue mold (Botrytis cinerea). Bacillus subtilis showed considerable results against Blue mold on apple fruits and could reduce the grey m o ld from 100% to less than 65% after twenty days. In addition, it was shown that bacterial strains could not only control the disease but they are a reliable replacement of a chemical pesticide called Thiabendazol. Laboratory trials were conucted by Cossentine (2003) to study how Bacillus thuringiensis subsp. kurstaki treatments on apple may be timed to maximize the survival of parasitoids of the obliquebanded leafroller (Choristoneura rosaceana) in the southern interior of British Columbia, Canada. The consumption of B. thuringiensis-treated leaves by host larvae significantly increased the percentage of dead host larvae in all parasitized and un-parasitized treatments. Previously stated, the most destructive pest in the apple cultivation is the codling moth (Cydia pomonella)(Pemsel et al. 2010). The Granulo virus has been proven to reduce codling moth development considerably. Virus uptake was found to be independent of active feeding and larvae became infected simply by walking or browsing on sprayed leaf disc surfaces in little time (Ballard, Ellis, and Payne 2000). However, its commercial development and use has been limited because of their high costs, slow action, short persistence and specificity relative to broad spectrum pesticides. The widespread development of strains of codling moth multi-resistant to insecticides and the desire to reduce dependence on pesticides have improved the commercial prospects of Granulo virus and use is likely to increase. Development of cheaper mass production techniques and possibly in vitro production are expected (Cross et al. 1999). 18 3.2 Economics of biological control Ex ante impacts of biological control are measured by cost-benefit analysis but, for impact analysis, a production function, lately including an integrated damage control function, is a standard procedure in agricultural production economics. In addition to this regression, a pesticide use function is often estimated to measure the substitution effect between biological control and chemical pesticides. This, if it is well specified, can also serve as an instrumental variable to avoid endogeneity in the production function using the two- or three-stage least squares regression (2, 3SLS). One key feature in current production functions is the distinction between inputs classified as standard factors of production (e.g. labor, land, capital, etc.) and damage control agents (e.g. pesticide, biological control). This distinction is important because the second group does not enhance productivity directly as standards inputs do but contribute indirectly by reducing output losses due to pest development (Lichtenberg and Zilberman 1986). Econometric investigations about damage control have had the tendency to rely on generic econometric models rather than to focus on knowledge about the physical and biological processes involved to specify the relevant functional form. This may generate biases of big proportions in estimates of productivity and unreliable conclusions about efficient input usage. This phenomenon has been occurring depending on the analysis? approach. Theoretical and normative empirical models of pest management at macro levels have incorporated the available entomological knowledge in their specifications and have derived optimal management practices and policy recommendations based on this premise. In contrast, Econometric measurements of pesticide productivity have been derived from standard production theoretical models such as the Cobb-Douglas specification (Lichtenberg and Zilberman 1986). 19 Damage control inputs should be incorporated into production analysis in a different way than regular production inputs. Models of biological and physical processes are used to obtain specifications of production processes with damage control inputs. These specifications are appropriate for micro level analysis (farm or field level). Heterogeneity among producers and different climatic conditions mean that a proper aggregation procedure should be incorporated to derive a appropriate regional analysis. Specification of the role of damage control agents in production functions has two important implications for theoretical and empirical work. The first one is that commonly used types of production functions specifications overestimate the productivity of damage control inputs even in larger samples. This upward bias happens because of a misspecification of the shape of the marginal factor productivity curve of damage control inputs which decrease more rapidly in the economic range than standard specifications assume. Second, damage control specifications have a different way to handle changes in damage control productivity through time. Using pesticides as an example, the spread of resistance through a pest population is an important problem. Thus treating pesticide in the same way as a regular production input will lead to predict behavior contrary to observed fact. In standard production functions, decreasing input effectiveness is reflected in decreasing marginal input productivity and thus in reduced level of input. In damage control specifications, decreasing effectiveness may increase input demand. This is exactly what is observed looking at pesticide use trends (Lichtenberg and Zilberman 1986). Litchenberg and Zilberman (1986) established different possible specifications of damage control function such as the Pareto distribution, the exponential distribution, the logistic distribution and the Weibull distribution. 20 Jankowski et al. presented a paper at the conference on international agricultural research for development held on Tropentag, Germany (Jankowski et al. 2007). In the paper, they analyzed the impact of a biological control agent (insect) on the diamondback moth in cabbage production in Kenya and Tanzania. They presented a pesticide use (cost) function and a production function. It was found that pesticide expenditure was 34% lower in areas where biological control was present; however, the production function showed mixed results. The biological control coefficient was positive and significant for the exponential damage control function but negative and significant for the logistic one which generates seriously questions the correctness of these results. Several studies on the impact and economics of GMOs will now be reviewed. Studies on using Bacillus thuringiensis (Bt) technology will specifically be discusses as they work very similar to MI. Bt crops produce proteins that are toxic to larvae of some insect species making it a pest-control agent that can be used, to some extent, as a substitute for chemical insecticides. Therefore, MI and Bt crops have similar properties and similar effects. All of these studies have incorporated this technology as a dummy variable, so did this thesis. Some studies made in China are very useful for this thesis. Of the list of developing countries, China was the only one that had introduced Bt-cotton on a large scale. Recognizing the negative externalities of excessive pesticide use, China?s government has made an effort to regulate pesticide production, marketing and application since the 1970s. The experience with regulation, however, has shown that when officials only promulgate rules and monitoring costs are high, reductions in the use of pesticides, the elimination of banned toxic ones, or the increase in the adoption of safe application procedures do not always follow. As a result, real reductions in the use of pesticides may have to depend on alternative approaches, such as the introduction of 21 new technologies. An observation on the background that emerges in several studies in China is that regardless of whether farmers use Bt or non-Bt varieties, the actual level of pesticide use dramatically exceeded its economically optimal level as computed from estimated factor productivity. The authors attribute this overuse to anecdotal evidence about misguided extension advice. Since part of the income of extension workers stems from pesticide sales they have an incentive to encourage farmers to use more pesticides. They cite some studies where some other authors found that the majority of farmers in China still considered the cotton bollworm as a problem although all were using Bt-cotton. Such observations show that although the economic benefits of Bt-cotton in China were demonstrated at an early stage of adoption, the sustainability of these benefits can be questioned. They also indicate that pesticide reduction requires other (supplementary) means such as a policy changes. This observation may be of interest for this study. A study made by Pemsl et al. (2005) used panel data of 150 farm households in the Shandong province in China for cotton production. Using the exponential damage control function, they found that there was a prevailing high level of insecticide use, despite Bt-cotton adoption. They offered the situation in the local seed markets as a possible explanation of this behavior. A vast number of different Bt varieties are available in local markets, with striking differences in price. They explained that this difference in Bt seed prices can only be explained by counterfeit varieties, thus not expressing the actual or aggregate impact of the technology. On the production function side, they found that the impact of Bt toxin on cotton yields was positive but not significant. Another study of cotton productivity in China was made by Huang et al. (2002) showing different results. This time they surveyed 282 cotton farms using cross-sectional data but, in this 22 case, putting more emphasis on provinces where Monsanto seed varieties were commercialized (to avoid any counterfeit issues). Their pesticide use regression analysis got a negative and highly significant coefficient on the Bt cotton meaning that Bt cotton farmers sharply reduce pesticide use when compared to non-Bt cotton farmers. Ceteris paribus, Bt cotton use allowed farmers to reduce pesticide use by 35.4 kilograms per hectare. For the production function, they used a regular Cobb-Douglas and also a Weibull and an exponential damage control function. All production functions obtained positive and significant coefficients for the Bt technology. Other developing countries are using this technology as well. Studies of Argentinean cotton (Qaim and de Janvry 2005) and Indian cotton (Qaim 2003) have shown similar results. In the first one, panel data with 299 cotton farmers having 89 adopters and 210 non-adopters was used, while in the second there was a cross sectional sample of 157 farm households chosen randomly from seven different states. In both studies, the technology decreases insecticide use significantly being the net effect a saving of 1.2 kg per hectare and 0.4 kg per acre respectively. In addition, Bt technology also affected the outputs positively in both studies, whether under the Cobb-Douglas or the damage control specifications. In the review made by Qaim (2009) about the economics of GMOs, he shows that available impact studies of insect-resistant crops show that these technologies are beneficial to farmers and consumers, producing large aggregate welfare gains as well as positive effects for the environment and human health. Bt crops can contribute significantly to global food security and poverty reduction. However, Bt does not completely eliminate the need for insecticide sprays because some crop damage still occurs when the technology is used. The reason is that Bt toxins are very specific to certain pest species, whereas other insect pests, remain unaffected. His compile of results confirm that both insecticide-reducing and yield-increasing effects can be 23 observed internationally. Studies already reviewed are compared. As we already saw, yield effects of Bt cotton are highest in Argentina and India. For Argentina, his explanation is simple: Conventional cotton farmers underutilize synthetic insecticides, so that insect pests are not effectively controlled. In contrast, in India, insecticide use in conventional cotton is much higher. He suggests that factors other than insecticide quantity influence damage control in conventional cotton and, thus, the yield effects of Bt technology. These factors include insecticide quality, insecticide resistance, and the correct choice of products and timing of sprays. So we have seen reviews about GMOs applications at farm level studies, especially Bt cotton, as they have a similar impact as MI on production. However, MI has a big advantage over GMOs because the second group has aroused significant opposition. According to Qaim (2009), public reservations about this technology are very strong in Europe and are gradually moving over to other countries and regions through trade regulations, public media, and outreach efforts of anti-biotech lobbying groups. The major concerns about the use of these biological control agents are related to potential environmental and health risks (Scientifics think that biological processes may be lose, and also there are no long term studies on human effect of this technology), but there are also fears about adverse social implications. Quoting Qaim (2009) ?some believe that this technology could undermine traditional knowledge systems in developing countries. Given the increasing privatization of crop improvement research and proliferation of intellectual property rights (IPRs), there are also concerns about the potential monopolization of seed markets and exploitation of smallholder farmers.? All of these does not happen with MI as it is a more ?nature providing? technology. 24 Chapter 4: Methodology and Data This chapter describes the methods behind the model development, as well as the details of the model including the data used. A pesticide use function and several types of production functions, including a stochastic frontier, are estimated using STATA 12 and SAS 9.2. 4.1 Data USDA?s 2007 Agricultural Resource Management Survey (ARMS) data on apple production was used for this study. This survey contains information on the production practices, inputs and costs, and financial performance of America?s farm households. Most of direct inputs and household characteristics come from the Phase 2 part of the survey while other variables such as yields and area harvested come from the Phase 3 part of the survey. The ARMS data has 4 specific unique characteristics which make it a valuable tool for this study: ? The ARMS survey has a broad coverage, including all major States producing a particular commodity, and generally covers more than 90 percent of the acreage of targeted commodities. 25 ? The ARMS survey uses a stratified random sample where each farm represents a known number of similar farms in the population based on its probability of being selected. Each farm is weighted by the number of farms it represents so that the ARMS sample can be expanded to reflect the targeted population. ? ARMS enterprise costs-of-production data contain sufficient detail about specific inputs to isolate the seed and pest control costs used to produce a given commodity. ? Enterprise costs of production can be estimated for each observation in the ARMS data so that a distribution of costs can be developed. Summarizing, in this study, each farm is weighted by the number of farms it represents so the sample can be expanded to reflect the targeted population. Only conventional (non organic) farmers were considered as the intent was to estimate the technology? impact on regular pesticide usage. Under the ?pest management practices? section of the production practices and costs reports (phase 2) of the survey, an item referring to biological control was used as the variable of interest. This is a binary choice variable taking the value of ?1? if the farmer was using the technology and ?0? otherwise. It would have been advantageous to use a quantitative measure of the MI applications but as only a small percentage of farmers were using this technology, a dummy variable seems more appropriate. In the sample of 547 conventional farms, 197 farms were using on average 3 biological control products, from which the main ingredient included one of the following: Granulovirus, Bacillus thuringensis, Bacillus subtilis, Bacillus pumilus and Thricoderma sp. Figure 5 shows the percentage represented by each biological agent, from which, 96% fall into the MI definition. 26 Source: Compiled by author. 2007 apple ARMS data Figure5. Percentage and type of biological control used. MI provides good resistance to different varieties of insects and diseases for apples. The main microbial pesticide used is any type of bacillus, especially Bacillus thuringensis (Bt), due to its ability to suppress many pests at the same time. For example, the Granulovirus is used against Codling moth (Cydia pomonella), but Bacillus thuringensis has been proven to work against Codling moth, Apple pandemis, Leafrollers, Western tussock moth, Velvetbean caterpillar and Green fruitworm (University of California 1999). Bacillus subtilis has been proven to work against Fire Blight, Botrytis, Sour Rot, Rust, Sclerotinia, Powdery Mildew, Bacterial Spot and White Mold (Peighamy-Ashnaei et al. 2008; Sundin et al. 2009). However, there are many other pests and diseases to which MI agents do not provide resistance. Therefore, MI does not completely eliminate the need to use chemical pesticides. Seven states were represented in the survey: Michigan, Oregon, New York, Pennsylvania, North Carolina, California and Washington. Washington was used as the base for 47% 32% 4% 11% 2% 4% % of Biological Products Used Granulovirus B t . K u r s t a k i Bacillus subtilis Bacillus pumilus Thricoderma sp. O t h e r s 27 its continuous, successful production history, and because it is the state with more total production (ERS 2012). 4.2 Pesticide use function As it was stated before, MI provides a good alternative to control some of the most important apple damaging agents such as the Codling moth. However, there are some major apple pests to which the technology does not provide resistance to, such as scab, powdery mildew and fire blight (provides only mid resistance). Therefore, MI does not completely eliminate the need to spray chemical pesticides in order to avoid pest damage. That is t why, it is not totally accurate to say that chemical pesticide usage may be reduced by the application of this technology as different pests are more prone to happen in different regions. As a first step, the summary statistics of those farmers using and not using the technology are compared to have a quick look at what might have been happening. The variable pesticide includes insecticide and fungicide applications (that are the ones that can behave as pesticides substitutes) not including any biological control product. In order to confirm the findings, a more precise quantification was needed. A double-log type functional form was estimated using OLS regression to quantify the technology?s impact on the pesticide use. A linear type functional form was also estimated for comparison purposes. These regressions were calculated using plot and farmer characteristic. The quantity of pesticide (pest) application is expressed in pounds per acre. The double-log model expressed in its linear form is as follows: Log (Pest) = A + ? 1 Log (price) + ? 2 Log (size) + ? 3 (MI) + ? ? i (K) + ? (1) 28 where A is the intercept and Price is a proxy for pesticide?s price. Size is used to reflect farm?s characteristic and refers to the actual farm size. MI is a dummy variable which takes the value of one for MI plots and zero otherwise. K is a vector of other determining factors such as experience (characteristic), an index reflecting pest pressure, and state area variable (dummy) as proxy for the different agro-climatic conditions found in these areas. Lastly, ? is the random error term with zero mean. Although only a single cross section of farms is used, large variations in the price of pesticides exist among the respondents, reflecting the differences in pesticide quality, pesticide prices at different times during the growing season, and the pesticide composition. Price is measured as the unit value price of pesticide purchased by the farmer. I calculate the unit value price for each farm by dividing the value of their pesticide purchases by the quantity that they purchased. Direct production inputs were not included as exogenous assuming damage control expenditures is a separate budget category. Yields are also not included because, as it was said already, a production function is estimated in the next step. This approach was taken because endogeneity of inputs is a potential problem with production functions estimates based on farm survey data. Pesticides in particular may be problematic if they are applied in response to high pest pressure as high levels of infestations may be correlated with lower yields (Huang et al. 2002; Qaim 2003; Qaim and de Janvry 2005). To avoid this possible econometric problem, the Instrumental Variable (IV) approach was adopted. An instrument for pesticide application in this case is a variable that is highly correlated with actual pesticide use but is not correlated to output except through its impact on pesticides. In this case, the predicted value of the pesticide use is 29 used. As long as the variables explaining pesticide use do not have any independent explanatory power on yields, the IV approach should allow me to better examine the impacts of MI and pesticides on apple output. Following Huang (2002) and Qaim (Qaim 2003, 2009) and Qaim and De Janvry (Qaim and de Janvry 2005), to implement the IV identification strategy, a number of control variables ? such as experience and the six states dummy variables ? were included in both the yield and pesticide use equations. The IV passed the Hausman-Wu exclusion restriction statistical test. 4.3 Production function and stochastic production frontier A production function or frontier is defined as function that, given available technology, specifies of the maximum amount of output possible for a given input mix. Production functions can be estimated from sample data (in this case cross-sectional data). Different types of production functions are estimated to measure the impact of the MI technology on apple production. Production function The first step in any parametrical empirical application is to select an appropriate functional form for the production function. A few mathematical forms of production functions are commonly used (those that are easy to manipulate). Every analyst should first appeal to technical (biological, chemical, nutritional, etc.) theory for specification of the functional form for modeling the particular production process in question. Following Beattie et al. (Beattie, Taylor, and Watts 2009), we use the Cobb-Douglas functional form. 30 ? = ?? ? ? ? ? ? ? ? (2) where A is a scalar referred to as a measure of total factor productivity, ? ? is one of the production factors and ? ? is the parameter to be estimated (same treatment for subscript 2). The Cobb-Douglas is easy to estimate and mathematically manipulate but is restrictive in the properties it imposes upon the production structure. For example, it has convex to origin and negative slope isoquants (input bundles for any given output) but it has unitary elasticity of substitution; it does not allow for technically independent or competitive factors. Marginal physical productivity (MPP) and Average Physical productivity are monotonically decreasing functions for all x given 0