Memoire Online - Using the WACC methodology to improve the assessment of projects in the french farming industry. Empirical evidences from farm's results of Isère

Dissertation / Project / Internship Report: International Management Project 2011-2012 Student Name: Bibard Anaël

Summary: The Purpose of this Thesis is to calculate the cost of capital for a typical farm in Isère. This cost of capital is use as an actualization rate to estimate the NPV of projects or the value of a farm based on this profitability.

Using the WACC Methodology to

Improve the Assessment of Projects

in the French Farming Industry

Empirical Evidences from Farm's Results of

Isère

Table of Content

1.2 The Profitability Heavily Relies on Subsidies, Not on Investment Decision 14

4.2 Calculation of the Historical Results Using the Data from CERFRANCE Isère 38

4.3 Calculation of the Capitalization Rate Using Bond Yields Plus Risk Premium 49

Abstract

The remarkable rise of the soft commodities prices in 2007/2008 represented just a reminder of the importance of agriculture for humanity. The Arab spring found its base on this mold and pulled down many regimes considered as stable few months before, like Tunisia or Egypt. Eventually, these price records could be reached again in 2012, due to a severe drought in the USA. To meet the challenge of feeding the world while developing more sustainable forms of production, advisory services must play a great role to help farmers in their decisions regarding their production as well as their investments. However, it appears that some financial consultants use really low actualization rates for agriculture, even lower than the risk-free rate. The purpose of this paper is therefore to present the different methods that could be used to determine more appropriate discount rates for farming businesses, and particularly the Weighted Average Cost of Capital and the bond yield plus risk premium model as presented by the French tax authority.

Historical results of farms from Isère over 5 years were studied to determine if the leverage has a significant impact on the profitability, and if signs of financial distress could be identified for the higher leverages. The statistical tests performed confirmed the underlying hypothesis of the WACC theory: the cost of capital is reduced by the increase of the leverage, up to a limit where financial distress overcomes the advantage of the lower cost of debt. The optimal capital structure for farms of Isère seems to range between 40-60% leverage, and up to 60-80% in the cases of the most stable production such as dairy farms.

This paper provides indications about how to establish the actualization rates for agricultural consultants, based on the WACC methodology and the method recommended by the tax authorities. The data analysis confirmed that the actual rates used by practitioners are clearly undervalued, leading to an over-valuation by two to four times in their asset valuation studies. The WACC and the bond yield plus risk premium methodology both seem to be applicable for small and medium farming businesses. However, further researches are necessary to validate the robustness of the WACC methodology in the agricultural context for non listed companies.

Abbreviations

CNCER: Conseil National des CERFRANCE (national council of the CERFRANCE) EBITDA: Earnings Before Interest, Taxes, Depreciation and Amortization

FADN: Farm Accountancy Data Network, or RICA: Réseau d'Information Comptable Agricole FAO: Food and Agriculture Organization

OECD: Organization for Economic Cooperation and Development REP: Required Equity Premium

SAFER: Société d'Aménagement Foncier et d'Etablissement Rural SAS: Statistical Analysis System

SMIC: Salaire Minimum Interprofessionnel de Croissance. The minimum salary in France. SPSS: Statistical Package for the Social Sciences

Table of Illustrations

Table 1: Subsidies and recurring net profit before tax in France per farm, in K€Source: FADN (RICA) 16 Table 2: Assumptions and recommendations of the 129 books that assume that REP = EEP. Source:

Table 3: Equity premiums recommended and used in textbooks. Source: Fernandez 2010 22

Table 11: Number of companies studied for the Beta estimation. Source: Bloomberg 35

Table 14: Answers of question 7: «how do you choose the discount factors you use?» 37

Table 15: Answers of question8 «what discount factor do you use usually for your customers? (ex given: 11%)» 37 Table 16: Answers of question 9: «If you were proposed a tool to estimate the WACC of the farms of your region to use it as a discount factor, how would you consider this tool?» 37 Table 17: Results of the Mood's median tests summary for ROE by groups of leverage for each years

Table 20: results of the Mood's median tests summary for ROA by groups of leverage for each years

Table 23: Calculation of the Beta for the grain specialization (oilseed is considered as grain farming in

Table 24: Calculation of the beta for the cattle ranching and farming specialization. Source:

Table 25: Calculation of the beta for poultry and fruits & vegetables Source: Bloomberg 52

Table 26: Summary of the beta for the different specialization and regions Source: Bloomberg 52

Table 27: Characteristic of each specialization in Isère *EBITDA/total annuity 53

Table 28: WACC estimation for each specialization (leverage is based on table 27) 53

Table 29 : WACC estimation for each specialization for group 5 (0 to 20% leverage) 54

Table 30: WACC estimation for each specialization for group 2 (60 to 80% leverage) 54

Table 31: NPV of a refinancing operation to reduce the WACC and increase the wealth of a dairy farm

Figure 1: The optimal amount of debt to reduce the company cost of capitalSource: Principles of

Figure 4: The percentage of the EU budget allocated to the CAP Source: Eurostat, in Clipici 2011 15

Figure 8: Volatility of the price of wheat after application of Agenda 2000Source: Offre et Demande Agricole 18 Figure 9: Fluctuation of the share of agro-food exports by key countries in the global volume of agrofood exports 18 Figure 10: Moving average (last 5 years) of the REP used or recommended in 150 finance and valuation textbooks. Source: Fernandez 2010 22 Figure 11: Small firm premium (bottom 10% of market cap in US) between 1927-2010. Source:

Figure 16: Box and whisker plots with median notch for ROE by year (means are shown with a red cross) 40 Figure 17: Box and whisker plots with median notch for ROE by specialization (means are shown with a red cross) 40 Figure 18: Median ROE for each year and each group of leverage for dairy* means significantly higher than ** 41

Figure 19 : Median ROE for each year and each group of leverage for grain* means significantly

Figure 20: Median ROE for each year and each group of leverage for cattle* means significantly

higher than ** 43 Figure 21 : Box and whisker plots with median notch for ROA by years (means are shown with a red cross) 45 Figure 22 : ROA for each year and each group of leverage for dairy producers* means significantly higher than ** 46 Figure 23: ROA for each year and each group of leverage for grain producers* means significantly higher than ** 47

Figure 24: ROA for each year and each group of leverage for cattle farming* means significantly

Appendix 1: Illustration of the database extraction. File name: "extraction 2006_1" 70

Appendix 2: Illustration of the database extraction, balance sheet information. File name: "extraction

Appendix 4: Normality test of the datasets for ROE, extraction from statgraphics 75

Appendix 5: Normality tests of the datasets for ROA, extraction from statgraphics 77

Introduction

Food safety and agricultural commodities prices played a great role in the recent Arab spring (Breisinger, Ecker, & Al-Riffai, 2011). The Arab spring, also called the Arab revolution, is a wave of protestations that started in Tunisia in December 2010 with the suicide of a student who couldn't pay for the fruits he was trying to sell in the street as a subsistence job. This situation is not a surprise for many observers, as the farming business is facing once again one of its historical challenges: feed the world and its increasing population. The access to the essential commodities, such as energy and food, is taken for granted nowadays in the developed countries. However, the land resource is finite and even subject to a regular contraction due to urbanization and salinization. Moreover, the increasing consumption of meat is less efficient in terms of global productivity than the direct human consumption of grain. The remarkable rise of the soft commodities prices in 2007/08 represented just a reminder of the importance of agriculture for humanity and the inelasticity of this market. The world market prices are oriented in the same direction in 2012, and the records reached in 2007 are getting closer due to a severe drought in the USA (Damgé, 2012). The OECD and the FAO consider that these high prices will become usual in the next decade (OECD-FAO, 2010).

Some new challenges arose more recently, and they are not easy to meet. The public opinion demands farmers to develop new forms of production more sustainable and respectful to the natural environment (Faure & Compagnone, 2011). Moreover, they have to maintain an economic activity in rural areas (Mundler, Labarthe, & Laurent, 2006). To achieve these challenges, the advisory services play a great role to deliver the best advices, for agronomical innovation as well as economical consulting (Faure, Desjeux, & Gasselin, 2011).

However, some financial consultants for farming businesses use low discount rates. This empirical observation has been made in a company member of the network of CERFRANCE, the leading accounting network in France, even ahead of the big four for the French accounting market. This leading position is due to the really high market share in agriculture and small businesses. Someone could have expected to find really innovative financial methods to produce better financial advices for farmers, but the consultants do not use the NPV method a lot, mainly because of the lack of information about the calculation of appropriate discount factors. Research in the field seems to be not sufficient to promote this method among consultants.

The purpose of this paper is therefore to do an exploratory research about the Weighted Average Cost of Capital (WACC) methodology and its applicability to the agricultural sector. To reach this main objective, many other research questions are addressed:

- Which methods are used by the practitioners to determine actualization rates for farming project in France?

- What is the optimal structure for a small or medium farm, considering the risk of financial

This research is based on the existing work produced by other authors, mainly in other economic sectors, but also on primary data extracted from the database of the CERFRANCE Isère. This exploratory work is subject to caution, because its subject hasn't been enough covered yet for the agricultural sector and the assumptions taken have a strong impact on the results. However, it is essential to improve the current knowledge about the cost of capital and the actualization rates in

agriculture. This paper aims to rough out this subject and to improve immediately the practices of the agricultural consultants of the CERFRANCE.

1 Significance of the Research

«Risk is like love; we all know what it is, but we don't know how to define it»

According to Allen, Myers & Brealey (2008, pp. 966-967) the capital asset pricing model (CAPM) and the Net Present Value (NPV) are ones of the most important concepts in finance. The CAPM allows investors to identify the non-diversifiable risks, by measuring the impact of a change in the aggregate value of the overall economy on the value of their investments. The NPV on the other side helps managers to choose between different projects, or even to choose to kill or not a project. The CAPM relies on the calculation of betas, and NPV relies on the choice of the discount factor. However, there are many ways to evaluate this discount factor. First, when a manager has many options or projects, he can set the discount factor at the opportunity cost of capital which is the profitability of the best project to evaluate each project. But this method does not take into account the risk associated with each projects and is not helpful when a manager has only one project, or when he just wants to evaluate the economic profitability of its company at a fair market price. In the last case, the appropriate way to set the discount factor is the CAPM and the calculation of the Weighted Average Cost of Capital (WACC) formula (Brealey, Myers, & Allen, 2008).

The WACC formula is of particular interest for our concerns, because it gives a good estimation of the company cost of capital, which can be reduced by financing the company using the optimal amount of debt as the traditional approach suggests (Brealey, Myers, & Allen, 2008, p. 504). Figure 1 illustrates this theory.

Figure 1: The optimal amount of debt to reduce the company cost of capital Source: Principles of Corporate Finance

In the French farming industry, the NPV methodology is nearly never used. When this method is
used, discount factors are chosen without clear explanations, and we can observe important

variations between practitioners, even in the same business unit. Is it because everyone considers that there is merely no risk in farming business? However, such hypothesis does not make sense considering that a farm manager needs to handle the weather risk, the crop's pests risk and the price volatility risk. Is the WACC methodology not applicable in farming business, or just the NPV method is not enough known and risks related to farming business poorly assessed?

1.1 The Farming Business in France

French agricultural exportations represent USD 17.5 billion (€ 13.2 billion) (Agreste, 2011; IMF.Stat, 2012), making France the 14^th biggest exporter of agricultural products in 2010 (WTO, 2011). Regarding food, France is the number three exporter in 2010, with USD 47.8 billion (€ 36.2 billion)(Agreste, 2011; WTO, 2011; IMF.Stat, 2012). France is also the number one agricultural producer in Europe, with € 62.0 billion in 2009 over a global European production of € 327 billion (Agreste, 2010). This high level of production and exportations is allowed by two factors: productivity and surface. The total cultivated area (TCA) in France represents 29.3 M ha in 2009, over the total 54.9 M ha of France (Agreste, 2010), and wheat yields of 7.04 T/ha are 88% superior to the world average of 3.74 T/ha (FAO, 2012).

However, the high level of production in France is not due to the size of the farms, because the average farm in France is only 55 ha in 2010 (Agreste, 2011 b) against 350 ha for US farms (Ambassade de France à Washington, 2009). The diversity of production is really high in France as we can see in Figure 1, as no production represent more than 16% of the overall production in value.

Regarding Isère, the production is also really diversified as we can see in Figure 3. The overall production represents € 516 million for 2010, for a total GDP of around € 33 billion in Isère for 2009 (CCI Rhône Alpes, 2011). The average size of the farms in the department is smaller compare to France with only 38 ha in 2010 (Agreste, 2011 c), and the TCA in Isère (241 300 ha) represents only 0.8% of the national TCA.

1.2 The Profitability Heavily Relies on Subsidies, Not on Investment Decision

To understand the farming business in France, it is necessary to know the Common Agricultural Policy (CAP) and its impacts on Agriculture. The CAP is the first and most important common policy for the European Union, and its cost jeopardized the overall EU budget for years (Figure 4). The CAP finds its origins in the 1950's, when European agriculture was on its knee after World War II. Charles De Gaulle considered this policy as essential to prevent major social events in France (Moravcsik, 1999), as he said that agriculture was as important as the troubles in Algeria. This Policy helped to increase the productivity, stabilized the agricultural markets, and secured food supplies of all European countries (Clipici, 2011; Zaharia, Tudorescu, & Zaharia, 2009; Howarth, 2000). According to Howarth (2000), its main objectives given at its start in 1962 were:

Figure 4: The percentage of the EU budget allocated to the CAP Source: Eurostat, in Clipici 2011

However, the European CAP had been heavily criticized over its history, either by European partners such as the Cairns group (Australia, Brazil...) and the USA (Spencer, 2003), or by European countries also such as the United Kingdom or Germany (Howarth, 2000; Elekes & Halmai, 2009; Moravcsik, 1999). The major points of friction are the market distortions induced by the CAP, the cost of this policy and the impact on the poorest economies (Howarth, 2000; Spencer, 2003; Borrel & Hubbard, 2000; Rickard, 2001). Some authors even argued that world agricultural prices could rise by 38% if subsidies were turned off (Borrel & Hubbard, 2000). Many negotiations rounds took place over the last decades to reduce these distortions, and the CAP was reformed many times since 1992 (see Figure 5).

In 1992, the CAP was reformed to decrease export subsidies and market supports to shift to coupled
direct payments. Those payments (direct aids in Figure 6) were proportional to the surface cultivated

by farmers, but not directly to the volume of production. This reform helped to reduce the production surpluses, and world agricultural markets distortions were reduced. However, the distortions were still significant after the reform as the overall amount of subsidies was really high, around € 35 billion in 1992 (Clipici, 2011). All the other reforms had the objective to reduce these distortions, such as the decoupled payments. New conditions for granting were also added to the CAP such as environmental protection, risk management, animal welfare, budget reduction, etc...

According to the Farm Accountancy Data Network (FADN), the amount of subsidies represents 80% of the average recurring net profit for French farms (see Table 1). Besides, grain producers depend even more on the CAP than other types of farms.

Table 1: Subsidies and recurring net profit before tax in France per farm, in K€ Source: FADN (RICA)

This dependence on subsidies is negative for innovation, as all the production is oriented by the CAP and not by the market demand. Moreover, the profitability depends on the capacity of the farmer to maximize the subsidies. New-Zealand is an example of a country that stopped all subsidies in 1984 (Gardner, 1994; Saunders, Wreford, & Cagatay, 2006; Sandrey & Scobie, 1994), and enjoys nowadays an enviable situation in the world agricultural market. This small country is the world largest milk exporter (Evans, 2008), and the second largest sheep producer (FAO, 2012).

Another element limits the innovative capacity of French farms, and also their adaptability: the control of the surface. As a matter of fact the SAFERs, Societé d'Aménagement Foncier et d'Etablissement Rural, have the power to control the land market. According to their mission, these structures have 3 main objectives:

- Protect and revitalize agriculture: this objective means to help young farmers to start their

activity, avoid the concentration on big farms, and develop a peri-urban agriculture and organic farming. The major tool used here is the preemption, which allows the SAFER to break a sale between to farmer to sell the land to another farmer. Small and young farmers have the priority in this system.

- Develop the vitality of the town and country planning policy: the SAFER has the ability to

- Protect the environment: this last objective is really wide, and goes from rehabilitation of

The result of this policy is easily presented in Figure 7. More liberal countries such as Denmark, the Netherlands or the United Kingdom have much bigger farms, with an average capital assets per farm respectively of 1 820 000 €, 1 700 000 € and 1 370 000 €. On the contrary, French farms have an average capital asset per farm below 400 000 €.

Some investments cannot be amortized in such small structures, like methanation units for milk farms or precision farming for field crops. For a methanation unit, which allows to reduce the impact on the environment by producing electricity, the total investment is around 1 200 000 € for 250 kWe (AREC, 2011). This type of unit is not very important in terms of productivity and can be 10 to 20 times bigger, but is already out of reach for a small farm.

1.3 A New Approach to Find

The CAP has created an artificial market for French farmers. As seen in Figure 8, the guaranteed wheat price (in blue) was well above the world market price (in red) before 1992. The price volatility of agricultural commodities has always existed, with a fall from 165 €/T in 1988 to 80 €/T at the end of 1990, but French farmers were protected from this high volatility.

Figure 8: Volatility of the price of wheat after application of Agenda 2000 Source: Offre et Demande Agricole

In fact, French farmers were affected by the world market price in 1995/96, when the prices reached 180 €/T, and in 2007 when all agricultural commodities prices soared... The price fall of 2008 and the new peak in 2011 and 2012 illustrates well that the volatility is not coming to an end.

Another element will modify the market for French farmers: the end of the quotas for milk and sugar beet producers in 2015 (Pinson, 2012; Hénin, 2011). Nowadays, the volume of production is limited for each European milk and sugar producers, in order to avoid surpluses and price falls. This suppression of the quotas will modify the milk and sugar market, and price volatility will surely increase ( terrenet.fr, 2012; Hénin, 2011).

Moreover, many other countries are more competitive than France for milk production (Rhein, 2009). The example of Germany is often cited in France because the price of milk is lower, farms are bigger, more productive and specialized (You, 2010; Waintrop, 2010). However, even Germany is considered as non-competitive in the world market (Rhein, 2009). In fact, both countries are declining in terms of market share for total agricultural products exportation (see Figure 9). However, the case of France is more complicated, as from the second world exporter in 1995 we lost market shares almost continuously because of a lack of competitiveness (Momagri, 2012).

Figure 9: Fluctuation of the share of agro-food exports by key countries in the global volume of agro-food exports

Finally, the CAP shall be reformed once more in 2014-2015: the decoupled payments should converge at the European average and more dependant to environmental conditions (Le Boulanger, 2011; Hénin, 2012). The effect will be important for French producers, as the decoupled payments will be reduced on average, particularly for grain producers. Moreover, these payments will be conditioned to environmental obligations which are not always respected so far.

Therefore, this is essential for farmers to analyze the source of their actual profitability if they want to face the challenge of a globalized agriculture combined with the reduction of subsidies! It is necessary to evaluate each projects and farms with more accuracy than before, because all indicators suggest that the market risk will increase.

2 Literature Review

2.1 Risk

2.1.1 Risk in Agriculture

The risks associated with the agricultural activities can be divided into 4 types of risks according to the OECD (Harwood, Heifner, Coble, Perry, & Somwaru, 1999; Holzman & Jorgensen, 2001):

- Market/price risks: these risks can affect directly a farm or a region through the variation of

- Production risks: hail or frost can affect directly a plot, but other climatic hazards can have a

strong impact on production. Technology evolution also can affect the production risks, by

controlling it (dams to prevent flood, anti-hail net, drainage, phytosanitary products...)

- Financial risks: some farmers have non agricultural revenues. A change in these revenues may affect the profitability. Farms are also exposed to financial risks through the interest rate on their loans.

- Institutional or legal risks: these risks are really hard to measure, as it goes from the

Overall, the risks in agriculture are really wide, and farmers cannot handle them all by themselves. Production risks for example cannot be covered only by insurance, because reinsurance companies refuse to cover the risk of insurance companies for this type of multi-risks insurance. The portfolio risks of these insurance is ten times higher than more conventional sectors (Chambers, 1989; Miranda & Glauber, 1997). Therefore, policy makers should integrate these elements into the CAP reform of 2014-2015.

On the other side, the market risks can be managed by the farmers themselves, the futures markets can be used by farmers to reduce their risk exposition (OECD, 2009). Actually, only 1% of all transactions made on the futures markets are effectively physically traded, but not many farmers use this tool (Rose, 2008). Even if the futures markets are not an efficient tool to predict the future physical prices, they remain efficient to reduce the risk exposure (OECD, 2009).

Regarding financial risks, it is harder to handle with agricultural policies or tools such as the future markets. However, farmers can with the help of their consultants estimate their financial exposition and reduce it by choosing fixed interest rates when they subscribe new loans. It is also possible to act on the optimal leverage level by choosing the appropriated financing decision for new projects or equipments.

Finally, institutional or legal risks are really hard to predict or reduce. For responsibility risk of course, it can be reduced through insurances and the choice of equipments in accordance with the legislation. However, institutional risks such as CAP reforms cannot be easily managed by farmers themselves.

2.1.2 Risk Premium

The equity risk premium, also called the risk premium, represents four different concepts according to the literature as stated by Fernandez (2010):

- Historical equity premium (HEP): it represents the difference between the historical results

of the stocks over government bonds or treasuries. This value really depends on the time-frame chosen. Damodoran (2011) recommends using a long time frame, as well as the French tax authority (2006).

- Expected equity premium (EEP): it is the expected difference between the same elements

than the HEP. However, in this case the historical results are not the base of the calculations. - Required equity premium (REP): this method is used by investors to estimate the

«incremental return of a diversified portfolio (the market) over the risk-free rate».

- Implied equity premium (IEP): it is in fact the REP assuming the market price of the stock

Out of the 150 textbooks studied by Fernandez (2010), 129 authors consider that REP has to be equal to EEP. This is the most common assumption as the CAPM states this assumption too, and 119 authors recommend using the CAPM (Fernandez, 2010). Out of those 129, 82 consider that EEP is also equal to HEP. We can consider that this position is the most commonly used, as 27 authors do not explain in their textbooks how they obtained their EEP (see Table 2). The overall recommended risk premium average is 6.7%. However, we can see than the difference between the min and max is really wide, ranging from 3.0% to 10.0%.

Table 2: Assumptions and recommendations of the 129 books that assume that REP = EEP. Source: Fernandez 2010

For our calculation, we will retain from this meta-analysis the minimum and maximum average for the most common assumption which is EEP = HEP. Therefore, the most probable range for the risk premium would be [5.5% : 6.7%] for the geometric mean and [7.0% : 8.5%] for the arithmetic mean. The geometric mean is presented by many authors as less dependent of the fluctuations and somehow more reliable. Fernandez itself recommends 3.8% to 4.3% for Europe and US.

According to the famous authors Brealey & Myers with Allen in their editions of Principles of Corporate Finance, they recommend to use REP between 5.0% and 8.5% (see Table 3). However, the major tendency is to use a REP between 6.0% and 8.0%.

Table 3: Equity premiums recommended and used in textbooks. Source: Fernandez 2010

What is more interesting is the reduction in the recommended risk premium in the textbooks over the last years. Figure 10 shows that this reduction was drastic between 1988 and 1993, and after 2002. Nowadays, the recommendation seems to be just below 6%. It is common to see in the literature some claims that these EEP are overvalued at 5.0 or 6.0%. Claus & Thomas (2001) states that after 1985 it is not realistic to take long period of centuries, and that the EEP is closer to 3.0% in reality. For France, for the period 1985-1998, they consider that EEP should be 2.6%, basing their calculation on the abnormal earnings approach. Quiry & Le Fur (2012) come to a similar conclusion even with a long time frame (111 years) with a 3.2% HEP.

Figure 10: Moving average (last 5 years) of the REP used or recommended in 150 finance and valuation textbooks. Source: Fernandez 2010

The French tax authority does not have a clear position on the subject (Direction Générale des Impots, 2006), however it gives the precision that someone could use the HEP of a long-run period: 5% over the last 100 year in France. For the French market, Damodoran (2011) and Salomons & Grootveld (2003) recommend to use 4.91%, using a 1976-2001 timeframe. The Credit Suisse in its Global Investment Returns Sourcebook of 2011 (Damodoran, 2011) recommends to use 6.0% for France (for a timeframe of 1900-2010, including the 2008 crisis for short-term Governments). Table 8 summarizes the different HEP obtained for the French market. Most of the observations are in the range between 3.8% and 5%.

The small-capitalization premium is subject to controversies in the early literature (Damodoran, 2011), however it seems that overall we can observe some significant premia for small or micro-capitalization. Figure 11 represents the premium observed for micro-capitalization compare to all-US stocks. We can easily see that during the 1980's and the major part of the 1990's the small firms did not perform more than the overall US stocks. However, since 1999 it seems that this small firm premium is starting again to be observed.

The other element we can learn from Figure 11 is that the variability is really strong. Therefore, it explains easily that we find really different positions in the literature, ranging from no premium to extreme values. For example Bergstrom, Frashure & Chisholm (1991) consider that for the French market this premium is 8.8% for a timeframe of 1955 to 1984. For the same time frame, they found only 3.0% for the German market.

Dimson & Marsh (2001) considers that the micro-cap geometric premium is 5.4% over all equities for a time-frame of 1955-1999 in UK (the EEP obtained is 6.2% for all equities for the same time frame). This premium for the US market seems to be lower according to Leonard (2009), with a premium ranging from 2.38% to 2.47% for US micro-cap stocks over the overall market (time frame: 1938- 2007). This micro-cap US premium has increased after the Dotcom crisis, reaching 5.63 to 5.83% for the period Oct-2002 to Sept-2007 (Leonard, 2009). Israelsen (2009) found 1.81% premium for small-cap compare to mid-cap and 2.8% for small-cap compare to large-cap for a 1980-2008 time frame. Switzer (2010) considers that this premium has even increased more when we consider a 2001-2010 time frame reaching 7.74% in the US. For the 1926-2010 time frame, this premium is 2.03% for the US. For Canada, this small-cap premium seems to be lower, ranging from 0.10% (1987-2010) to 4.12% (2001-2010). The most important finding of this work is that small-stocks have better premia during the periods of recoveries (after the dotcom crisis, the 2008 crisis, or the 1975 crisis). This is particularly essential today, as the overall economy is still recovering from the 2008 financial crisis.

Figure 11: Small firm premium (bottom 10% of market cap in US) between 1927-2010. Source: Damodoran 2010

Overall, the small and micro-cap premium observed ranges from 0.1% to 8.8%, depending on the market and the time frame studied. However, it seems that this premium is increasing since 2000, because the economy has experienced two important crisis since 2001. Moreover, it seems that the premium seems to be higher in France from what have been observed by Bergstrom, Franshure and Chisholm (1991). Table 5 presents a summary of our findings in the literature.

Authors		Time frame	Small and micro-cap premium
Bergstrom, Frashure	&	1955-1984, France	8.8 %
Chisholm (1991)
Israelsen (2009)		1980-2008 US	1.8% to 2.8%
Leonard (2009)		2001-2007 US	5.63% to 5.83%
Leonard (2009)		1938-2007 US	2.38% to 2.47%
Switzer		2001-2010 Canada	4.12%
Switzer		1926-2010 Canada	0.10%
Switzer		2001-2010 US	7.74%
Switzer		1926-2010 US	2.03%
Mean			3.97%

For the recent time frame, which is the most appropriate time frame considering the fact that the economy is in recovery since the major crisis of 2008, authors considers that the HEP is ranging from 4.12% to 7.74% (Table 5, Switzer, 2010). Moreover, this choice is supported by the fact that many analysts agree on the fact that the soft commodities prices should remain at a higher price in the coming years (OECD-FAO, 2010). The expected market return should increase in the following years.

2.2 The WACC Theory

2.2.1 The Importance of the WACC Theory.

The WACC is not considered easy to compute, but is one of the fundamental elements in modern finance (Quiry & Le Fur, 2012; Brealey, Myers, & Allen, 2008). The NPV calculation and the determination of the value of the stock are based on the results of the WACC, which underlines its importance (Quiry & Le Fur, 2012, p. 699). The underlying theory of the WACC is the CAPM. This model objective is to measures capital assets value depending on the risk and the expected return of the company. According to Brealey, Myers & Allen (2008), the after-tax WACC represents more the reality of the company and can be described as follow:

The cost of equity in the WACC theory has to be calculated using the CAPM, which can be described as follow:

E(Re): expected return on equity. It corresponds to K(E) in the WACC formula [3: beta. It represents the risk.

E(Rm): expected return of the market in which the company operates. See part 2.1.2 page 20 on risk premium for the literature review dealing with risk premium. However, the risk premium is defined as the premium obtained compare to a risk-free asset.

Rf: risk free rate of return. The risk-free rate of return can be estimated by the long-term government bond yield (Brealey, Myers, & Allen, 2008; Quiry & Le Fur, 2012).

2.2.2 The Betas for Micro-Capitalizations

In this research, financial techniques are used to estimate the risk for listed companies to apply it for non-listed small companies. The bias introduced is consequent, as the beta has a negative relationship with the size of firms (Binder, 1992; Al-Rjoub, Varela, & Hassan, 2005; Shomir, Pat, & Jeong-gil, 2011). Therefore, small farms should have a higher Beta compared to listed companies operating in agriculture, as smaller firms present little or no diversification (Drew & Veeraraghavan, 2003). Table 6 presents the results found in the literature comparing large capitalizations and micro-capitalizations, which can be considered as the right comparison between the listed companies operating in agriculture which have activities in many countries and the small farms of Isère.

Authors	Time frame	Number of groups	micro-cap Beta premium
Al-Rjoub, Varela & Hassan (2005)	1970-2000	1^st decile vs 10^th d		0.51
Al-Rjoub, Varela & Hassan (2005)	1982-2000	1^st decile vs 10^th d		0.36
Al-Rjoub, Varela & Hassan (2005)	1990-2000	1^st decile vs 10^th d		0.24
Shomir, Pat & Jeong-gil (2011)	1980-2003	1^st quartile vs 4^th q		0.21
Bhardwaj & Brooks (1993)	1926-1988	1^st group vs 5^th g		0.72
Chan & Chen (1988)	1949-1983	1^st group vs 20^th g		0.69
Dongcheol (1993)	1926-1990	1^st decile vs 10^th d		0.58
Jegadeesh (1992)	1954-1989	1^st group vs 20^th g	0.32	- 0.62
Mean				0.47

As presented in Table 6, the effect of size is quite important on the beta. It is important to notice that the standard deviation also increases with the beta (Drew & Veeraraghavan, 2003; Dongcheol, 1993; Chan & Chen, 1988). From the results found in the literature, for micro-capitalizations compare to large capitalizations, the beta increases on average by 0.47, which is really significant regarding the impact of the beta on the WACC formula.

2.3 Discount Rates and the CAPM Used in Agriculture

The CAPM and the actualization rates have already been used in agriculture long ago. The Capital Asset Pricing Model has been tested for example to test the farm real estate market by Barry P. J. in the 1980's (Barry P. J., 1980a). His conclusions were that the low betas (0.19 for the US) observed for farm real estate made this investment valuable as a diversification tool for portfolios and that the required rate of return for the US farmland was 8.76%. Other studies have been done on the same subject to test the interest for investors to diversify into farmland (Hanson & Myers, 1995). It seems that these conclusions remain true nowadays, as the foreign investments in farmland increased strongly after the rise of the grain market price in 2007.

Since then, the diversification into agricultural commodities have been tested either. The results are interesting, because thanks to negative correlation with other markets, this diversification reduces the risk and are considered as profitable for investors (Sminou, 2010). The opposite have been tested also, to see if farmers who own their lands have an interest to diversify their investments with traditional financial tools (Nartea & Webster, 2008). According to their results for New Zealand after the deregulation of the market, the expected return of the farmers would increase thanks to diversification, even if the expected return on farmland due to capital gains (9.83%, 1988-2003) were higher than the expected returns of ordinary shares (5.59%, for the same period). This potential gain comes from the good diversification of the portfolio which reduces the risk but not the overall return according to the authors.

On the other side, some studies have been done in agriculture using the NPV method. Most of them
do not explain clearly how they chose their discount factors. For example, investments in peach
orchards in India have been tested with three discount rates proposed by NGOs: 6, 9 and 12% to

compare the NPV method with the amortization method (Gangwar, Singh, & Mandal, 2008). Finally the authors retained the 12% discount factor.

Barry P. J., at the beginning of the 1980's, has done a research to evaluate the impacts of government support price programs on the financial structure of the farms of US (Barry P. J., 1980b). In his calculation, he used three pre-tax discount factors of 10%, 12% and 14%, corresponding to an after tax discount factor of 7.8%, 8.16% and 7.0% respectively. However, here the author selected the three pre-tax discount factors arbitrarily.

More recently, Johnson M. R. (Johnson, 2002) presented its method to calculate the capitalization rate for Kansas in a case study. He starts first with the loan rate offered to farmers, and then he adds a discretionary increase based on the risk premium and the tax effect. He obtains a capitalization rate of 14.71% in 2000 (with a loan rate level of 8.94% and a corporate tax rate at 30%). Using his method we would obtain around 10.5% of capitalization rate for France (with a loan rate level of 3.5% and a corporate tax rate at 33%). The discount factors found in the literature are therefore ranging from 7.0% for the lowest, to 14.71% for the most recent study found from the US. However, it would be hard to apply these methods to the French context, because most of these discount factors are somehow based on discretionary risk premiums or they are chosen totally arbitrarily.

3 Research Methodology

One of the main objectives of this research is to give new decision tools for farmers' consultants to evaluate the expected profitability of their customers' projects or farms. In this section, the preparation of this research will be presented.

3.1 Survey Construction

A small survey designed to collect different perspectives from accounting and consulting companies, mainly from the network CEFRANCE:

- First, the network of the CERFRANCE represents more than 50% of the market of farming consulting. The remaining market shares are shared by many different organizations with no comparable size to the leader.

- Then, the methodologies used in the different CERFRANCE are based on the same recommendations from the head of the network (the CNCER). Therefore, we can obtain a clear view of the methodologies used by consultants without a large survey.

3.1.1 Questionnaire

- Confirm of reject my initial observation which stands that farm consultants use discount rates based on loan rates plus a risk premium around 2%.

- Test the willingness of farm consultants to use the WACC to approximate the cost of capital of their customers in their region.

To test these hypotheses, an internet-based survey of 9 questions was sent to 25 consultants in France. In order to improve the answer rate, the questionnaire was sent only to farm consultants who were linked to me in a professional social network (Viadéo^® or Linked In^®). As these consultants are all French, it has been written in French (see Appendix 3).

This choice of consultants introduces a first bias: mostly young professionals use professional social networks. However, as young consultants are less experimented, it is common sense to believe that they are taught with the last methods developed or chosen by their consulting companies. Therefore, this choice of population was motivated by its efficiency to collect up to date information at the minimal cost.

3.1.2 Results Analysis

As the number of participants was reduced, analysis of the results was only descriptive (percentages, average and medians) and no statistical tests were performed. However, these results were sufficient to be conclusive.

3.2 Historical Analysis of the Results of Farms in Isère

3.2.1 Data Collection

The data used in this part of the research have been extracted from the database of the CERFRANCE
Isère. This database has 5 years of results available, for 2 000 farms operating in Isère. CERFRANCE
Isère is the leading accounting company in the area, serving more than 50% of the 3 999 professional

farms of the department (Agreste, 2011). Therefore, we can consider the results extracted from this sample as representative of the population of farmers of Isère. This is why the inclusion criteria for the extractions were as wide as possible:

- All companies which are taxed at the «bénéfice réel», which means that they are obligated to produce a balance sheet, a profit and loss statement and a cash flow statement to the tax administration. All the companies with a turnover above 76 300 € are concerned by this obligation. This inclusion criterion is obvious because without books accounts profitability or leverage ratios cannot be calculated.

- All companies that are taxed under the «Bénéfice agricole» regime, which means that more than 70% of their turnover comes from agricultural activities. This inclusion criteria has been chosen to minimize the impact of forestry, environmental services and other activities practiced by some farmers.

The software used at the CERFRANCE Isère (Panda) cannot produce extractions for several years at one time neither an extraction with all the data about a year in one file. Therefore, two extractions were made for each year:

- One excel sheet with all the balance sheet information available (146 data per farm, from

surface, number of labor units to asset valuation. See Appendix 1 for illustration),

- A second excel sheet with all profit & loss and cash flow statement information (143 data per farm. See Appendix 2 for illustration)

As presented in Table 7, the number of companies available is not constant. This effect is due to the information system structure: depending on the way the accounting books are send to the tax administration, the results of the farm are automatically or manually uploaded into the database. The only exclusion criterion was to keep a constant sample over the five years. Therefore all farms which did not have results for 1 or more year were excluded. This exclusion criterion reduced the population of observed farms to 620.

The 620 farms have been classified into 7 specializations, based on the production most observed in the department of Isère (see Figure 12). A farm generating 2/3 of its turnover from one main production was considered as specialized in this production. Cattle ranching corresponds to the beef production (sheep or other ruminants were considered as diversified production as they do not represent enough farms to constitute a group). Walnut productions, vegetables and fruit growing

were not analyzed as the number of farms observed was too low to make groups based on their financial leverage.

3.2.2 Data Processing

The raw data extracted from the database were to be analyzed. First, the self-employed labor unit costs are not always taken into accounts into the books, due to accounting regulations. Therefore, the data have been transformed to include a labor cost for each labor unit at the SMIC (minimum salary level in France for a FTE, Table 8):

Another element had to be transformed in order to have reliable results: social security contributions. This cost is tax deductible, but depending on the legal form of the company it can be treated outside of the balance sheet and profit & loss statements. The calculation of these contributions depends on many factors, such as the year of installation of the farmer, his average profits for the last 3 years (an option exists to calculate it only with last year results). An important element is that only farmers who operate in limited liability legal form have the possibility to treat these contributions off-balance sheet. Their results are most of the time significantly higher, justifying that the added contributions should be on average higher than the contributions in books. The contributions were calculated based on the results of the annual accounting results (in reality, the years' results are taxed over 3 years: N-1, N-2 and N-3 for most farms). This recalculation introduced a bias in this research, but the accounting principles of the company have been adapted to avoid this bias for future researches. For 2012 and after, all social contributions will be recorded in the database.

Years	# of companies without social security contribution in books		Average contribution added			Average contribution in books		Average contribution for the population
2006	77	(12%)	7	061	€	7	193 €	7	176 €
2007	110	(18%)	11	871	€	7	009 €	7	871 €
2008	139	(22%)	12	633	€	7	371 €	8	551 €
2009	153	(25%)	10	124	€	8	230€	8	697 €
2010	162	(26%)	6	539	€	7	488 €	7	241 €

3.2.3 Research Questions

The main goal of this analysis is to test the traditionalist approach which states that an optimal level of debt helps to reduce the WACC (Brealey, Myers, & Allen, 2008, p. 485). If this theory holds in agriculture, a positive link between the debt level and profitability or productivity should be observed. Therefore, the research question can be stated with two hypotheses.

H0: ROA and ROE tend to increase with the level of debt of the farms in Isère.

H1: ROA and ROE have no positive links with the level of debt of the farms in Isère

H0: Financial distress may occur when leverage increases, reducing ROA or ROE. H1: No signs of financial distress are clearly observed when the leverage increases.

The second hypothesis is common sense. However it has to be tested for agriculture. Milk production was really stable for years, because the market was controlled thanks to the system of quotas. Therefore, as the volume of production is somehow easy to control, financial performances of the farms are easy to predict, limiting strongly the risks of financial distress.

3.2.4 Data Analysis

The WACC theory assumes that there is an optimal debt level to reduce the cost of capital of the company and then increase its overall performance (see Figure 1 page 12). Therefore, ROE and ROA were analyzed by groups of leverage in order to verify if this theory holds in agriculture. The groups more leveraged should have better results in terms of ROE and ROA than low leverage groups. We should also observe more variability and/or lower results for groups with really high leverages like group 1 or 2 (Table 10). This observation would be considered as financial distress.

As presented in part 3.2.2, the data had to be processed to be exploited. The ROE, ROA and leverage also have to be calculated considering the specificity of the accounting standards used at the CEFRANCE Isère. Therefore, ROE and ROA were calculated as follow:

- Wc: cost of labor (each FTE self-employed unit of labor received a cost of the SMIC),

- E: shareholders equity. As no financial market exists to estimate a market value of farm's
equity in agriculture, the book value had to be retained,

- Pca: partners current accounts. These amounts are considered as debt from an accounting perspective. However, it is not considered by convention as a debt but as shareholder's equity from a consultant perspective because these «partners» are the shareholders,

- D: total debts. The value retained was book value, as the database do not includes

information allowing a recalculation at market value. However, the interest rate was analyzed at the market value,

In order to compare the groups, a set of statistical tests were performed with the software Statgraphic Centurion:

- Shapiro Wilk test and Kolmogorov test for the normality of the datasets. Density trace and histograms were also used for sample sizes bigger than 2 000 farms (Shapiro Wilk test cannot be performed for this size of samples). Normality was tested before each test.

- One-way ANOVA and multiple range comparison to compare the means of the different groups when the assumptions of normality and homogeneity of variance were verified.

- When the assumption of normality or homogeneity of variance was rejected, a non parametric test was used (Kruskal-Wallis). As normality and equivalence of the variance are the base-assumptions of the ANOVA, the means cannot be compared if these two hypotheses are not verified. The Kruskal-Wallis test compares the rank of each observation to overcome this limitation, and tests if the medians are significantly different or not.

- Mood's and Median test to analyses the medians. It is a non-parametric test which compares the distribution of each sample around the overall median of all the groups. Therefore, this test can estimate if the medians are significantly different.

First, the effect of time was tested. As time has an effect on ROE or ROA, separate tests were performed for each year in order to isolate the time effect and analyze the effect on leverage more accurately. Then, the effect of specialization was tested. As specialization has an effect on ROE or ROA, separate tests were performed for each specialization to improve the robustness of the results. The risk error (alpha) represents the risk to reject the null hypothesis tested when it should be

considered actually true. Because 10% is an acceptable level of risk error in finance, this alpha was used for all statistical tests in this research.

3.3 Bond Yield Plus Risk Premium Model

The first method to estimate the capitalization rate was the method used by the French tax authority. Farming consultants have to keep in mind that their valuation methods can be used as the base of the calculation for a capital gain on the sale of a company. The tax authority therefore has the right to modify the tax base if it considers that the results do not represent a fair value at the time of the transaction. The most common tax adjustment in those cases is the hidden donation, when the tax authority considers that the company or its shares (not publicly tradable in the case of most farms) are undervalued. This is why it is necessary to include the methods used by the administration in this study. In reality, this method is based on the CAPM, as the risk premium is the difference with the total expected return minus the risk-free return. The actualization rate formula can be presented in this way:

As presented by Sharpe (1964) it was assumed that everyone can borrow or lend its funds at a «pure rate of interest». This risk-free rate is considered by the tax authority as the government bonds rate (Direction Générale des Impots, 2006), and was assumed in this paper as equivalent to the French government debt reference rate, given by the French central bank (a 12 month average of the 10 years bonds monthly average was used).

3.3.2 Beta and Risk premium

The risk free rate has to be increased by a risk premium, justified by the necessary supplementary yield that a risky asset needs to deliver compare to a pure interest rate. The tax authority recommends using the historical risk premium observed in the French economy, which can be increased to consider the size of the company studied. For the calculations, the references of the tax authority were used, and an increase to consider that we are dealing with micro-capitalization.

The first bias that arises is easy to consider, as stated by Damodoran (2011), would be to include 2008 stock results in our time-frame: the major crash observed this year would diminish the results significantly, leading investors to think that the risk would be lower after the 2008 crisis than before. However, from a pragmatic point of view, this could be considered as logical: there is less risk of a major historical crisis after it happens than before, like there are less risks of a centenary flood after ones happen.

The Beta is used to consider the risk relative to the company and the sector in which it operates (Direction Générale des Impots, 2006). Once again, the tax authority produced some recommendation and we will base our calculation on it. We will compare this beta with the beta observed by another author for agriculture in France.

3.4 WACC Estimation

Regarding the WACC estimation, 2 types of data will be used: primary and secondary data. For primary data, we will use:

- Results of the farms from Isère extracted from the database of the CERFRANCE Isère to calculate the weighted average leverage by specialization. The time frame of the dataset is 2006-2010, with 620 farms. The characteristics of this dataset are presented in detail in part 3.2.

- Sample collection of 46 recent loan's rate for all specialization in the database of the

- The market capitalization and the beta calculated by Bloomberg for all the companies listed

in North America and Western Europe for the following activities: Oilseed farming, grain farming, cattle ranching and farming, poultry, fruits & vegetables. Those activities have been chosen because they are really close to the specialization of the farms of Isère.

- Risk-free rate of return: the 10 year government bonds yield calculated by the French central bank will be used. The 12 month average of this rate will be used. This choice of long term maturity was determined by the timeframe of the investments of French farms. Even in grain production, investments are made for 7 years on average, and the rotation timeframe for the crops is 3 to 4 years. For dairy farms, constructions are amortized over 20 years on average.

The extraction from Bloomberg helped to identify 41 companies specialized in agriculture in Western Europe and US (see Table 11). The grain market is globalized today, and the meat market is really influenced by world prices either. These statements are not so true for the dairy market, because the cost of refrigerated transportation is quite high regarding the price of milk. The European market is a more representative market for dairy production therefore.

The beta of each firm will be weighted averaged with the market capitalization, in order to obtain a beta for each specialization and each region. However, for grain farming we will retain the beta of oilseed and grain farming for Western Europe and the US together. The grain market is so interconnected between Europe and US that the market-risk is really close. For the grain farming in Isère, we will retain the beta of oilseed farming and grain farming, because oilseed production is considered as grain production in the French classification. For dairy and cattle farming, we will retain the beta of cattle ranching and farming of Western Europe. For diversified production and the average farm of Isère, we will retain the average beta of the 41 companies extracted from the Bloomberg database.

Table 11: Number of companies studied for the Beta estimation. Source: Bloomberg

The last element of the WACC formula is the corporate tax rate. This element is really hard to determine in agriculture, because farmers have the choice between the corporate tax rate at 15% (up to 38 120 € of net income and 33.33% after) and the personal income tax with its progressive tax rate from 0% to 41%, depending on the total income per capita of the household. Moreover, the net income is also subject to social tax, at an average rate of 32.1%. Therefore, to simplify our calculation we will assume a corporate tax rate at 33.33%, which is the usual corporate tax rate in France (only companies held at 75% by persons have the benefit of the 15% tax rate up to 38 120 €). This rate is chosen in most studies about profitability for other industries in France. Therefore, it will simplify comparisons. As the French farms are small companies, it is not obvious that this tax rate could increase in the coming years. The actual government considers that SEM's are too much taxed compare to the blue chips.

4 Data Analysis

4.1 Results of the Survey

Fourteen farm consultants answered to the survey, from almost all regions of France (Figure 13). Therefore, the results are not influenced by only one CERFRANCE.

Regarding the age and the gender, the major part of the respondents are male consultants aged between 26 and 35 years old as we can see in Table 12.

Out of the 14 respondents, 13 said that they realized project business plans for their customers. However, only 79% said that they used discount factors to evaluate the value of their customers' companies in question 5. Less surprising are the answers to question 6, where 89% answered that they used the NPV method to estimate the profitability of a project (see Table 13).

Regarding question 7, only 9 out of the 14 respondents answered to the question (Table 14). It seems that the most popular method among the farm consultants is to choose a discount factor based on the experience, and loan's rate plus risk premium is the second most popular method.

Table 14: Answers of question 7: «how do you choose the discount factors you use?»

Regarding the most important question of the survey, the question 8, not all consultants are specialized in all the fields concerned by the question. This explains why the number of answers is reduced (9 consultants gave at least one answer for this question). However, the answers tend to confirm that almost all farm consultants use really low discount factors (see Table 15). The highest answer is 15% for grain producers. However, even this answer is surprising, because this consultant uses a 5% discount factor for cattle ranching, which seems to be a difference too important. There are merely no reasons for such a difference in terms of risk premium between the two different productions, and it is probable that this answer is only a typing error.

Table 15: Answers of question8 «what discount factor do you use usually for your customers? (ex given: 11%)»

Another element is surprising: means and medians are really low for all production, with discount factors equivalent to the French 10 years Government bond yield. This rate was recently at its record low at 2.422% (Dobson, 2012). The risk premiums used by some practitioners are actually lower than the 2% used at the CERFRANCE Isère. The low number of bankruptcy observed in the sector (1.6%o in 2011 for Rhone Alpes) may explain that practitioners using rates based on experience use really low risk premium. Most of the consultants do not use a loan's rate plus 2% risk premium, as this method would give results closer to a 5 to 6% interval (regarding the loan's rate observed in the sector).

Table 16: Answers of question 9: «If you were proposed a tool to estimate the WACC of the farms of your region to use it as a discount factor, how would you consider this tool?»

Finally, question 9 (Table 16) implicitly confirmed that consultants are not really secured with the tools they use. 78% of the respondents considered that a tool to estimate the WACC of the farms of their region would be necessary to improve the discernment of their recommendation, and the remaining 22% considered it as useful.

- Consultants use the loan rate plus risk premium model. The risk premiums used seem to be lower than 2%,

- Most of the consultants would find necessary to have new tools to improve their methods.

4.2 Calculation of the Historical Results Using the Data from CERFRANCE Isère

The objective of these calculations is to see if the hypothesis of the traditionalist's approach is verified in agriculture: does the performance of the firm increases with leverage? Therefore, the results of firms divided in 5 groups of leverage were analyzed (group 1 high leverages, group 5 lowest leverages. See section 3.2.4 page 31 for details).

4.2.1 Normality of the Datasets

Figure 14 presents the distribution of values for ROE to compare it with a normal distribution. The distribution does not fit with the normal distribution line looking at the histogram, because the central values have an abnormally high frequency. This impression is confirmed by the KolmogorovSmirnov test, because the P-Value is lower than 0.000. We can reject with 99% confidence that ROE comes from a normal distribution. The Shapiro-Wilk test cannot be performed because the sample is bigger than 2000 values.

The distributions of ROE have been tested for each year's separately, and within each year's for each
production either. Over the 25 tests, ROE could come from a normal distribution only for dairy

producers and cattle producers in 2010. Therefore, only the Kruskal-Wallis and Mood's Median tests were used to analyze ROE, and the results of the Anova's were not presented.

Figure 15 presents a histogram of the frequency of the distribution for ROA to compare it with a normal distribution. From a graphic perspective, it seems that ROA is closer than ROE to a normal distribution, as the normal distribution line is closer to the frequency histogram. However, the Kolmogorov-Smirnov test P-Value was lower than 0.000 as well, therefore we can also reject at 99% confidence that ROA comes from a normal distribution.

Same as the ROE, the distributions of ROA have been tested for each year and within each year for each production separately. All the tests performed reported the same results: ROA does not come from a normal distribution at 99% confidence. Therefore, only the results of the Kruskal-Wallis and Mood's Median tests were discussed for ROA.

4.2.2 ROE

4.2.2.1 ROE Tested by Time and Specialization

The dataset have been tested first to see if time had an effect on ROE. The results of the Mood's
median and Kruskal-Wallis tests were quite logical: as someone could have predicted, time has a
significant effect on the ROE, with really low P-Values. The years 2006 and 2009 were two years of

bad results for farmers in Isère. The price of agricultural commodities soared in 2007 and 2008,
explaining that these two years were better. The effect was similar in 2010, when commodity prices

YEAR	Sample Size	n<=	n>	Median	90,0% lower CL	90,0% upper CL
2006	568	326	242	0,0195117	0,00305438	0,041546
2007	568	245	323 5	0,0910952 55	0,0759334	0,110756
2008	568	250	318	0,0837851	0,065602	0,0957051
2009	568	328	240	0,0176906	0,00193535	0,0337506
2010	568	271	297	0,0734001	0,0512998	0,091961

Test statistic = 46,0986 P-Value = 2,3492E-9 Kruskal-Wallis Test for ROE by YEAR

Figure 16: Box and whisker plots with median notch for ROE by year (means are shown with a red cross)

Figure 16 presents the box and whisker plots with median notch at 90% confidence intervals, showing that the results of the years 2006 and 2009 were significantly lower than in 2007, 2008 or 2010. Therefore, in order to isolate the time effect, ROE has to be analyzed year by year.

Figure 17: Box and whisker plots with median notch for ROE by specialization (means are shown with a red cross)

The Mood's median test for ROE by specialization illustrates that the specialization has a significant impact on the ROE, with a really low P-Value illustrating that we cannot reject the hypothesis that the medians are different at 99% confidence. The cattle production has a median ROE lower than the other productions. The P-Value for the Kruskal-Wallis test is 3,96451E-8, giving the same results. Therefore, ROE was tested separately for each specialization and each year, in order to isolate the effect of time and specialization.

4.2.2.2 ROE for Dairy Producers

The group of dairy producers represents 141 different farms studied over the 5 years. The Figure 18 presents the results of the Mood's median tests performed for the ROE of the different groups. First, we can see that the groups 4 and 5, which are the lowest leverages, have the lowest medians. Then, the group 1 shows sign of financial distress: it has the best results in 2007 with a ROE higher than 15%, but is close to be the worst group in 2006 with a ROE close to 0,0%. Groups 2 and 3 seem to be more stable and closer in terms of results, with an advantage for group 2 which is higher than group 3 (except in 2006, a bad year for agriculture in Isère as already presented, explaining that financial distress may have occurs for the most leveraged farms).

Figure 18: Median ROE for each year and each group of leverage for dairy * means significantly higher than **

In 2007, group 1, 2 and 3 were significantly better than group 5. In 2009, it was group 2 and 3 which were significantly better than groups 4 and 5. In 2010, only the group 2 over performed the group 4. For all the other years, there is no statistical difference between groups, and the differences between the medians cannot be interpreted due to high standard deviation (see Appendix 8 for details page 89).

4.2.2.3 ROE for Grain Producers

The group of grain producers represents 126 farms of Isère. The sample is constant over the 5 years.
The Figure 19 represents the ROE for each year and each group. First, group 1 over performed in
2006, 2009 and 2010. This is quite surprising, because 2009 wasn't a good year in terms of price of

grains, and we could have expected signs of financial distress. However, except in 2010, there is no statistical difference showing that group 1 is better than any other groups. This could be surprising regarding the major difference seen in the histograms, but standard deviations are really high for group 1, with a lot of extreme values (see Appendix 10 page 99 for standard deviations).

Figure 19 : Median ROE for each year and each group of leverage for grain * means significantly higher than **

In 2007, the group 2 has significantly better results than the group 5. For 2010, it is group 1 which has better results than groups 4 and 5. For 2006, 2008 and 2009, no statistical conclusion can be made because the Mood's Median and Kruskal-Wallis tests did not reveal any significant results, with P-Values higher than 10% (see Appendix 10 page 99).

4.2.2.4 ROE for Cattle Ranching

For the group of cattle producers, only 45 specialized producers were represented. As the sample is really small, the differences between the groups are in most cases not significant due to the limited number of values in each group of leverage (see Appendix 9 page 95). In figure 18 group 2 seems to over perform the other groups. Years 2007, 2008 and 2010 seem to illustrate that performance increases with leverage, up to a limit at 80% where financial distress may occurs. However, Group 2 is significantly higher than group 5 only in 2008. We cannot conclude anything else from the results of the statistical tests.

Figure 20: Median ROE for each year and each group of leverage for cattle * means significantly higher than **

4.2.2.5 ROE for diversified production

This group is the biggest in terms of number of values, with 256 farms analyzed for 5 years. However, this is a really heterogeneous group, with farms that cannot be compared between each others. Therefore the diversified group was not analyzed in details because the results would not be useful in terms of managerial implications.

4.2.2.6 Results Summary for ROE by Groups of Leverage

Table 17 presents a summary of the results of the statistical tests. The group 2 showed the best results in terms of ROE, followed by group 3. Groups 4 and 5 on the contrary have really often significantly lower results than groups 2, 3 and in one case group 1.

Table 17: Results of the Mood's median tests summary for ROE by groups of leverage for each years

Table 18 presents the average ROE for all specializations and all years. Diversified and dairy farming have really close results regarding their overall means. However, it seems that diversified production is much more volatile, moving from 13.9% ROE on average in 2007 to 4.0% in 2009. For milk, the range starts from 5.4% in 2006 and goes to 11.7% in 2008. It seems that this specialization is less risky than diversified production or grain farming. Cattle's farming, on the contrary, has always the lowest results except in 2010. It seems that the economical performances of these farms are

chronically insufficient. 2010 is an exception, with the rise of the price of meat linked with the new export opportunities to Turkey. This improvement of the market conditions for cattle farming still has to be confirmed, because it depends so far on a customer representing more than 50% of the exports: if Turkey stops its importations as it did in the past, prices could fall sharply.

*Specialization*	*2006*	*2007*	*2008*	*2009*	*2010*	*Mean*
Dairy	5,4%	9,3%	11,7%	7,8%	5,6%	8,4%
Cattle	-5,4%	3,5%	3,6%	1,5%	13,7%	2,2%
Grain	2,0%	21,2%	18,6%	2,7%	7,7%	11,3%
Diversified	4,0%	8,7%	13,9%	4,3%	12,3%	8,3%
Mean	3,2%	10,3%	13,0%	4,8%	10,1%	8,2%

However, a really different outcome can be observed looking at Table 19, which presents the ROE by specialization and by groups of leverage (all years are considered in this table). The cattle specialization, which obtains low results every year except in 2010, is profitable when the leverage is higher than 40%. On average, the farms of Isère are economically more profitable when the leverage increases, particularly for milk and cattle. We can observe a fall in the results for group 1, particularly for dairy farms, which can be explained by the financial distress that can occurs when leverage is too high.

*Specialization*	*Group 1*	*Group 2*	*Group 3*	*Group 4*	*Group 5*	*Mean*
Dairy	-2,4%	11,7%	9,3%	6,3%	3,7%	8,4%
Cattle	1,2%	7,3%	7,4%	1,3%	-4,8%	2,2%
Grain	9,4%	12,5%	11,5%	12,0%	10,5%	11,3%
Diversified	5,7%	12,4%	11,1%	4,4%	5,4%	8,3%
Mean	4,6%	11,8%	10,2%	5,9%	4,8%	8,2%

These results are totally consistent with the initial hypothesis which states that there is an optimal level of debt for farming. From the ROE results, group 2 and group 3 obtain the best results, which is an indication that the optimal leverage lies between 40% and 80%. Group 3 has similar results than group 2, except for dairy production. Therefore, the optimal leverage must be higher for this specialization.

- Financial distress may occur for all production for leverage higher than 80%.

4.2.3 ROA

4.2.3.1 ROA Tested by Time and Specialization

We applied the same tests for ROA than for ROE. The time effect was first tested as well as
specialization to see if it was necessary to isolate their effects on the ROA. The results were very
similar than for the ROE, which is really logical as ROE and ROA have the same numerator. Again, the

box and whisker plot in Figure 21 illustrates that year 2006 and 2009 were bad in terms of results for farming in Isère with ROA median close to 1%, which is lower than the 1.5% inflation in 2006 (Le Monde.fr, 2007).

Test statistic = 73,6591 P-Value = 0 Mood's Median Test for ROA by YEAR Total n = 2840

YEAR	Sample Size	n<=	n>	Median	90,0% lower CL	90,0% upper CL
2006	568	339	229	0,00945275	-0,000555481	0,0218933
2007	568	242	326	0,0570648	0,0460463	0,0689174
2008	568	248	320	0,0528593	0,0436125	0,0621654
2009	568	327	241	0,0121847	0,000872383	0,0184358
2010	568	264	304	0,044225	0,0338877	0,0561194

Figure 21 : Box and whisker plots with median notch for ROA by years (means are shown with a red cross)

Regarding the effect of specialization, once again the P-Values were really low, illustrating that we cannot reject the hypothesis that medians are significantly different at 99% confidence. The two tables of the Kruskal-Wallis and Mood's median tests illustrate that cattle's ranching presented significantly lower results than all other production in Isère.

Therefore, we had to isolate the effect of specialization also on ROA. The tests were performed for each specialization and each year separately.

4.2.3.2 ROA for Dairy Producers

The farms analyzed here are the same than the ones for the ROE for dairy producers: 141 dairy farms of Isère. Figure 22 presents the results of the different tests performed to compare the median ROA for dairy farms between 2006 and 2010. The results are close to what can be observed for the ROE: groups 2 and 3 seem to over-perform the other groups.

Figure 22 : ROA for each year and each group of leverage for dairy producers * means significantly higher than **

For year 2007, the group 2 and 3 are significantly better than the group 5. For 2009 and 2010, group

2 is significantly higher than group 4 and 5. Group 1 median on the contrary is lower than group 2 or

3 for all years, and has the lowest median of all groups in 2006, 2008 and 2009. We cannot affirm that this difference is significant as the standard deviations are really high for group 1 (Appendix 11 page 103). However, these elements are clear signs of financial distress.

4.2.3.3 ROA for Grain Producers

Figure 23 represents the ROA medians for each group of leverage for the grain producers. For this specialization, the results are not really similar with what we can observe for the ROE. First of all, group 2 is never statistically higher than any other groups, and is even the lowest performance in 2008 and 2009 (without a statistical difference however). Only 2006 and 2010 tend to be consistent with the theory which states that performance increases with leverage. In 2010 group 1 was

significantly higher than group 4. However, there was no significant difference with group 5 even if its median was lower, because of the higher standard deviation in group 5 (see appendix Appendix 13 page 111).

Figure 23: ROA for each year and each group of leverage for grain producers * means significantly higher than **

The only statistical difference between groups tends to prove that leverage increases ROA, but only one year out of five. Therefore, it will be hard to make conclusions on these results for grain producers, except that it seems that the optimal leverage range is 40-60% for this specialization.

4.2.3.4 ROA for Cattle Farming

The results for the ROA are consistent with what have been observed for the ROE for cattle farming. Group 2 has a higher median than group 4 and 5 in 2008, and performance seems to increases with the leverage (see Figure 24). Once again, group 1 do not present stable results over the year, which could be the consequence of financial distress for those farms which are more leveraged than 80%. However, it will be hard to make more interpretation from these results as only 45 farms are studied for this specialization, and significant differences are observed only in 2008.

Figure 24: ROA for each year and each group of leverage for cattle farming * means significantly higher than **

4.2.3.5 Results Summary for ROA by Groups of Leverage

The median of group 2 is usually significantly higher than the medians of groups 4 and 5 (Table 20). Group 3 in two cases over-performs group 4 or 5, and group 1 is better than group 4 in one case. The same elements are observed for the ROA than for the ROE: the performance increases with the leverage, even if the differences between groups are not always significant. The high heterogeneity intragroup and between groups, illustrated by the high standard deviations (see Appendix 11, Appendix 12 and Appendix 13 page 103, 107 and 111) explains why the differences cannot be considered as significant even if it could have been anticipated so only by looking at the means and medians of the different groups.

Table 20: results of the Mood's median tests summary for ROA by groups of leverage for each years

Table 21 presents the average ROA for all specializations and all years for the studied farms. The first point that can be noticed is that the average ROA is 3.1% for the average farm in Isère. The ROA reaches 13% for grain farms in 2007, but fall to 1.7% in 2009 for the same specialization. It is interesting to compare the dairy and the grain production: both have a close weighted average for ROA, but variability is really higher for grain than for dairy farming (standard deviation for dairy is 1.7%, but reaches 5.6% for Grain). Grain farming can be considered as a riskier specialization. In 2008, the average farm in Isère reaches an ROA of 8.2%, 6.4% in 2007, and 6.2% in 2010. However,

the average cattle farm on the long term has an ROA close to the level of inflation. 2010 is the exception with the best result of all specialization with 9.0% ROA.

*Specialization*	*2006*	*2007*	*2008*	*2009*	*2010*	*Mean*
Dairy	3,3%	5,8%	7,3%	4,8%	3,4%	*5,2%*
Cattle	-3,6%	2,4%	2,5%	1,0%	9,0%	*1,5%*
Grain	1,1%	13,0%	11,9%	1,7%	5,1%	*6,9%*
Diversified	2,4%	5,3%	8,5%	2,5%	7,3%	*5,0%*
*Mean*	*1,9%*	*6,4%*	*8,2%*	*2,9%*	*6,2%*	*3,1%*

Table 22 presents the results from a different perspective: the high leverage groups obtain better results, except for group 1 which presented strong signs of financial distress. The grain specialization on the contrary have opposite results, with an ROA decreasing with the leverage. However, when we consider this element year by year, this result cannot be observed, showing that time has to be taken in consideration. Except for the grain specialization, these results are consistent with what we observed for the ROE, showing that the economical performance increases with the leverage, and that 80% is a higher limit. Even cattle production obtains positive results when the leverage is between 40% and 80% (group 2 and 3). In the case of this specialization, the results of group 3 are better (without significant difference however) than group 2.

*Specialization*	*Group 1*	*Group 2*	*Group 3*	*Group 4*	*Group 5*	*Mean*
Dairy	-1,1%	6,1%	5,7%	4,5%	3,3%	5,2%
Cattle	0,7%	4,0%	4,8%	0,9%	-4,2%	1,5%
Grain	3,8%	5,8%	6,8%	8,6%	9,0%	6,9%
Diversified	2,2%	6,6%	6,6%	3,2%	4,8%	5,0%
*Global Mean*	2,0%	6,2%	6,2%	4,3%	4,2%	3,1%

Finally, we can say that H0 cannot be rejected for the first hypothesis and the second hypothesis tested:

- ROA and ROE tend to increase with the level of debt of the farms in Isère.

- Financial distress may occur when leverage increases, reducing ROA or ROE. Looking at the ROE, 80% seem to be the upper limit. The results for the ROA tend to show that 60% is probably more appropriate (except for dairy production).

4.3 Calculation of the Capitalization Rate Using Bond Yields Plus Risk Premium

This method, presented in part 3.3, estimates the capitalization or actualization rate through this equation:

Rp: Risk premium for the French market. It has to be increased for the micro-capitalization premium

The risk-free rate of the 10 years Governments bonds average is 3.00% (Source French central bank, average of monthly average from July 2011 to June 2012). It is interesting to notice that this rate reached an average of 2.569 % in June 2012, and it has never been so low (Dobson, 2012). The 12 month average will surely decrease in the following months, but on the long term it cannot be considered that this rate will remain at this low level.

Regarding the Beta, the tax authority recommends 0.8 to 1.1 for the food-processing and agricultural sector (Direction Générale des Impots, 2006). On their side, Fama & Fench (Laur, 2010) consider that the beta for the agricultural sector is 0.81 (time frame 1929-2009). For shorter periods of time, the agricultural sector beta remains around 0.70 to 0.91 (Laur, 2010). In our calculation the value of 0.8 was retained because it both answers to the necessity to cope with the recommendation of the tax authority and the necessity to be easy to used for consultants (these information are easy to access).

- The risk premium for the French market: from the literature review, we found a range of [3.8% : 5.0%] (see part 2.1.2)

- The small or micro-cap premium: from the literature review, we found a range of [4.12% : 7.74%] (see part 2.2.2).

Therefore the actualization rate has to be contained in the following range:

The results obtained with this method are really higher than what consultants use normally. The benefit of this fact is the following: consultants are not exposed to tax adjustment, because the administration cannot consider their results as hidden donation. The administration itself would retain a lower tax base for its calculations (as the overall price of the company is reduced by a higher actualization rate). However, this method is only an approach based on secondary data, and need to be compared with other approaches such as the WACC.

4.4 Approach of the WACC for Farming in Isère

4.4.1 Calculation of the Betas

Table 23 presents the results of the calculation of the betas for the grain farming specialization. Over the 24 companies concerned, the average beta obtained is 0.804, really close to the 0.8 to 1.1 proposed by the tax authorities. However, it is important to notice that this beta is a weighted average for the companies operating in grain and oilseed, with an average market capitalization of 843 M$ (around 693 M€ for 2012). The average total assets of the farms concerned in our study is 279 000 €.

Another element is worth noticing: the company size. The Average grain company in Western Europe has a market capitalization of 714 M$, compared to 1 997 M$ for North America.

*Member Name*	*Total Mkt Cap ($)*	*adjusted beta*	R²	*SE beta*	*â weigthed average*
SOCFIN	440 363 745	1.106	0.425	0.134
ODET	1 990 689 644	0.622	0.176	0.093
SIPEF NV	550 707 800	0.729	0.152	0.139
SOCFINASIA	598 033 371	0.419	0.010	0.126
ANGLO-EAST PLNTS	395 519 023	0.574	0.056	0.147
SOCFINAF	383 174 522	0.626	0.100	0.131
REA HOLDINGS	224 078 524	0.641	0.054	0.191
ADECOAGRO SA	834 190 505	1.220	0.246	0.289
M P EVANS GROUP	331 529 995	0.517	0.048	0.123
CONTINENTAL FARM	47 960 000	0.289	0.003	0.173
NARBOROUGH PLTNS	14 030 701	0.352	0.000	0.137
	*5 810 277 829*			0.144*	*0,722*
ADECOAGRO SA	834 190 505	1.220	0.246	0.289
TRIGON AGRI A/S	106 214 029	0.870	0.175	0.174
BLACK EARTH-SDR	129 321 577	1.017	0.217	0.194
SOCFINASIA	598 033 371	0.419	0.010	0.126
SOCFIN	440 363 745	0.499	0.020	0.182
BONIFICA FERR	190 783 713	0.574	0.010	0.353
CONTINENTAL FARM	47 960 000	0.289	0.003	0.173
AGRICOLE DE CRAU	29 847 642	-0.166	0.008	1.226
BOLLORE	4 053 210 291	0.785	0.344	0.093
	*6 429 924 873*			0.145*	*0,780*
ANDERSONS INC	663 031 853	1.132	0.377	0.153
VITERRA INC	4 573 354 378	0.834	0.142	0.184
SANDERSON FARMS	928 292 250	0.501	0.023	0.162
SEABOARD CORP	1 826 143 352	1.106	0.425	0.134
	*7 990 821 833*			0.167*	*0.882*
	20 231 024 535			0.154*	0.804

Table 23: Calculation of the Beta for the grain specialization (oilseed is considered as grain farming in France). Source: Bloomberg

For the cattle ranching, presented in Table 24, we see that there is more difference between Europe and North America. The average beta is really higher for the four European companies, with a beta close to 1. Once more, the range of references proposed by the Tax authorities corresponds to what is observed in Bloomberg. It is worth noticing the difference of the size of the companies between the two continents, but in the opposite direction with the grain specialization. European companies are more important, with an average market capitalization of 319 M$ compared with an average of 68 M$ for North America. However, the market capitalization of the listed companies operating in cattle ranching and farming is smaller than the companies operating in grain.

*Member Name*	*Total Mkt Cap ($)*	*adjusted beta*	R²	*SE beta*	*â weigthed average*
YASHENG GROUP	71 968 101	0.517	0.007	0.331
ALICO INC	133 115 937	0.929	0.259	0.150
CANAL CAPITAL CO	39 135	-2.083	0.018	3.955
	*205 123 173*			0.214*	*0.784*
NATIONAL MILK	3 871 194	0.274	0.011	0.085
ADECOAGRO SA	834 190 504	1.222	0.246	0.289
TRIGON AGRI A/S	106 214 028	0.870	0.175	0.174
M P EVANS GROUP	331 529 994	0.517	0.048	0.123
	*1 275 805 720*			0.236*	*1.007*
	1 480 928 893			0.233*	0.976

cattle ranching and farming NA
cattle ranching and farming NA
cattle ranching and farming NA

cattle ranching and farming WE cattle ranching and farming WE cattle ranching and farming WE cattle ranching and farming WE

Table 24: Calculation of the beta for the cattle ranching and farming specialization. Source: Bloomberg

Table 25 presents the companies specialized in poultry and fruits and vegetables. These
specializations are not studied in this paper, because the number of farms concerned is lower in

Isère. However, some regions are more concerned with these types of farms, and some consultants will find useful to work on these beta.

*Member Name*	*Total Mkt Cap ($)*		*adjusted beta*	R²	*SE beta*	*â weigthed average*
CAL-MAINE FOODS		715 100 119	0.814	0.238	0.128
SANDERSON FARMS		928 292 250	0.501	0.023	0.162
	1	*643 392 369*			0.147*	*0.637*
LDC		677 446 072	0.512	0.083	0.088
		*677 446 072*			0.088*	*0.512*
DOLE FOOD CO INC		633 782 418	1.081	0.254	0.191
VILLAGE FARMS IN		27 137 995	0.479	0.014	0.182
FRESH DEL MONTE	1	106 613 718	0.881	0.401	0.100
ALICO INC		133 115 937	0.929	0.259	0.150
	1	*900 650 068*			0.135*	*0.945*
CAMPOSOL HOLDING		84 172 880	0.558	0.043	0.159
CONTINENTAL FARM		47 960 000	0.289	0.003	0.173
BONIFICA FERR		190 783 713	0.574	0.010	0.353
		*322 916 593*			0.276*	*0.528*

fruits&vegetables NA fruits&vegetables NA fruits&vegetables NA fruits&vegetables NA

Table 25: Calculation of the beta for poultry and fruits & vegetables Source: Bloomberg

Table 26 presents a summary of the beta for the companies specialized in agriculture for each specialization. It is interesting to notice that the average market capitalization of North American companies is bigger than the Western European ones, but their beta is higher. This element is not consistent with what can be found in the literature about the relationship between the market size of a company and its beta. Therefore, we can conclude that the risk must be lower for farming in Western Europe than in North America.

Total oilseed farming in WE Total grain western europe Total grain north america

Table 26: Summary of the beta for the different specialization and regions Source: Bloomberg

4.4.2 Calculation of the WACC for the Different Specialization

Table 27 presents the results of the different leverage for each specialization. This leverage was be
used in the calculation of the estimation of the WACC for each specialization. Other financial
information are presented in this table, such as the net salary per capita (in €/month) which can be

compared to the SMIC which is around 1 029 € net per month for the time period 2006-2010. Therefore, the average cattle farmers in Isère earn less than the minimum salary in France.

The average leverage is really similar for each specialization, around 50%, but the debt coverage on the contrary is different depending on the specialization. The grain farmers have a debt coverage of 3.16, which is really different than the 2.08 for the cattle farmers. It shows again that cattle farming was the less profitable specialization in Isère for 2006-2010.

Table 27: Characteristic of each specialization in Isère *EBITDA/total annuity

Regarding the WACC, Table 28 presents the details of the calculation and the range of WACC for each specialization. First, the cost of debt is really low for farmers in France (see Appendix 14 page 115). The first element which could explain this fact is that loans are guaranteed by the assets in the farming business, reducing therefore the risk for the bank and the interest rate as well. The other element which can explain this fact is that the first bank in agriculture is the Crédit Agricole, which is really linked with the agricultural sector. His regional boards are mostly composed with farmers.

One could object that this low cost of debt doesn't reflect the real risk associated with the debt in Agriculture because of the specificity of this sector. However, the rate bankruptcy is really low: 1.6%o in Rhone Alpes (Coface, 2011). This element is linked with the culture of the French farmers. Going bankrupt is clearly not conceivable for most of them, and we can put this statement in relation with their level of salary. As presented in table 27, cattle farmers on average earn less than the minimum salary in France, but they maintain their activity and continue to pay their annuities. From an economical perspective, this attitude is totally counter-productive, because selling the farm and getting whatever job would generates higher returns. This fact illustrates the farmers' attachment to their farm and may explain the really good interest rates they obtain from their banks.

Table 28: WACC estimation for each specialization (leverage is based on table 27)

The cost of equity is higher for dairy and cattle farms (see Table 28), linked with the higher beta observed for these specializations. Finally, the WACC range for the average farm in Isère is [7.4% ; 10.3%]. This method gives lower results than the method proposed by the tax authorities. However, for some specializations, the range is almost similar, particularly for cattle and dairy production.

Another element has to be considered. As a matter of fact, the WACC methodology takes into
account the leverage, which reduces the overall cost of capital. Table 29 presents the WACC for the

farms of the group 5, for which the leverage range is 0-20%. Here the WACC increases tremendously, reaching almost 20% for the higher estimation for cattle and dairy production.

Table 29 : WACC estimation for each specialization for group 5 (0 to 20% leverage)

Group 2 (60-80% leverage) was usually the best group in terms of results (see parts 4.2.2 and 4.2.3). Regarding this group, the WACC falls to lower levels, between 5.6% and 7.4% for the average farm in Isère. This massive reduction of the WACC is due to the really low cost of debt observed in agriculture in France. The level is still higher for cattle and dairy farms of course, pushed by the higher beta for this production. The WACC for this group of leverage is closer to the actualization rate used by practitioners (2.5-4.5%), but still significantly higher.

Table 30: WACC estimation for each specialization for group 2 (60 to 80% leverage)

As presented in the previous sections, the signs of financial distress are strong in group 1 (leverage higher than 80%), and even for the group 2. Therefore, the real optimal WACC is maybe closer to the results presented in Table 28 for leverages around 50%. Farming consultants should use these ranges of WACC to determine their actualization rates in their feasibility studies.

5 Discussion

5.1 Survey Results

The results of the survey confirmed my initial observations: farm consultants do not often use the WACC method, and the discount factors used are really low generally, around 2.5% to 4.5%. Moreover, the estimations based on experience are the most popular, and we can make the assumption that they are influenced by the low level of bankruptcy in the sector. The means of the discount factors used by practitioners was compared to the ROE and ROA statistically observed in Isère and the required risk premium based on our calculation. As practitioners use lower discount factors than the required risk premium, it signifies that the estimations they produce are overestimated, which would lead farmers to base their decisions on overoptimistic simulations.

5.2 ROE and ROA Analysis the Farms of Isère

The statistical tests performed in part 4 confirm the underlying hypothesis of the WACC methodology: firm's performance increases with leverage up to a certain limit where financial distress occurs. The standard deviation increases in group 1, with lower median results. It can be safely assumed that 80% leverage is a limit that should not be exceeded in agriculture.

Groups 2 and 3 are the groups which have results significantly better than the lower groups, meaning that the optimal level of debt could be between 40% and 80%. As group 3 shows good results either, better than group 1 which shows strong signs of financial distress some years, it is common sense to think that the optimal debt level is closer to 60% than 80%. Except for the dairy specialization, it could be recommended to target a leverage of 40 to 60%.

For grain and cattle producers, the optimal level of debt is surely between 40% and 60%. The median ROA of the group 2 is never significantly higher than any other groups, and is even lower some years (without significant differences). The group 3 on the contrary presents more stable results in terms of ROA for these specializations (grain vs cattle). This element may be explained by the difference in nature between these two productions: dairy farms have to invest in expensive buildings and equipments, necessary to increase the volume of production (the size of the animal housing is directly linked with the level of production); However, for grain production, the most expensive equipments are the tractors, and there is no such linear relationship between the investment and the productivity. Land is more linked with the productivity, but availability of land is not linked with the financial capacity of farmers, and most of the time farmers rent the land. Therefore, lands are most of the time not financed by loans and have no impact on the leverage. For cattle farming, the animal housing is less necessary than for dairy farms, depending on the surface of pasture available for the herd. Therefore, a high leverage for these farms does not directly guarantee that the farm will benefit of a higher potential of production.

Other elements can be learned from the results of the analysis of the ROE and ROA of farms in Isère. The average farm in Isère has not so low results in terms of ROE and ROA: 8.2% and 3.1% respectively. ROE and ROA reaches respectively 11.8% and 6.2% for group 2, which is higher than what consultants use. If we consider that the historical ROA should be the minimum actualization rate, the difference with the actualization rate used by practitioners is quite high.

On average, the respondents answered that they used discount factors around 2.5% to 4.5% depending on the specialization. Moreover, the results obtained in this analysis are based on historical results: it does not take all the risks into account and reflects only the past results of the farms of Isère. Therefore, most agricultural consultants use overoptimistic discount factors, leading them to consider that some projects are profitable when not.

5.3 Capitalization Using the Bond Yield Plus Risk Premium

The range obtained using this method is [9.3% ; 12.7%], far from the 2.5% to 4.5% commonly used by the practitioners in agriculture, and even the historical ROA around 4.0% to 6.8% observed for the groups 2 and 3 of the farms from Isère (leverage between 40 and 80%). On the contrary, some specializations reached this range in terms of ROE, with 11.3% for grain. The first conclusion we can draw from these results is that farmers of Isère do not reach the required rate of return calculated for some specializations. The second conclusion is that we can now safely assume that most practitioners do not use the appropriate actualization rate in their consulting activities, leading farmers to base their decisions on overoptimistic results.

It is common sense to think that the first observation comes partially from the second. As a matter of fact, most of the farmers from the database studied in this research based their financial decision partially on the business plans calculated by the CERFRANCE Isère. Now, it is established that the risk premium was not enough taken into account in these business plans, leading some farmers to start a project without an appropriate economical assessment. This statement is particularly true when farms are sold to new farmers, because the valuation methods used by practitioners tend to overestimate the value of the company. In this case, the new farmer buys a farm which is too expensive regarding its expected performances, which decreases strongly the net present value of his investment. Eventually, this leads the overall sector to underperform on the long run.

This eventuality is hard to prove, but has to be considered. In Isère, the consultants calculate the price of farm using two valuation methods, as recommended by the tax authorities (Direction Générale des Impots, 2006). The most commonly used are the patrimonial method and the profitability method. The first one is easy to implement and requires the evaluation of the net assets of the farm. Buildings and equipments are valued at their resale price. In one sense, this is the value of the company if it had to be sold parts by parts. The second one is based on the actualization rate: we divide the reproducible net cash flow generated by the actualization rate. It is the NPV of a perpetual cash flow. A difference between 3% and 9% multiplies the value of the farm by 3 with this method! The most usual discount factors used by practitioners are 2 to 4 times lower than the range [9.3% ; 12.7%] obtained with the method recommended by the tax authorities, which means that the difference can be really important.

5.4 The WACC Methodology

The elements observed in part 5.3, regarding the results of the method bond yield plus risk premium proposed by the tax authorities, are consistent but higher than what is obtained with the WACC methodology for the average farm in Isère. This method confirms that the actualization rates used by practitioners are undervalued. However, it gives really different results depending on the leverage of the companies. Therefore, it gives a strong advantage to this approach for the practitioner, who will be able to adapt his actualization rate to the particularity of each farm. Another advantage is the possibility to choose the best financing solution for a new investment, considering the overall

leverage of the farm in order to modify its capital structure to reduce the WACC. The consequence of this modification should be the increase of the profit generated by the farm.

On the other side, the method proposed by the tax authorities has the strong advantage to be easier to implement and to learn for the consultants. The other advantage of the bond yield plus risk premium as presented by the French tax authority is its origin itself: in case of litigation with the administration, the choice of the actualization rate will be easier to defend with the method they recommend. The power of this argument is reduced by the recognition of legitimacy of other methods from the administration itself, without describing or citing them however (Direction Générale des Impots, 2006).

Finally, we can observe that the WACC methodology should be more accepted by the practitioners, as it gives lower results for the average farms in Isère. This element has to be considered, because such a strong modification of the actualization rates used by the consultants would signify that all their previous studies were significantly wrong. It could lead to a movement of rejection from these practitioners.

6 Utilization of These Results

6.1 Managerial Implications

The first conclusion that can be learned from this research is that agricultural consultants do not use adapted discount factors. The lack of research and publication on the subject is leading consultants to adopt discount factors based on the cost of debt or inflation or even only on risk-free assets, and not on the cost of capital or a real risk premium. The first managerial implication that should be drawn is the necessity of further research on the topic. CERFRANCE developed a partnership with a network of engineering school in agriculture to specialize 40 students in consulting for farmers every year (CNCER, 2012). This partnership with teaching and research facilities could be the base of the discussion about the necessity to improve the knowledge on the topic. The engineering schools in agriculture are the perfect support for a deeper research on the subject, and this option should be considered as the partnership with the CERFRANCE would help the researchers to access a unique database that contains all economical results of more than 50% of all French farmers. Finally, the CNCER (the central element of the CERFRANCE network) should initiate a project to uniform the practices among the different members of the network. It is clearly a weakness of the network so far.

The other implication induced by this research is the necessity to improve the methods actually used by most practitioners of the CERFFRANCE. Our survey is not exhaustive, therefore it cannot be said that all the CERFRANCE are concerned, neither all its consultants. Some answers prove that some of them use higher rates. However, generally speaking it is obvious that the different methods to determine appropriate actualization rate are not well understood by most of the consultants. In-house training should be more focused on these methods, as well as on the consequences of choosing extremely low discount factors. This type of training could start without waiting for other results, because even the descriptive statistics show that for some groups of leverage the average ROE and ROA are really higher than the discount factors used by practitioners. The ROE cannot be the right estimation for a discount factor. However, it seems to be misleading to choose it two times lower than the ROA for the best groups of farms.

Consultants and farmers should also adapt their methodologies for the management of the working capital. This element is really important in agriculture, and represents 34.5% of the total assets in dairy farming in Isère and 43% for the cattle farms. This working capital is principally composed by the animals (66% for dairy, and 79% for cattle farms). This working capital, for the farms that are not enough leveraged, could be financed with debt instead of being financed with equities, which is usually the case in agriculture. This financing is possible to put in place, as the financial institutions could take a guarantee on the herd of cattle. The farmers would be able to recuperate in cash a part of its equities, as well as reducing its overall cost of capital. The benefit could be important, both for the financial institution which would increase its activities, and also for the farmers who would be able to increase the NPV of their farms and their personal wealth. An estimation of the possible gain for the farmer is presented in Table 31. The hypotheses of this calculation are:

- 4% interest rate, higher than the average rate for farms from Isère, because the loan would be riskier for the bank than a loan guaranteed by lands or constructions,

- 40 000 € total loan, (45% of the total value of the animals for an average dairy farm), - 10% actualization rate,

- 6 years maturity, considering that a herd is renewed at 25% each year on average and calves become productive in their third year.

	0			1		2			3			4			5		6
40	000
		-7		510	-7	510	-7		510		-7	510		-7	510	-7	510
				429		356			281			203			122		44
40	000	-7		080	-7	153	-7		229		-7	307		-7	388	-7	466
1,000			0,909		0,826			0,751		0,683			0,621			0,564
40 000			-6 437		-5 912			-5 431		-4 991			-4 587			-4 214
8 429 €

Table 31: NPV of a refinancing operation to reduce the WACC and increase the wealth of a dairy farm

The example presented in table 31 would not be interesting if the actualization rate was set at 3% like most consultants do. However, it becomes really interesting when the actualization rate is set in a normal range for dairy farms (between 8.3% and 11.7% as presented in section 4).

The only limitation of this strategy is to get the acceptation of the financial institution which would increase its risk exposure. So far, the risk associated with the working capital was mostly supported by the farmer itself. However, this limitation could be reduced if the farmers subscribe to a predefined percentage of the cash generated by the new loan on a financial product offered by the institution. This solution would also be interesting for the preparation of the retirement of the farmers, who are most of the time not well prepared considering the low level of the pensions in the sector.

6.2 Agricultural Policies Implications

The actual Common Agricultural Policy (CAP) does not take into consideration many elements of the risk in agriculture:

- The climate risk (some subsidies are now conditioned to the subscription to climatic

Our results show that some years are profitable, when we consider 2007, 2008 or even 2010. However, 2006 and 2009 were really poor in terms of results. The future CAP will be reformed in 2014 or 2015, depending on the complexity of the negotiations to reach a consensus among European countries. It could be a great opportunity to consider the evolution of the risk in agriculture during this reform. Some analysts suggest that countercyclical tools could help farmers to maintain their revenue during downturns and limit the overall cost of the CAP (Momagri, 2012b). This paper shows that cyclicality of revenues and profitability are really strong. Instability comes from the price inelasticity regarding the soft commodities: 1 or 2% change in the level of global production can lead to a price variation of 50 to 100%, and even more (Momagri, 2012b). Therefore, the opportunity of changing the CAP to integrate countercyclical tools should be strongly considered, even if this would lead to add complexity to a system already considered as opaque and complex by non-initiated observers.

Other policies should be included in the CAP, which are the public subsidies for new farmers. So far, only the French government subsidizes the new farmers through loans at low interest rates and a front subsidy for the first year of activity. These elements could be generalized at the European level, particularly for the loan subsidy which could be an easy way to redirect the CAP payments to more socially acceptable subsidies. Lower interest rates and public support for loan's access can be as profitable in the long term for farming activities as decoupled payments or direct aids which represent a large part of the cost of the CAP (see Figure 6, page 16).This type of support would have a lower impact on the price of the grains than direct payments, as it would be more oriented on maintaining the number of young farmers high enough to maintain the age pyramid of this population.

7 Limitations and Further Research Implications

7.1 Survey Construction

The purpose of this survey is not to have an exhaustive view of the tools used by the practitioners in the field. As the questionnaire was sent to consultants of different CERFRANCE, the answers cannot be considered as representative for all accounting companies working with farmers. However, the CERFRANCE has a dominant position in this particular market all over the country and serve more than 50% of the farmers. Therefore, it could be appropriate for the CNCER to intensify its researches on the subject to uniform the practices across all CERFRANCE.

7.2 ROE and ROA Analysis

As we can see in Table 7 page 29, some years like 2009 count fewer companies available for analysis, even if the number of books accounts produced by the CERFRANCE Isère increased constantly over the 5 years. Therefore, some companies were not included in the research because of limitations of the information system. However, this limitation cannot be considered as enough relevant to invalidate the results obtained, mainly because the non-inclusion in the database for a company occurs randomly (there is no reason not to include results in the database).

Another limitation comes from the decision to work on a constant population. This constant population enables an interpretation of the results over several years, but it also removes the companies that went bankrupt during the studied period. Therefore, this exclusion criterion may have an effect on the results for the hypothesis 2 which analyses the effect of financial distress. However, the number of bankruptcy in agriculture is really low: only 63 farms went bankrupt in 2011 in Rhone Alpes (Coface, 2011) over 39 020 farms (Agreste, 2011), which represents 1.6%o bankruptcy in the sector for the year.

Another limitation comes from the number of farms available for the cattle specialization. These farms, specialized in beef production are not enough represented in Isère to obtain reliable results. Other regions such as Côtes d'Or or Saône-et-Loire should work on the subject.

Finally repeated measure statistical test could have been performed to test the data to try to evaluate more precisely the ideal debt leverage ratio. However, these mixed models are not available with traditional statistical software as they do no compare means or medians but probabilities (here probability to fall in financial distress in year N+1 depending on the debt level of year N). As this research paper tries to serve as a base for all the consultants of the CERFRANCE network, this analysis needs to be replicable with the results of other departments without expensive tools such as SAS or SPSS. However, this limitation justifies further researches on the topic.

7.3 Bond Yield Plus risk Premium Model

The bond yield plus risk premium model is easier to calculate than the WACC for the farms, which represents a strong advantage for this method. However, this model does not take into account the financial structure of the companies studied. This strong limitation reduces its applicability to the farms of Isère, which are really diversified.

The research was based on elements found in the literature for this part of the research. It has been
decided to choose this method in order to try to estimate what actualization rate could be chosen

without exhaustive researches on the subject. The purpose of this choice is to compare the actualization rate used by practitioners with the rates the tax authorities could use in case of litigation.

7.4 The WACC Methodology

The first limitation of this methodology is really easy to identify: this method, based on the CAPM, is really dependent on the beta of the company studied. However, the beta cannot be calculated for non listed companies, and it is necessary to extrapolate the results of listed companies to obtain an appropriate beta for the small and medium farming businesses studied in this paper. The other limitation of the method is directly linked to the first one: the market value of the firm should be estimated by its market capitalization. However, for non-listed companies this value cannot be calculated. The book value used in our research cannot replace fully the market capitalization, because it does not reflect the opinion of the market on the company.

The second limitation of this method is the size of the farms studied compare to the size of the listed companies used to calculate the betas. Even if a premium was added to the betas of the listed companies, it remains a bias not considered by the WACC methodology.

The last limitation is more linked with the choice of the expected return. It has been chosen on purpose to adopt the risk premium to estimate E(Rm) - Rf, in order to be able to compare the results of the two methodologies. However, it could have chosen to use the historical results of the farms from Isère of the first quartile (the 25% most performing farms).

7.5 Further Research Implications

First, there is a clear necessity to intensify the research about the optimal level of debt in Agriculture. Repeated measure statistical tests could be used to work deeper on the subject, in order to identify more clearly the optimal level of debt, but also to identify what are the other factors that have a strong impact on the financial performance of farms. It seems obvious that the productivity of farmers (number of hectares or cows per FTE) and the surface of the farm play a great role in the results. However, it is really hard to find quantitative research on the subject in France, at least researches that could be used easily by financial consultants.

The other further research implication is the necessity to test the methods presented here in the long run, in order to compare them. It would be interesting to follow many projects evaluated with the NPV method in order to see finally which method gave the most appropriate actualization rate. Moreover, other methods should be tested either to identify which ones should be chosen by practitioners. So far, this research tends to prove that the WACC methodology can be used by consultants, even if some theoretical limitations arose.

Finally, the same research methodology should be applied in other departments than Isère, with other specializations also. This would give further indications on the adaptability of the WACC methodology to the agricultural sector. It would allow also consultants to estimate the appropriate discount factors for farms specialized in fruits, wines or vegetables.

Conclusion

The WACC methodology tested in this research presents real limitations due to its underlying theories based on the market efficiency. This assumption holds only for the listed companies, introducing a bias in the calculation presented in this paper. However, the results obtained are consistent with the initial expectations regarding the relationship between leverage and financial performance or financial distress. Therefore, consultants should take more consideration for this methodology and use it in their feasibility studies.

The other achievement of this paper concerns the discount rates actually used by practitioners. All signs clearly indicate that many of them use abnormally low actualization rates, ranging from 2.5 to 4.5%. The impact of these low actualization rates in valuation methods, such as the profitability method, can be really important. The NPV of a farm can be over-estimated by two to four using such discount factors! Some consultants use more appropriate rates, but it is far from being a generality according to the results of the survey.

Then, it appears clearly that leverage has a positive impact on the financial performance of the farms of Isère. These expected results confirm that the WACC methodology should hold in the context of small and medium farming business. Therefore, consultants should consider the capital structure of the farms into consideration not only to avoid the risk of financial distress, but also to look for the optimal leverage, which appears to range between 40 and 60%. From the results presented in this paper, the 60-80% leverage group presented really good performances, but also more variability. Moreover, the bankruptcy risk was not studied because a sample of farms studied had to be constant. Therefore, it is safer to consider that the optimal debt level lies in the group 40-60%, may be closer to 60% regarding the good performance of the 60-80% group, particularly for the dairy specialization.

Finally, the consultants of the CERFRANCE Isère do not need to wait for other researches on the field to modify their methods. Main recommendation is to increase significantly their actualization rates. On the other side, it must be acknowledge that further researches on the topic are necessary to improve the methodology and determine more precisely the hypothesis that should be taken, regarding the risk premiums and the beta for example. The productivity constraints of the consultants working in the different CERFRANCE militate in favor of a partnership with the engineering schools specialized in agriculture. This partnership, already implemented in some schools, could be the starting point of others researches about the financial performances of farming businesses.

Bibliography

Agreste. (2011 b). Agreste Primeur, France Métropolitaine, Recensement agricole 2010. Montreuil Sous Bois: Agreste.

Agreste. (2011). Commerce extérieur agroalimentaire. Agreste Conjoncture , 2 (12), pp. 1-5.

Agreste. (2011 c). Premiers résultats en Isère. Agreste Rhône Alpes, Coup d'Oeil , 129, 4 p.

Al-Rjoub, S., Varela, O., & Hassan, M. (2005). The size effect reversal in the USA. Applied Financial Economics , 15 (17), 1189-1197.

Ambassade de France à Washington. (2009). Les Etats Unis en chiffre. Washington DC: République Française.

Barry, P. J. (1980a). Capital Asset Pricing and Farm Real Estate. American Agricultural Economics Association , 62 (3), 549.

Barry, P. J. (1981). Capital Asset Pricing and Farm Real Estate: Reply. American Agricultural Economics Association , 63 (3), 580.

Barry, P. J. (1980b). Financial Impacts of Government Support Price Programs: Comment. American Journal Of Agricultural Economics , 62 (3), 599-604.

Bergstrom, G., Frashure, R., & Chisholm, J. (1991). The Gains from international small-company in Global Portfolios: Quantiative Strategies for Maximum Performance. Homewood: R.Z. Aliber and B.R. Bruce, Business One Irwin.

Bhardwaj, R. K., & Brooks, L. D. (1993). DUAL BETAS FROM BULL AND BEAR MARKETS: REVERSAL OF THE SIZE EFFECT. Journal of Financial Research , 16 (4), 269-284.

Binder, J. J. (1992). BETA, FIRM SIZE, AND CONCENTRATION. Economic Inquiry , 30 (3), 556-563.

Birner, R., Davis, K., Pender, J., Nkonya, E., Anandajayasekeram, P., & Ekboir, J. (2009). From best practice to best lit: a framework for designing and anlyzing pluralistic agricultural advisory services worldwide. Journal of Agricultural Education and Extension , 15, 341-355.

Borrel, B., & Hubbard, L. (2000). Global Economic Effects of the EU Common Agricultural Policy. Economic Affaris , 20 (2), 18-26.

Brealey, R. A., Myers, S. C., & Allen, F. (2008). Principles of corporate finance (Ninth international edition ed.). New York: McGraw Hill.

Breisinger, C., Ecker, O., & Al-Riffai, P. (2011). Economie du Printemps Arabe. De la révolution à la transformation et la sécurité alimentaire. Politiques alimentaires en perspective , 18, 2-4.

Carter, C. (1981). Capital Asset Pricing and Farm Real Estate: Comment. American Agricultural Economics Association , 63 (3), 578.

CCI Rhône Alpes. (2011). Les Chiffres Clés de l'Isère, édition 2010-2011. Grenoble: Chambre de Commerce et d'Industrie de Grenoble.

Chambers, R. G. (1989). Insurability and Moral Hazard in Agricultural Insurance Markets. American Journal of Agricultural Economics , 85 (2), 604-616.

Chan, K., & Chen, N.-F. (1988). An Unconditional Asset-Pricing Test and the Role of Firm Size as an Instrumental Variable for Risk. Journal of Finance , 43 (2), 309-325.

Claus, J., & Thomas, J. (2001). ). Equity Premia as Low as Three Percent? Evidence from Analysts' Earnings Forecasts for Domestic and International Stock Markets. Journal of Finance , 56 (5), 1629- 1666.

Clipici, E. (2011). Directions of Common Agricultural Policy. Agricultural Management , 13 (1), 173- 178.

Dimson, E., & Marsh, P. (2001). U.K. Financial Market Returns, 1955-2000. Journal of Business , 74 (1), 1-31.

Dongcheol, K. (1993). The Extent of Nonstationarity of Beta. Review of Quantitative Finance & Accounting , 3 (2), 241-254.

Drew, M. E., & Veeraraghavan, M. (2003). Beta, Firm Size, Book-to-Market Equity and Stock Returns. Journal of the Asia Pacific Economy , 8 (3), 354-380.

Elekes, A., & Halmai, P. (2009). The "Hottest Topic" of the budgetary review: Will the CAP survive? Intereconomics , 44 (5), 300-308.

Faure, G., & Compagnone, C. (2011). Les Transformations du conseil face à une nouvelle agriculture. Cahiers de l'Agriculture , 20 (5), 321-326.

Faure, G., Desjeux, V., & Gasselin, P. (2011). Revue Bibliographique sur les recherches menées dans le monde sur le conseil en agriculture. Cahiers de l'Agriculture , 20 (5), 327-342.

Gangwar, L., Singh, D., & Mandal, G. (2008). Economic Evaluation of Peach Cultivation in North Indian Plains. Agricultural Economics Research Review , 21 (1), 123-129.

Gardner, B. L. (1994). Liberalization in New-Zealand agriculture: Discussion. American Journal of Agricultural Economics , 76 (5), 1053.

Hanson, S. D., & Myers, R. J. (1995). Testing for a time-varying risk premium in the returns to U.S. farmland. Journal of Empirical Finance , 2 (3), 265-276.

Harwood, J., Heifner, R., Coble, K., Perry, J., & Somwaru, A. (1999). Managing Risk in Farming: Concepts Research and Analysis. Agricultural Economic Report , 774.

Hénin, F. (2011, December 10). Impact de la fin des quotas sur les accords de prêts aux éleveurs. Retrieved April 01, 2012, from www.web-agri.fr: http://www.web-agri.fr/actualite-
agricole/economie-social/article/les-banques-apprennent-a-travailler-avec-la-fin-des-quotas-1142- 76501.html

Holzman, & Jorgensen. (2001). Social Risk management: A New conceptual framework for social protection, and beyond. International Tax and public Finance , 8, 529-556.

Howarth, R. (2000). The CAP: History and attempts at reform. Economic Affairs , 20 (2), 4.

Israelsen, C. (2009). The Value Premium. Financial Planning , 39 (6), 108-110.

Jegadeesh, N. (1992). Does Market Risk Really Explain the Size Effect? Journal Of Financial & Quantitative Analysis , 27 (3), 337-351.

Johnson, M. R. (2002). Kansas's Agricultural Use Value Method. Assessment Journal , 9 (3), 23-29.

Laur, P. (2010). Le Beta varie-t-il pour de bonnes raisons ? L'influence des caractéristiques spécifiques des firmes et du contexte macroéconomique sur le coefficient de risque systématique ß. Retrieved July 10, 2012, from www.vernimmen.net: http://www.vernimmen.net/ftp/Memoire_Pierre_Laur.pdf

Leonard, S. (2009). U.S. Equity Returns After Major Market Crashes. Journal of Financial Planning , 22 (10), 84-94.

Miranda, M., & Glauber, J. (1997). Systemic Risk, Reinsurance, and Failure of Crop Insurance Markets. American Journal of Agricultural Economics , 79, 206-215.

Moravcsik, A. (1999). Le grain et la grandeur : les origines économiques de la politique européenne du général de Gaulle (1ère partie). Revue Française de Science Politique , 49 (4-5), 507-544.

Mundler, P., Labarthe, P., & Laurent, C. (2006). Les disparités d'accès au conseil: le cas de la région Rhône-Alpes. Economie Rurale (291), 26-34.

Nartea, G., & Webster, P. (2008). Should farmers invest in financial assets as a risk management strategy? Some evidence from New Zealand. Australian Journal Of Agricultural & Resource Economics , 52 (2), 183-202.

OECD. (2009). Managing Risk in Agriculture: a Holistic Approach. Paris: Editions OCDE.

Petit, M. (2006). L'exploitation agricole familiale: leçons actuelles de débats anciens. Cahiers Agricultures , 15 (6), 486-490.

Quiry, P., & Le Fur, Y. (2012). Finance d'Entreprise. Pierre Vernimmen 10ème édition. Paris: Dalloz.

Rose, F. (2008). A note on the role of investment capital in the US agricultural futures markets and the possible effect on cash prices. Groupe sur les céréales, les aliments du bétail et le sucre. OECD.

Salomons, R., & Grootveld, H. (2003). The equity risk premium: Emerging vs Developed Markets. Emerging Markets Review , 4, 121-144.

Sandrey, R. A., & Scobie, G. M. (1994). Changing international competitiveness and trade: Recent experience in New Zealand agriculture. American Journal Of Agricultural Economics , 76 (5), 1041.

Saunders, C., Wreford, A., & Cagatay, S. (2006). Trade Liberalisation and greenhouse gas emissions: the case of the dayring in the European Union and New-Zealand. Australian Journal of Agricultural & Resources Economics , 50 (4), 538-555.

Sharpe, W. F. (1964). Capital asset prices: a theory of market equilibrium under conditions of risk. Journal of Finance , 19 (3), 425-442.

Shomir, S., Pat, O., & Jeong-gil, C. (2011). FURTHER INVESTIGATION ON THE VARIABILITY OF INDIVIDUAL STOCK BETA AND PORTFOLIO SIZE. International Journal Of Business, Accounting, & Finance , 5 (1), 91-102.

Sminou, K. (2010). Stock Market and Agricultural Futures Diversification: An International Perspective. Journal Of Alternative Investments , 12 (4), 36-57.

Switzer, L. (2010). The behaviour of small cap vs. large cap stocks in recessions and recoveries: Empirical evidence for the United States and Canada. The North American Journal of Economics and Finance , 21 (3), 332-346.

Waintrop, M. (2010, August 18). Pourquoi le lait Allemand est-il moins cher que le Français. Retrieved March 15, 2012, from www.la-croix.com: http://www.la-croix.com/Actualite/S-
informer/Economie/Pourquoi-le-lait-allemand-est-moins-cher-que-le-francais-_NG_-2010-08-18- 555741

WTO. (2011). International Trade Statistics 2011. Geneva: World Trade Organization.

Zaharia, I., Tudorescu, N., & Zaharia, C. (2009). Have the Cap Reforms Delivered Significant Changes. Economics, Management & Financial Markets , 4 (2), 107-110.

L'objectif de cette enquête est d'avoir un aperçu des techniques utilisées par les conseillers d'entreprise agricole pour actualiser les flux. Les résultats de cette enquête seront analysés de manière anonyme, et seront utilisés dans le mémoire « L'utilisation du CMPC en vue d'optimiser l'évaluation du coût du capital dans les exploitations agricoles en France. Constatations empiriques à partir des résultats économiques d'exploitations Iséroises ». Une version électronique de ce mémoire pourra être distribuée aux personnes intéressées d'ici octobre.

Question 3: Dans le cadre de votre profession, rédigez vous des études prévisionnelles pour vos adhérents afin d'estimer la rentabilité de leurs projets ?

Question 4: est-ce que vous utilisez des taux d'actualisation pour évaluer la valeur d'une entreprise agricole ? (actualiser un flux perpétuel: flux prévisionnel reproductible / taux d'actualisation)

Question 5: Est-ce que vous utilisez la méthode de la Valeur Actuelle Nette (VAN) pour estimer la rentabilité d'un projet ? (projet photovoltaïque ou méthanisation par exemple)

Question 6: comment choisissez vous un taux d'actualisation ? Cochez les méthodes utilisées dans votre structure.

Question 7: Dans quelle fourchette de pourcentage se situent les taux d'actualisation que vous utilisez principalement pour votre clientèle agricole ?

Question 8: Comment qualifieriez vous les méthodes utilisées dans votre structure pour comparer et évaluer des projets d'investissement ?

Question 9: Si l'on vient a vous proposer un outil pour estimer le coût moyen pondéré du capital des exploitations de votre région afin de l'utiliser comme taux d'actualisation, comment considéreriez vous cet outil ?

Appendix 4: Normality test of the datasets for ROE, extraction from statgraphics

The Shapiro Wilk test for normality cannot be performed. Therefore we observed the data graphically, and considered that the data follow a normal distribution (density trace and histogram of frequency look normal).

This analysis shows the results of fitting a normal distribution to the data on ROEc. The estimated parameters of the fitted distribution are shown above. You can test whether the normal distribution fits the data adequately by selecting Goodness-ofFit Tests from the list of Tabular Options. You can also assess visually how well the normal distribution fits by selecting Frequency Histogram from the list of Graphical Options. Other options within the procedure allow you to compute and display tail areas and critical values for the distribution. To select a different distribution, press the alternate mouse button and select Analysis Options.

This pane shows the results of several tests run to determine whether ROEc can be adequately modeled by a normal

distribution. The Shapiro-Wilk test is based upon comparing the quantiles of the fitted normal distribution to the quantiles of the data. The Shapiro-Wilk test was not performed because the sample size was greater than 2000.

This pane shows the results of tests run to determine whether ROEc can be adequately modeled by a normal distribution. Since the smallest P-value amongst the tests performed is less than 0,05, we can reject the idea that ROEc comes from a normal distribution with 95% confidence.

Appendix 5: Normality tests of the datasets for ROA, extraction from statgraphics Uncensored Data - ROA by time

Too much data for the Shapiro test. We do it visually, and we can conclude that it is normally distributed! 3046 values ranging from -4,37674 to 3,08145

This analysis shows the results of fitting a normal distribution to the data on ROA. The estimated parameters of the fitted distribution are shown above. You can test whether the normal distribution fits the data adequately by selecting Goodness-ofFit Tests from the list of Tabular Options. You can also assess visually how well the normal distribution fits by selecting Frequency Histogram from the list of Graphical Options. Other options within the procedure allow you to compute and display tail areas and critical values for the distribution. To select a different distribution, press the alternate mouse button and select Analysis Options.

This pane shows the results of several tests run to determine whether ROA can be adequately modeled by a normal
distribution. The Shapiro-Wilk test is based upon comparing the quantiles of the fitted normal distribution to the quantiles of

the data. The Shapiro-Wilk test was not performed because the sample size was greater than 2000.

This pane shows the results of tests run to determine whether ROA can be adequately modeled by a normal distribution.

Since the smallest P-value amongst the tests performed is less than 0,05, we can reject the idea that ROA comes from a normal distribution with 95% confidence.

This procedure performs a one-way analysis of variance for ROE. It constructs various tests and graphs to compare the mean values of ROE for the 5 different levels of ANNEE. The F-test in the ANOVA table will test whether there are any significant differences amongst the means. If there are, the Multiple Range Tests will tell you which means are significantly different from which others. If you are worried about the presence of outliers, choose the Kruskal-Wallis Test which compares medians instead of means. The various plots will help you judge the practical significance of the results, as well as allow you to look for possible violations of the assumptions underlying the analysis of variance.

ANNEE	Count	Average	Standard deviation	Coeff. of variation	Minimum	Maximum	Range	Stnd. skewness
2006	568	-0,0079864	0,541224	-6776,82%	-3,25	3,7011	6,9511	1,84708
2007	568	0,0882099	0,498518	565,15%	-3,90675	3,31882	7,22556	-5,94064
2008	568	Scat 0,07103	by Leve 0,60985	858,581%	-3,3045	3,703	7,0075	-0,658518
2009	568	0,00166802	0,489926	29371,7%	-3,68929	3,90305	7,59234	-11,2575
4 2010	568	0,0906529	0,628869	693,71%	-3,47483	3,57	7,04483	3,45006
Total	2840	0,0487149	0,557841	1145,11%	-3,90675	3,90305	7,80979	-2,20759
2 ANNEE	Stnd. kurtosis
2006	72,8003
2007	101,535
0 2008	78,8042
2009	114,527
2010 2	67,3229
Total	187,103

This table shows various statistics for ROE for each of the 5 levels of ANNEE. The one-way analysis of variance is

primarily intended to compare the means of the different levels, listed here under the Average column. Select Means Plot

WARNING: The standardized skewness and/or kurtosis is outside the range of -2 to +2 for 5 levels of ANNEE. This indicates some significant nonnormality in the data, which violates the assumption that the data come from normal distributions. You may wish to transform the data or use the Kruskal-Wallis test to compare the medians instead of the means.

The ANOVA table decomposes the variance of ROE into two components: a between-group component and a within-group component. The F-ratio, which in this case equals 4,23794, is a ratio of the between-group estimate to the within-group estimate. Since the P-value of the F-test is less than 0,05, there is a statistically significant difference between the mean ROE from one level of ANNEE to another at the 95,0% confidence level. To determine which means are significantly different from which others, select Multiple Range Tests from the list of Tabular Options.

					Stnd. error
ANNEE	Count		Mean		(pooled s)	Lower limit	Upper limit
0 2006	568		-0,0079864		0,0233533	-0,0351483	0,0191755
2007	568		0,0882099		0,0233533	0,061048	0,115372
2008 004	568		0,07103		0,0233533	0,043868	0,0981919
2009	568		0,00166802 2007		,233533 2008 2	-0,0254939 2010	0,0288299
2010	568		0,0906529		0,0233533 NNEE	0,063491	0,117815
Total		2840		0,0487149

This table shows the mean ROE for each level of ANNEE. It also shows the standard error of each mean, which is a measure of its sampling variability. The standard error is formed by dividing the pooled standard deviation by the square root of the number of observations at each level. The table also displays an interval around each mean. The intervals currently displayed are based on Fisher's least significant difference (LSD) procedure. They are constructed in such a way that if two means are the same, their intervals will overlap 95,0% of the time. You can display the intervals graphically by selecting Means Plot from the list of Graphical Options. In the Multiple Range Tests, these intervals are used to determine which means are significantly different from which others.

ANNEE	Count		Mean		Homogeneous Groups
2006	568		-0,0079864		X
2009	568		0,00166802		X
2008	568		0,07103		X
2007	568		0,0882099		X
2010	568		Box- 0,0906529		W h X
Contrast		Sig.		Difference		+/- Limits
2006 - 2007 2006		*		-0,0961963		0,0543239
2006 - 2008		*		-0,0790164		0,0543239
2006 - 2009				-0,00965442		0,0543239
2007 2006 - 2010		*		-0,0986393		0,0543239
2007 - 2008				0,0171799		0,0543239
2008 2007 - 2009		*		0,0865419		0,0543239
2007 - 2010				-0,00244305		0,0543239
28 - 2009 2009		*		0,0693619		0,0543239
2008 - 2010				-0,019623		0,0543239
2009 - 2010 2010		*		-0,0889849		0,0543239

This table applies a multiple comparison procedure to determine which means are significantly different from which others. The bottom half of the output shows the estimated difference between each pair of means. An asterisk has been placed next to 6 pairs, indicating that these pairs show statistically significant differences at the 95,0% confidence level. At the top of the page, 2 homogenous groups are identified using columns of X's. Within each column, the levels containing X's form a group of means within which there are no statistically significant differences. The method currently being used to discriminate among the means is Fisher's least significant difference (LSD) procedure. With this method, there is a 5,0% risk of calling each pair of means significantly different when the actual difference equals 0.

	Test		P-Value
Levene's	1,56665		0,180425
Comparison		Sigma1		Sigma2	F-Ratio	P-Value
2006 / 2007		0,541224		0,498518	1,17867	0,0506
2006 / 2008		0,541224		0,60985	0,787605	0,0045
2006 / 2009		0,541224		0,489926	1,22038	0,0179
2006 / 2010		0,541224		0,628869	0,740685	0,0004
2007 / 2008		0,498518		0,60985	0,668216	0,0000
2007 / 2009		0,498518		0,489926	1,03539	0,6790
2007 / 2010		0,498518		0,628869	0,628409	0,0000
2008 / 2009		0,60985		0,489926	1,54948	0,0000
2008 / 2010		0,60985		0,628869	0,940428	0,4649
2009 / 2010		0,489926		0,628869	0,606932	0,0000

The statistic displayed in this table tests the null hypothesis that the standard deviations of ROE within each of the 5 levels of

ANNEE is the same. Of particular interest is the P-value. Since the P-value is greater than or equal to 0,05, there is not a

statistically significant difference amongst the standard deviations at the 95,0% confidence level.

The table also shows a comparison of the standard deviations for each pair of samples. P-Values below 0.05, of which there are 7, indicate a statistically significant difference between the two sigmas at the 5% significance level.

The Kruskal-Wallis test tests the null hypothesis that the medians of ROE within each of the 5 levels of ANNEE are the same. The data from all the levels is first combined and ranked from smallest to largest. The average rank is then computed for the data at each level. Since the P-value is less than 0,05, there is a statistically significant difference amongst the medians at the 95,0% confidence level. To determine which medians are significantly different from which others, select Box-and-Whisker Plot from the list of Graphical Options and select the median notch option.

0,1 ANNEE	Sample Size	n<=	n>	Median	90,0% lower CL	90,0% upper CL
2006	568	326	242	0,0195117	0,00305438	0,041546
0,08 2007	568	245	323	0,0910952	0,0759334	0,110756
2008	568	250	318	0,0837851	0,065602	0,0957051
0,06 2009	568	328	240	0,0176906	0,00193535	0,0337506
2010 ⁰⁰⁴	568	271	297	0,0734001	0,0512998	0,091961

Mood's median test tests the hypothesis that the medians of all 5 samples are equal. It does so by counting the number of

observations in each sample on either side of the grand median, which equals 0,0583012. Since the P-value for the chi-

square test is less than 0,1, the medians of the samples are significantly different at the 90,0% confidence level. Also

included (if available) are 90,0% confidence intervals for each median based on the order statistics of each sample.

This procedure performs a one-way analysis of variance for ROA. It constructs various tests and graphs to compare the mean values of ROA for the 5 different levels of ANNEE. The F-test in the ANOVA table will test whether there are any significant differences amongst the means. If there are, the Multiple Range Tests will tell you which means are significantly different from which others. If you are worried about the presence of outliers, choose the Kruskal-Wallis Test which compares medians instead of means. The various plots will help you judge the practical significance of the results, as well as allow you to look for possible violations of the assumptions underlying the analysis of variance.

ANNEE		Count	Average	Standard deviation	Coeff. of variation	Minimum	Maximum	Range	Stnd. skewness
2006		568	-0,013372	0,223258	-1669,59%	-2,05464	0,701545	2,75619	-31,7124
2007		568	0,045657	0,198647	435,086%	-1,41503	0,677594	2,09262	-20,8092
2008		568	Scatte 0,0677319	by Level 0,328621	485,179%	-2,07651	2,83152	4,90803	33,4605
2009		568	-0,0042941	0,204192	-4755,18%	-1,50563	0,600911	2,10654	-23,9402
2,9 2010		568	0,0572362	0,292304	510,697%	-1,41254	2,82314	4,23567	28,9587
Total		2840	0,0305918	0,256688	839,075%	-2,07651	2,83152	4,90803	33,8379
ANNEE	Stnd. kurtosis
2006	118,296
0,9 2007	63,4508
2008	169,523
-0,1 209	65,4003
2010	147,887
-1,1 Total	386,057

This table shows various statistics for ROA for each of the 5 levels of ANNEE. The one-way analysis of variance is

primarily intended to compare the means of the different levels, listed here under the Average column. Select Means Plot

The ANOVA table decomposes the variance of ROA into two components: a between-group component and a within-group component. The F-ratio, which in this case equals 11,9622, is a ratio of the between-group estimate to the within-group estimate. Since the P-value of the F-test is less than 0,05, there is a statistically significant difference between the mean ROA from one level of ANNEE to another at the 95,0% confidence level. To determine which means are significantly different from which others, select Multiple Range Tests from the list of Tabular Options.

			Stnd. error
0,01 ANNEE	Count	Mean	(pooled s)	Lower limit	Upper limit
2006 0,01	568	-0,013372	0,0106882	-0,0258033	-0,000940728
2007	568	0,045657	0,0106882	0,0332257	0,0580883
2008 003	568	0,0677319	0,0106882	0,0553006	0,0801632
2009	568	-0,0042941 2007	0,106882 2008 2	-0,0167254 2010	0,00813719
2010	568	0,0572362	0,0106882 NNEE	0,0448049	0,0696675
Total	2840	0,0305918

This table shows the mean ROA for each level of ANNEE. It also shows the standard error of each mean, which is a measure of its sampling variability. The standard error is formed by dividing the pooled standard deviation by the square root of the number of observations at each level. The table also displays an interval around each mean. The intervals currently displayed are based on Fisher's least significant difference (LSD) procedure. They are constructed in such a way that if two means are the same, their intervals will overlap 95,0% of the time. You can display the intervals graphically by selecting Means Plot from the list of Graphical Options. In the Multiple Range Tests, these intervals are used to determine which means are significantly different from which others.

ANNEE	Count		Mean		Homogeneous Groups
2006	568		-0,013372		X
2009	568		-0,0042941		X
2007	568		0,045657		X
2010	568		0,0572362		X
2008	568		Box- 0,0677319		X h X
Contrast		Sig.		Difference		+/- Limits
2006 - 2007 2006		*		-0,059029		0,0248626
2006 - 2008		*		-0,0811039		0,0248626
2006 - 2009				-0,00907792		0,0248626
2007 2006 - 2010		*		-0,0706082		0,0248626
2007 - 2008				-0,0220748		0,0248626
2008 2007 - 2009		*		0,0499511		0,0248626
2007 - 2010				-0,0115791		0,0248626
28 - 2009 2009		*		0,072026		0,0248626
2008 - 2010				0,0104957		0,0248626
2009 - 2010 2010		*		-0,0615303		0,0248626

	Test		P-Value
Levene's	2,58661		0,0352025
Comparison		Sigma1		Sigma2	F-Ratio	P-Value
2006 / 2007		0,223258		0,198647	1,26314	0,0055
2006 / 2008		0,223258		0,328621	0,461557	0,0000
2006 / 2009		0,223258		0,204192	1,19546	0,0337
2006 / 2010		0,223258		0,292304	0,583375	0,0000
2007 / 2008		0,198647		0,328621	0,365405	0,0000
2007 / 2009		0,198647		0,204192	0,946424	0,5123
2007 / 2010		0,198647		0,292304	0,461846	0,0000
2008 / 2009		0,328621		0,204192	2,59007	0,0000
2008 / 2010		0,328621		0,292304	1,26393	0,0054
2009 / 2010		0,204192		0,292304	0,48799	0,0000

The statistic displayed in this table tests the null hypothesis that the standard deviations of ROA within each of the 5 levels of

ANNEE is the same. Of particular interest is the P-value. Since the the P-value is less than 0,05, there is a statistically significant difference amongst the standard deviations at the 95,0% confidence level. This violates one of the important

assumptions underlying the analysis of variance and will invalidate most of the standard statistical tests.

The table also shows a comparison of the standard deviations for each pair of samples. P-Values below 0.05, of which there are 9, indicate a statistically significant difference between the two sigmas at the 5% significance level.

The Kruskal-Wallis test tests the null hypothesis that the medians of ROA within each of the 5 levels of ANNEE are the same. The data from all the levels is first combined and ranked from smallest to largest. The average rank is then computed for the data at each level. Since the P-value is less than 0,05, there is a statistically significant difference amongst the medians at the 95,0% confidence level. To determine which medians are significantly different from which others, select Box-and-Whisker Plot from the list of Graphical Options and select the median notch option.

ANNEE	Sample Size	n<=	n>	Median	90,0% lower CL	90,0% upper CL
59 2006	568	339	229	0,00945275	-0,000555481	0,0218933
2007	568	242	326	0,0570648	0,0460463	0,0689174
2008	568	248	320	0,0528593	0,0436125	0,0621654
39 2009	568	327	241	0,0121847	0,000872383	0,0184358
2010	568	264	304	0,044225	0,0338877	0,0561194

Mood's median test tests the hypothesis that the medians of all 5 samples are equal. It does so by counting the number of

observations in each sample on either side of the grand median, which equals 0,034619. Since the P-value for the chi-square

test is less than 0,1, the medians of the samples are significantly different at the 90,0% confidence level. Also included (if

available) are 90,0% confidence intervals for each median based on the order statistics of each sample.

Groups	Count	Average	Standard deviation	Coeff. of variation	Minimum	Maximum	Range	Stnd. skewness
1	10	-0,0069795	0,262229	-3757,13%	-0,484615	0,252975	0,73759	-1,14121
2	48	0,142481	0,180874	126,946%	-0,412968	0,474718	0,887686	-2,98642
3	43	0,123011	0,208889	169,813%	-0,37387	0,642566	1,01644	0,773958
4	28	0,0362009	0,195441	539,878%	-0,268354	0,67475	0,943103	2,75521
5	12	0,0452324	0,0883501	195,325%	-0,0649112	0,241946	0,306857	1,52055
Total	141	0,0965617	0,197818	204,861%	-0,484615	0,67475	1,15936	-0,42451
Groups	Stnd. kurtosis
1	-0,367154
2	2,91906
3	0,924502
4	3,11992
5	0,672821
Total	2,31948

Groups	Sample Size		n<=		n>		Median		90,0% lower CL		90,0% upper CL
1	10		6		4		0,0749859		-0,379136		0,23285
2		48		17		31		0,162987		0,111378		0,214087
3		43		20		23		0,121517		0,0528025		0,196325
4		28		19		9		0,0175846		-0,083556		0,106835
5		12		9		3		0,0220671		-0,0257561		0,128591

Groups	Count	Average	Standard deviation	Coeff. of variation	Minimum	Maximum	Range	Stnd. skewness
1	10	-0,17879	1,37918	-771,394%	-3,68929	1,89707	5,58636	-2,33501
2	42	0,11621	0,185857	159,932%	-0,376288	0,596295	0,972584	0,502
3	39	0,0630057	0,137776	218,672%	-0,181477	0,29342	0,474897	-0,76982
4	36	0,0100035	0,175893	1758,32%	-0,318248	0,462527	0,780774	2,18408
5	14	-0,0197315	0,0917217	-464,848%	-0,213078	0,123559	0,336637	-0,648606
Total	141	0,0399575	0,389822	975,59%	-3,68929	1,89707	5,58636	-26,7144
Groups	Stnd. kurtosis
1	3,83774
2	1,17713
3	-1,46433
4	0,88851
5	0,0493783
Total	153,54

Test statistic = 13,3289 P-Value = 0,00977558 Mood's Median Test for ROE by Groups

Groups	Sample Size	n<=	n>	Median	90,0% lower CL	90,0% upper CL
1	10	5	5	0,0550638	-0,618245	0,387143
2	42	13	29	0,099151	0,0719154	0,155127
3	39	16	23	0,0898162	0,0108911	0,150199
4	36	26	10	-0,00936615	-0,0647502	0,0233567
5	14	11	3	-0,0151896	-0,103506	0,0445331

Groups	Count	Average	Standard deviation	Coeff. of variation	Minimum	Maximum	Range	Stnd. skewness
1	6	0,075066	2,21034	2944,52%	-3,267	3,703	6,97	0,288721
2	33	0,106858	0,284379	266,127%	-1,04792	0,891704	1,93962	-3,53472
3	48	0,103257	0,150159	145,423%	-0,177508	0,584271	0,761778	2,14386
4	40	0,0920779	0,125932	136,767%	-0,156094	0,329752	0,485846	-0,006572
5	14	0,121409	0,358122	294,972%	-0,115502	1,34073	1,45623	5,31902
Total	141	0,101531	0,4658	458,776%	-3,267	3,703	6,97	3,27998
Groups	Stnd. kurtosis
1	1,24469
2	10,951
3	1,51367
4	-0,748698
5	9,66003
Total	108,366

Groups	Sample Size	n<=	n>	Median	90,0% lower CL	90,0% upper CL
1	6	3	3	0,0677642
2	33	11	22	0,125305	0,0811245	0,185447
3	48	25	23	0,0847578	0,0506366	0,131847
4	40	22	18	0,0781741	0,0264757	0,146814
5	14	10	4	0,0356748	-0,0276925	0,118881

Groups	Count	Average	Standard deviation	Coeff. of variation	Minimum	Maximum	Range	Stnd. skewness
1	7	-0,20666	1,02638	-496,653%	-2,52135	0,386503	2,90785	-2,79018
2	34	0,177157	0,250627	141,472%	-0,362507	1,06101	1,42352	3,11097
3	49	0,0841043	0,146422	174,096%	-0,481832	0,37319	0,855022	-3,64391
4	38	0,0371798	0,147275	396,114%	-0,270465	0,39789	0,668355	0,0203899
5	13	-0,0100111	0,0819924	-819,015%	-0,105939	0,165918	0,271857	1,15795
Total	141	0,0707839	0,28493	402,535%	-2,52135	1,06101	3,58236	-24,7291
Groups	Stnd. kurtosis
1	3,6496
2	5,01567
3	4,98283
4	0,0175206
5	0,00589768
Total	120,25

Groups	Sample Size	n<=	n>	Median	90,0% lower CL	90,0% upper CL
1	7	2	5	0,1569
2	34	10	24	0,140842	0,102307	0,193956
3	49	23	26	0,120696	0,0469401	0,164298
4	38	24	14	0,0287984	-0,0153515	0,119707
5	13	12	1	-0,0275842	-0,0825783	0,0486049

Groups	Count	Average	Standard deviation	Coeff. of variation	Minimum	Maximum	Range	Stnd. skewness
1	8	-0,143292	0,641699	-447,826%	-1,38481	0,676675	2,06148	-1,25734
2	35	0,0225383	0,240142	1065,48%	-1,09395	0,276005	1,36996	-7,54682
3	50	0,0475295	0,194926	410,116%	-0,699781	0,443904	1,14368	-3,02011
4	34	0,00565164	0,126919	2245,71%	-0,254159	0,257111	0,511269	0,526975
5	14	0,0271876	0,10212	375,611%	-0,0965745	0,303625	0,4002	2,70992
Total	141	0,0183813	0,233466	1270,13%	-1,38481	0,676675	2,06148	-12,3857
Groups	Stnd. kurtosis
1	0,679283
2	16,2943
3	4,88459
4	-0,227566
5	2,80584
Total	31,5881

Groups	Sample Size	n<=	n>	Median	90,0% lower CL	90,0% upper CL
1	8	5	3	-0,000383141	-1,3004	0,615859
2	35	14	21	0,0600382	-0,0171497	0,142537
3	50	20	30	0,0771671	0,00399078	0,126858
4	34	22	12	-0,0159893	-0,0583432	0,0413279
5	14	10	4	0,0124171	-0,0434617	0,0456193

Groups	Count	Average	Standard deviation	Coeff. of variation	Minimum	Maximum	Range	Stnd. skewness
1	7	-0,0824401	0,59067	-716,484%	-1,40816	0,248772	1,65693	-2,73727
2	6	-0,133896	0,230839	-172,402%	-0,567189	0,0267818	0,593971	-1,70149
3	7	-0,100753	0,239325	-237,537%	-0,526313	0,16275	0,689062	-1,00633
4	16	-0,158071	0,283015	-179,042%	-1,0385	0,149503	1,188	-3,41863
5	9	-0,138262	0,411863	-297,886%	-1,21844	0,131187	1,34963	-3,43557
Total	45	-0,130205	0,34697	-266,48%	-1,40816	0,248772	1,65693	-6,25348
Groups	Stnd. kurtosis
1	3,53339
2	1,45072
3	0,225223
4	4,876
5	5,01709
Total	7,65586

Groups	Sample Size	n<=	n>	Median	90,0% lower CL	90,0% upper CL
1	7	1	6	0,149788
2	6	3	3	-0,0463576
3	7	4	3	-0,0429657
4	16	10	6	-0,0797749	-0,286345	0,0373879
5	9	5	4	-0,0364259	-0,719997	0,122984

Groups	Count		Average		Standard deviation		Coeff. of variation		Minimum		Maximum		Range		Stnd. skewness
1	5		0,350147		0,452136		129,127%		-0,0783535		0,965862		1,04422		0,501517
2	4		0,0319552		0,274144		857,902%		-0,374624		0,203506		0,57813		-1,52757
3	10		-0,00885641		0,230547		-2603,16%		-0,436945		0,272862		0,709807		-1,60867
4	14		-0,0810269		0,28336		-349,711%		-1,03602		0,134314		1,17033		-5,12914
5		12		-0,119093		0,347957		-292,173%		-1,08706		0,167536		1,25459		-3,22219
Total		45		-0,0171888		0,329044		-1914,29%		-1,08706		0,965862		2,05292		-2,20465
Groups		Stnd. kurtosis
1		-0,792416
2		1,43611
3		0,476563
4		9,17569
5		3,97617
Total		6,68529

Groups	Sample Size	n<=	n>	Median	90,0% lower CL	90,0% upper CL
1	5	2	3	0,250584
2	4	1	3	0,149469
3	10	3	7	0,0563751	-0,408074	0,18652
4	14	10	4	-0,0117706	-0,0781489	0,0366616
5	12	7	5	0,0195701	-0,364224	0,0845263

Groups	Count	Average	Standard deviation	Coeff. of variation	Minimum	Maximum	Range	Stnd. skewness
1	3	-0,137126	0,268858	-196,066%	-0,442407	0,0643751	0,506782	-1,04524
2	6	0,200511	0,0814852	40,6389%	0,124808	0,341432	0,216623	1,16574
3	12	0,193629	0,925176	477,808%	-0,481933	3,053	3,53493	4,4236
4	10	-0,0391609	0,0765327	-195,431%	-0,166521	0,0709146	0,237435	-0,429557
5	14	-0,233825	0,391322	-167,357%	-1,13931	0,081154	1,22046	-2,50106
Total	45	-0,0122204	0,546856	-4474,94%	-1,13931	3,053	4,19231	10,3766
Groups	Stnd. kurtosis
1
2	0,433379
3	7,37526
4	-0,621184
5	1,23119
Total	31,9128

Groups	Sample Size	n<=	n>	Median	90,0% lower CL	90,0% upper CL
1	3	2	1	-0,0333468
2	6	0	6	0,177259
3	12	5	7	0,0251912	-0,299153	0,149681
4	10	6	4	-0,0288041	-0,128674	0,0316859
5	14	10	4	-0,0589653	-0,468499	0,0204429

Groups	Count	Average	Standard deviation	Coeff. of variation	Minimum	Maximum	Range	Stnd. skewness
1	7	0,0674309	0,133599	198,127%	-0,0885807	0,339021	0,427602	1,678
2	6	-0,00554691	0,302544	-5454,28%	-0,536908	0,316999	0,853907	-1,1561
3	7	0,0118722	0,235413	1982,89%	-0,321133	0,434852	0,755984	0,734801
4	13	-0,0072047	0,257595	-3575,37%	-0,37133	0,729829	1,10116	2,94153
5	12	-0,204717	0,385528	-188,322%	-1,13015	0,0444478	1,1746	-2,77536
Total	45	-0,0450762	0,292856	-649,69%	-1,13015	0,729829	1,85998	-3,41959
Groups	Stnd. kurtosis
1	1,87505
2	0,748515
3	0,758146
4	4,36867
5	1,95883
Total	7,26813

Groups	Sample Size	n<=	n>	Median	90,0% lower CL	90,0% upper CL
1	7	1	6	0,0295297
2	6	3	3	0,0492473
3	7	4	3	-0,0559836
4	13	8	5	-0,0533762	-0,147412	0,0818418
5	12	7	5	-0,0593088	-0,518098	0,0304725

Groups	Count	Average	Standard deviation	Coeff. of variation	Minimum	Maximum	Range	Stnd. skewness
1	5	-0,0655322	0,312964	-477,572%	-0,608197	0,156568	0,764764	-1,74174
2	8	0,0281413	0,327645	1164,28%	-0,58715	0,465494	1,05264	-0,979147
3	7	0,058023	0,17472	301,122%	-0,122793	0,384158	0,506951	1,1234
4	15	0,033406	0,200963	601,577%	-0,152073	0,678656	0,830729	4,12163
5	10	-0,111302	0,656989	-590,277%	-1,49246	1,18739	2,67985	-0,314283
Total	45	-0,00685106	0,368239	-5374,92%	-1,49246	1,18739	2,67985	-2,12124
Groups	Stnd. kurtosis
1	1,74629
2	0,476915
3	0,634647
4	6,42088
5	2,12482
Total	10,2089

Groups	Sample Size	n<=	n>	Median	90,0% lower CL	90,0% upper CL
1	5	2	3	0,0190313
2	8	3	5	0,0835481	-0,544271	0,440487
3	7	3	4	0,094098
4	15	9	6	-0,00436701	-0,0863766	0,091969
5	10	6	4	-0,00584826	-0,622008	0,214118

One-Way ANOVA - ROE by Groups (ANNEE=2006&SPECIALISATION="Grandes cultures")

Groups	Count	Average	Standard deviation	Coeff. of variation	Minimum	Maximum	Range	Stnd. skewness
1	26	0,179213	1,2686	707,873%	-2,7905	3,7011	6,4916	1,73148
2	25	-0,0607359	0,608729	-1002,26%	-1,40621	1,22074	2,62696	-0,970494
3	28	-0,229362	0,671868	-292,929%	-3,25	0,240467	3,49047	-7,95841
4	28	-0,0463455	0,30485	-657,778%	-0,85709	0,622275	1,47936	-0,288213
5	19	0,00180915	0,119182	6587,76%	-0,343083	0,196914	0,539997	-1,78401
Total	126	-0,0360658	0,728705	-2020,49%	-3,25	3,7011	6,9511	2,59487
Groups	Stnd. kurtosis
1	3,16843
2	0,267322
3	17,0465
4	1,31149
5	2,74246
Total	26,5316

Groups	Sample Size	n<=	n>	Median	90,0% lower CL	90,0% upper CL
1	26	9	17	0,127062	-0,323749	0,515738
2	25	11	14	0,0162612	-0,214373	0,217584
3	28	16	12	-0,0916287	-0,169954	0,051114
4	28	17	11	-0,0583804	-0,168101	0,083301
5	19	10	9	-0,00722626	-0,0502325	0,077573

One-Way ANOVA - ROE by Groups (ANNEE=2007&SPECIALISATION="Grandes cultures")

Groups	Count		Average		Standard deviation		Coeff. of variation		Minimum		Maximum		Range		Stnd. skewness
1	17		0,313318		1,35468		432,366%		-2,42969		3,31882		5,74851		-0,0972392
2	20		0,387628		0,374057		96,4989%		-0,302679		1,03553		1,33821		0,0513675
3	28		0,224728		0,341117		151,791%		-0,76952		0,759663		1,52918		-2,12864
4	30		0,0938323		0,434063		462,594%		-1,607		0,727135		2,33414		-4,80535
5		31		0,144244		0,188682		130,808%		-0,161782		0,682598		0,84438		1,8683
Total		126		0,211571		0,58687		277,387%		-2,42969		3,31882		5,74851		0,152769
Groups		Stnd. kurtosis
1		0,947297
2		-0,338419
3		1,27225
4		8,34027
5		1,03053
Total		24,1497

Groups	Sample Size	n<=	n>	Median	90,0% lower CL	90,0% upper CL
1	17	9	8	0,212239	-0,312194	1,10676
2	20	4	16	0,349002	0,227342	0,615027
3	28	11	17	0,324976	0,0618571	0,425479
4	30	18	12	0,159329	0,0492383	0,243204
5	31	21	10	0,110928	0,0475332	0,22415

One-Way ANOVA - ROE by Groups (ANNEE=2008&SPECIALISATION="Grandes cultures")

Groups	Count	Average	Standard deviation	Coeff. of variation	Minimum	Maximum	Range	Stnd. skewness
1	11	-0,0327873	1,60727	-4902,11%	-2,50663	3,0111	5,51773	0,0378886
2	19	-0,0564867	0,93133	-1648,76%	-2,69078	0,844573	3,53535	-3,89846
3	31	0,140444	0,397205	282,82%	-0,854586	0,968828	1,82341	-1,36906
4	36	0,173722	0,249045	143,359%	-0,480456	0,918507	1,39896	-0,402512
5	29	0,270691	0,585796	216,408%	-0,181367	3,181	3,36237	10,2369
Total	126	0,13511	0,690238	510,871%	-2,69078	3,181	5,87178	-2,50975
Groups	Stnd. kurtosis
1	0,226125
2	3,87705
3	0,740814
4	3,45077
5	25,9989
Total	23,3753

Groups	Sample Size	n<=	n>	Median	90,0% lower CL	90,0% upper CL
1	11	6	5	0,106996	-2,13698	1,11392
2	19	9	10	0,222666	-0,0259652	0,468206
3	31	13	18	0,19585	0,00799357	0,346097
4	36	18	18	0,157943	0,121049	0,248522
5	29	17	12	0,140075	0,0642507	0,268611

One-Way ANOVA - ROE by Groups (ANNEE=2009&SPECIALISATION="Grandes cultures")

Groups	Count	Average	Standard deviation	Coeff. of variation	Minimum	Maximum	Range	Stnd. skewness
1	16	-0,125603	0,867811	-690,918%	-3,0875	0,623308	3,71081	-4,80476
2	30	-0,127775	0,699474	-547,427%	-3,04993	1,08441	4,13434	-5,23813
3	24	-0,169858	0,567433	-334,063%	-2,0637	0,486368	2,55007	-3,81864
4	29	0,00564268	0,177867	3152,17%	-0,226196	0,608902	0,835098	4,06886
5	27	-0,0230683	0,343156	-1487,57%	-1,23032	0,707247	1,93757	-2,58084
Total	126	-0,0823706	0,547235	-664,357%	-3,0875	1,08441	4,17191	-13,5848
Groups	Stnd. kurtosis
1	8,1581
2	11,4772
3	4,52777
4	4,66555
5	5,88206
Total	32,3185

Groups	Sample Size	n<=	n>	Median	90,0% lower CL	90,0% upper CL
1	16	6	10	0,0520806	-0,166633	0,346845
2	30	19	11	-0,0973981	-0,308864	0,0792782
3	24	13	11	-0,0574204	-0,193187	0,139505
4	29	12	17	-0,0339269	-0,0779227	0,0381452
5	27	13	14	-0,0370192	-0,128781	0,0722118

One-Way ANOVA - ROE by Groups (ANNEE=2010&SPECIALISATION="Grandes cultures")

Groups	Count	Average	Standard deviation	Coeff. of variation	Minimum	Maximum	Range	Stnd. skewness
1	23	0,365149	1,55797	426,666%	-3,43153	3,04051	6,47204	-1,30814
2	28	-0,0165585	0,803741	-4853,95%	-2,79716	1,43732	4,23449	-4,06086
3	27	0,0836475	0,442804	529,369%	-0,876775	1,12613	2,0029	0,527303
4	22	0,165111	0,778352	471,412%	-0,764529	2,94949	3,71402	4,66934
5	26	0,0409212	0,394668	964,46%	-0,51794	1,15891	1,67686	3,03462
Total	126	0,118172	0,870417	736,569%	-3,43153	3,04051	6,47204	-0,99077
Groups	Stnd. kurtosis
1	1,41263
2	5,76644
3	1,80958
4	7,20079
5	2,1899
Total	13,7459

Groups	Sample Size	n<=	n>	Median	90,0% lower CL	90,0% upper CL
1	23	7	16	0,320624	0,0355358	0,546584
2	28	12	16	0,10634	-0,102615	0,378908
3	27	12	15	0,0793623	-0,0679452	0,247933
4	22	15	7	-0,0154434	-0,147291	0,088839
5	26	17	9	-0,0381151	-0,184124	0,10304

Groups	Count	Average	Standard deviation	Coeff. of variation	Minimum	Maximum	Range	Stnd. skewness
1	8	-0,0276876	0,142156	-513,428%	-0,286952	0,158607	0,445559	-0,534166
2	35	0,0286671	0,0880925	307,295%	-0,264655	0,154997	0,419652	-2,50623
3	50	0,0389244	0,121565	312,31%	-0,456612	0,275404	0,732016	-3,21185
4	34	0,00725538	0,0943411	1300,29%	-0,17114	0,224912	0,396052	1,06473
5	14	0,0252457	0,0962117	381,101%	-0,0883339	0,303035	0,391369	3,15877
Total	141	0,0236042	0,106445	450,959%	-0,456612	0,303035	0,759647	-2,64697
Groups	Stnd. kurtosis
1	0,348903
2	2,46083
3	6,51958
4	-0,0314555
5	4,01228
Total	6,72696

Groups	Sample Size	n<=	n>	Median	90,0% lower CL	90,0% upper CL
1	8	6	2	-0,0301937	-0,263351	0,155944
2	35	15	20	0,0256582	-0,00587675	0,0856697
3	50	20	30	0,0500655	0,00212655	0,0763213
4	34	21	13	-0,0110551	-0,0471129	0,0310625
5	14	9	5	0,0102684	-0,0378927	0,0391715

Groups	Count		Average	Standard deviation	Coeff. of variation	Minimum	Maximum	Range	Stnd. skewness
1	7		0,0431667	0,117686	272,63%	-0,206161	0,141439	0,3476	-2,13626
2	34		0,091015	0,0987241	108,47%	-0,089323	0,353563	0,442886	2,18268
3	49		0,0587761	0,0951325	161,856%	-0,250523	0,336835	0,587358	-0,867226
4	38		0,0305541	0,107431	351,609%	-0,209863	0,300593	0,510456	0,346439
5	13		-0,00667726	0,0742954	-1112,66%	-0,0941891	0,158182	0,252371	1,30607
Total	141		0,0521344	0,101742	195,153%	-0,250523	0,353563	0,604086	0,494216
Groups		Stnd. kurtosis
1		2,35332
2		1,70826
3		3,09405
4		0,132081
5		0,297842
Total		2,51387

Groups	Sample Size	n<=	n>	Median	90,0% lower CL	90,0% upper CL
1	7	3	4	0,0682113
2	34	11	23	0,0789014	0,0562679	0,103172
3	49	22	27	0,0688806	0,0244087	0,0981639
4	38	24	14	0,0202312	-0,0107817	0,0908195
5	13	11	2	-0,0253725	-0,069592	0,0417187

Groups	Count	Average	Standard deviation	Coeff. of variation	Minimum	Maximum	Range	Stnd. skewness
1	6	-0,13581	0,22382	-164,803%	-0,458807	0,0627273	0,521534	-0,719548
2	33	0,061246	0,11112	181,432%	-0,315704	0,356522	0,672226	-1,49312
3	48	0,0709067	0,102503	144,56%	-0,1078	0,38364	0,49144	2,8109
4	40	0,0697293	0,0930542	133,451%	-0,106346	0,250123	0,356469	0,516
5	14	0,114182	0,339191	297,062%	-0,0959154	1,27192	1,36783	5,36768
Total	141	0,063812	0,152396	238,82%	-0,458807	1,27192	1,73072	15,6065
Groups	Stnd. kurtosis
1	-0,816362
2	4,77391
3	1,78309
4	-0,706596
5	9,78065
Total	69,5517

Groups	Sample Size	n<=	n>	Median	90,0% lower CL	90,0% upper CL
1	6	5	1	-0,0544237
2	33	14	19	0,0606138	0,0374828	0,106412
3	48	24	24	0,0527917	0,024406	0,0916713
4	40	20	20	0,0551586	0,0217964	0,10954
5	14	8	6	0,0311118	-0,0237862	0,107481

Groups	Count	Average	Standard deviation	Coeff. of variation	Minimum	Maximum	Range	Stnd. skewness
1	10	-0,0530749	0,174992	-329,708%	-0,464359	0,116707	0,581066	-1,98846
2	42	0,0624122	0,0839266	134,471%	-0,101321	0,31074	0,41206	1,26242
3	39	0,0469517	0,0859704	183,104%	-0,0800366	0,221199	0,301236	0,211634
4	36	0,0117456	0,132892	1131,42%	-0,194892	0,350478	0,54537	2,56668
5	14	-0,0183905	0,0860311	-467,801%	-0,20497	0,115921	0,320892	-0,71293
Total	141	0,0289862	0,111116	383,342%	-0,464359	0,350478	0,814837	-1,10943
Groups	Stnd. kurtosis
1	1,81784
2	0,905741
3	-1,26116
4	0,830834
5	0,270994
Total	5,91865

Groups	Sample Size	n<=	n>	Median	90,0% lower CL	90,0% upper CL
1	10	6	4	-0,0185376	-0,202246	0,112088
2	42	14	28	0,0601934	0,0371256	0,0825121
3	39	16	23	0,0513118	0,00540596	0,0953483
4	36	26	10	-0,00644387	-0,0427011	0,0156736
5	14	9	5	-0,0137695	-0,0925586	0,0372823

Groups	Count	Average	Standard deviation	Coeff. of variation	Minimum	Maximum	Range	Stnd. skewness
1	10	0,0334214	0,0841506	251,786%	-0,112091	0,130723	0,242815	-0,717725
2	48	0,0807499	0,0909653	112,651%	-0,197911	0,323128	0,521039	-1,2612
3	43	0,0720914	0,112581	156,164%	-0,150311	0,328919	0,47923	0,716939
4	28	0,0258738	0,139871	540,588%	-0,206584	0,454306	0,66089	2,57093
5	12	0,041393	0,0791111	191,122%	-0,0596758	0,211972	0,271648	1,42801
Total	141	0,0605058	0,108654	179,575%	-0,206584	0,454306	0,66089	1,67491
Groups	Stnd. kurtosis
1	-0,512573
2	2,28223
3	-0,339197
4	2,46287
5	0,44387
Total	2,09408

Groups	Sample Size	n<=	n>	Median	90,0% lower CL	90,0% upper CL
1	10	5	5	0,0441637	-0,0820523	0,123533
2	48	19	29	0,0855747	0,0591239	0,11358
3	43	21	22	0,0697366	0,0250791	0,111734
4	28	18	10	0,0125945	-0,0636708	0,0666888
5	12	8	4	0,019074	-0,0212725	0,120007

Groups	Count	Average	Standard deviation	Coeff. of variation	Minimum	Maximum	Range	Stnd. skewness
1	7	-0,105765	0,453978	-429,234%	-1,13039	0,140396	1,27079	-2,80058
2	6	-0,0453944	0,078609	-173,169%	-0,177484	0,0229301	0,200414	-1,00598
3	7	-0,0473511	0,111711	-235,92%	-0,212802	0,0912202	0,304023	-0,371181
4	16	-0,111976	0,203313	-181,569%	-0,736663	0,11937	0,856032	-3,26129
5	9	-0,112163	0,336903	-300,368%	-0,993887	0,114128	1,10801	-3,40561
Total	45	-0,0921169	0,257116	-279,119%	-1,13039	0,140396	1,27079	-7,86363
Groups	Stnd. kurtosis
1	3,66696
2	0,113159
3	-0,655198
4	4,54897
5	4,95777
Total	12,0237

Groups	Sample Size	n<=	n>	Median	90,0% lower CL	90,0% upper CL
1	7	1	6	0,0516395
2	6	3	3	-0,0247187
3	7	3	4	-0,0179525
4	16	10	6	-0,0646852	-0,187257	0,028068
5	9	6	3	-0,0293466	-0,589257	0,107523

Groups	Count	Average	Standard deviation	Coeff. of variation	Minimum	Maximum	Range	Stnd. skewness
1	5	0,0746187	0,110029	147,455%	-0,0409659	0,217745	0,258711	0,141305
2	4	0,0162684	0,168061	1033,05%	-0,230272	0,138248	0,36852	-1,42673
3	10	0,00954246	0,130679	1369,44%	-0,222714	0,215079	0,437793	-0,964728
4	14	-0,0609528	0,218634	-358,694%	-0,79908	0,099552	0,898632	-5,15769
5	12	-0,0972349	0,295406	-303,806%	-0,922824	0,153846	1,07667	-3,2495
Total	45	-0,0330348	0,213964	-647,692%	-0,922824	0,217745	1,14057	-7,61542
Groups	Stnd. kurtosis
1	-0,824357
2	1,27345
3	0,42576
4	9,25855
5	4,09935
Total	12,5719

Groups	Sample Size	n<=	n>	Median	90,0% lower CL	90,0% upper CL
1	5	2	3	0,0907909
2	4	1	3	0,0785487
3	10	3	7	0,0371746	-0,201921	0,112828
4	14	10	4	-0,00844839	-0,0582445	0,0252546
5	12	7	5	0,0165747	-0,297092	0,0768496

Groups	Count	Average	Standard deviation	Coeff. of variation	Minimum	Maximum	Range	Stnd. skewness
1	3	-0,0681985	0,155976	-228,708%	-0,241266	0,0615116	0,302777	-0,81616
2	6	0,0936297	0,0367747	39,2768%	0,0291386	0,137466	0,108327	-1,05367
3	12	0,190185	0,765362	402,431%	-0,228707	2,5915	2,82021	4,68812
4	10	-0,027013	0,0593021	-219,532%	-0,128615	0,0663722	0,194987	-0,361463
5	14	-0,197005	0,329566	-167,288%	-0,941719	0,0759867	1,01771	-2,45214
Total	45	-0,00864	0,452028	-5231,81%	-0,941719	2,5915	3,53322	11,3001
Groups	Stnd. kurtosis
1
2	0,926372
3	7,97102
4	-0,321322
5	1,13895
Total	35,7779

Groups	Sample Size	n<=	n>	Median	90,0% lower CL	90,0% upper CL
1	3	2	1	-0,0248415
2	6	0	6	0,0991103
3	12	5	7	0,0165927	-0,183294	0,117365
4	10	6	4	-0,019054	-0,0979158	0,0233145
5	14	10	4	-0,0486992	-0,393363	0,0173552

Groups	Count	Average	Standard deviation	Coeff. of variation	Minimum	Maximum	Range	Stnd. skewness
1	7	0,0219456	0,051473	234,548%	-0,0646818	0,0906458	0,155328	-0,277781
2	6	0,0469839	0,122463	260,649%	-0,114153	0,186734	0,300888	-0,157948
3	7	0,0075369	0,128871	1709,86%	-0,171113	0,230123	0,401236	0,640108
4	13	-0,00644273	0,165268	-2565,19%	-0,223332	0,462336	0,685667	2,92308
5	12	-0,191032	0,366677	-191,946%	-1,05433	0,0394772	1,09381	-2,77349
Total	45	-0,0419523	0,232259	-553,625%	-1,05433	0,462336	1,51666	-7,34034
Groups	Stnd. kurtosis
1	0,277796
2	-1,10656
3	0,33783
4	4,0552
5	1,87133
Total	14,9154

Groups	Sample Size	n<=	n>	Median	90,0% lower CL	90,0% upper CL
1	7	1	6	0,0143045
2	6	2	4	0,0517442
3	7	4	3	-0,0355858
4	13	8	5	-0,0368864	-0,106846	0,0743713
5	12	8	4	-0,0506681	-0,489059	0,0293248

Groups	Count	Average	Standard deviation	Coeff. of variation	Minimum	Maximum	Range	Stnd. skewness
1	5	0,0150524	0,0949592	630,858%	-0,121916	0,107682	0,229598	-0,543456
2	8	0,0492815	0,133	269,878%	-0,13138	0,25763	0,38901	0,0269102
3	7	0,0338206	0,105709	312,558%	-0,0764125	0,221189	0,297601	0,818161
4	15	0,0330233	0,173375	525,008%	-0,115176	0,613893	0,729069	4,73907
5	10	-0,120915	0,592286	-489,838%	-1,41254	0,982268	2,39481	-0,732483
Total	45	-0,000167523	0,301365	-179895,%	-1,41254	0,982268	2,39481	-4,15775
Groups	Stnd. kurtosis
1	-0,285199
2	-0,375168
3	0,169954
4	8,06292
5	2,10848
Total	17,7737

Groups	Sample Size	n<=	n>	Median	90,0% lower CL	90,0% upper CL
1	5	2	3	0,0136008
2	8	3	5	0,0500813	-0,129276	0,244492
3	7	3	4	0,0634198
4	15	9	6	-0,00312077	-0,0566846	0,0635826
5	10	6	4	-0,00640764	-0,597499	0,188712

One-Way ANOVA - ROA by Groups (ANNEE=2006&SPECIALISATION="Grandes cultures")

Groups	Count	Average	Standard deviation	Coeff. of variation	Minimum	Maximum	Range	Stnd. skewness
1	26	-0,0210406	0,264087	-1255,13%	-0,717935	0,385031	1,10297	-1,9433
2	25	0,00810686	0,217864	2687,41%	-0,415281	0,488035	0,903316	-0,103519
3	28	-0,116418	0,412285	-354,143%	-2,05464	0,185067	2,23971	-8,91681
4	28	-0,031409	0,209427	-666,773%	-0,52968	0,400579	0,930259	0,126946
5	19	0,00295864	0,109342	3695,66%	-0,315782	0,194464	0,510247	-1,81182
Total	126	-0,0351374	0,270354	-769,419%	-2,05464	0,488035	2,54268	-16,0997
Groups	Stnd. kurtosis
1	1,11277
2	-0,0307004
3	20,8898
4	0,554087
5	2,91151
Total	55,7841

Groups	Sample Size	n<=	n>	Median	90,0% lower CL	90,0% upper CL
1	26	11	15	0,026324	-0,102639	0,0944078
2	25	11	14	0,00499102	-0,0805615	0,132682
3	28	16	12	-0,0494749	-0,0861457	0,038731
4	28	17	11	-0,0426167	-0,114584	0,0563211
5	19	8	11	-0,00581426	-0,04485	0,0720474

One-Way ANOVA - ROA by Groups (ANNEE=2007&SPECIALISATION="Grandes cultures")

Groups		Count		Average		Standard deviation		Coeff. of variation		Minimum		Maximum		Range		Stnd. skewness
1		17		-0,007065		0,34629		-4901,48%		-1,09085		0,420565		1,51142		-3,33089
2		20		0,153102		0,161233		105,311%		-0,146563		0,424526		0,571089		0,162655
3		28		0,142208		0,210249		147,846%		-0,374676		0,677594		1,05227		-0,432352
4	30		0,0629824		0,333997		530,302%		-1,35464		0,542145		1,89679		-6,01841
5	31		0,124657		0,159385		127,859%		-0,131722		0,550014		0,681736		1,66031
Total	126		0,100616		0,252418		250,873%		-1,35464		0,677594		2,03224		-10,5448
Groups	Stnd. kurtosis
1	4,71157
2	-0,605725
3	1,35227
4	12,217
5	0,561296
Total	26,2384

Groups	Sample Size	n<=	n>	Median	90,0% lower CL	90,0% upper CL
1	17	10	7	0,0704004	-0,154593	0,168393
2	20	11	9	0,101369	0,0783039	0,269983
3	28	9	19	0,188491	0,0397537	0,225054
4	30	16	14	0,101664	0,0372794	0,181625
5	31	17	14	0,0994869	0,0423933	0,189778

One-Way ANOVA - ROA by Groups (ANNEE=2008&SPECIALISATION="Grandes cultures")

Groups	Count	Average	Standard deviation	Coeff. of variation	Minimum	Maximum	Range	Stnd. skewness
1	11	0,132713	0,42949	323,624%	-0,372089	1,10397	1,47606	1,43867
2	19	-0,00170479	0,334047	-19594,6%	-1,11512	0,344811	1,45993	-4,38283
3	31	0,0933312	0,227565	243,825%	-0,452873	0,441864	0,894738	-1,51654
4	36	0,120531	0,176681	146,586%	-0,357934	0,7116	1,06953	0,39532
5	29	0,232773	0,487943	209,622%	-0,152033	2,64574	2,79777	10,0901
Total	126	0,122303	0,332737	272,058%	-1,11512	2,64574	3,76086	14,46
Groups	Stnd. kurtosis
1	1,03476
2	6,07373
3	-0,0995506
4	5,01875
5	25,4604
Total	63,1311

Groups	Sample Size	n<=	n>	Median	90,0% lower CL	90,0% upper CL
1	11	6	5	0,0986874	-0,315322	0,49457
2	19	12	7	0,0782904	-0,00894632	0,17541
3	31	15	16	0,112627	-0,000184656	0,246034
4	36	17	19	0,123432	0,0862352	0,170685
5	29	13	16	0,127359	0,0548796	0,224914

One-Way ANOVA - ROA by Groups (ANNEE=2009&SPECIALISATION="Grandes cultures")

Groups	Count	Average	Standard deviation	Coeff. of variation	Minimum	Maximum	Range	Stnd. skewness
1	16	0,011712	0,133853	1142,87%	-0,354306	0,234759	0,589065	-2,07989
2	30	-0,0478995	0,322939	-674,201%	-1,50563	0,400454	1,90608	-7,00551
3	24	-0,075446	0,303661	-402,488%	-1,08988	0,3173	1,40718	-3,70352
4	29	0,00888732	0,131824	1483,28%	-0,146988	0,436751	0,583739	4,25596
5	27	-0,0184289	0,289708	-1572,03%	-1,01514	0,555053	1,57019	-2,44534
Total	126	-0,0261916	0,256366	-978,807%	-1,50563	0,555053	2,06068	-10,8935
Groups	Stnd. kurtosis
1	2,45812
2	16,3957
3	4,61302
4	4,43736
5	5,06561
Total	26,7243

Groups	Sample Size	n<=	n>	Median	90,0% lower CL	90,0% upper CL
1	16	4	12	0,028231	-0,0257411	0,100183
2	30	19	11	-0,057458	-0,109656	0,0350027
3	24	12	12	-0,0297157	-0,100399	0,0758477
4	29	14	15	-0,0245652	-0,0545986	0,0302802
5	27	14	13	-0,0315569	-0,112985	0,0662393

One-Way ANOVA - ROA by Groups (ANNEE=2010&SPECIALISATION="Grandes cultures")

Groups	Count	Average	Standard deviation	Coeff. of variation	Minimum	Maximum	Range	Stnd. skewness
1	23	0,14274	0,225007	157,634%	-0,414526	0,629534	1,04406	-0,951261
2	28	0,0441773	0,235867	533,911%	-0,61943	0,448297	1,06773	-1,69751
3	27	0,0456094	0,239838	525,851%	-0,501173	0,67582	1,17699	0,174262
4	22	0,103995	0,495929	476,878%	-0,574792	1,77651	2,35131	4,02694
5	26	0,0271355	0,336543	1240,23%	-0,436735	0,900406	1,33714	2,7171
Total	126	0,0694037	0,314165	452,664%	-0,61943	1,77651	2,39595	7,10176
Groups	Stnd. kurtosis
1	1,41796
2	1,15054
3	2,04443
4	5,40591
5	1,53559
Total	15,4533

Groups	Sample Size	n<=	n>	Median	90,0% lower CL	90,0% upper CL
1	23	6	17	0,153682	0,0569592	0,230257
2	28	14	14	0,0352505	-0,0468007	0,207821
3	27	12	15	0,0373107	-0,0316293	0,116303
4	22	14	8	-0,012042	-0,110281	0,0668334
5	26	17	9	-0,0347476	-0,162994	0,0882766

The cost of capital has been estimated using the market value for the cost of debt. For some specialization like cattle farming, we do not dispose of enough farms to collect 15 recent interest rates.

Using the WACC methodology to improve the assessment of projects in the french farming industry. Empirical evidences from farm's results of Isère

Using the WACC Methodology to

Improve the Assessment of Projects

in the French Farming Industry

Empirical Evidences from Farm's Results of

Isère

Table of Content

Abstract

Abbreviations

Table of Illustrations

Introduction

1 Significance of the Research

1.1 The Farming Business in France

1.2 The Profitability Heavily Relies on Subsidies, Not on Investment Decision

1.3 A New Approach to Find

2 Literature Review

2.1 Risk

2.1.1 Risk in Agriculture

2.1.2 Risk Premium

2.2 The WACC Theory

2.2.1 The Importance of the WACC Theory.

2.2.2 The Betas for Micro-Capitalizations

2.3 Discount Rates and the CAPM Used in Agriculture

3 Research Methodology

3.1 Survey Construction

3.1.1 Questionnaire

3.1.2 Results Analysis

3.2 Historical Analysis of the Results of Farms in Isère

3.2.1 Data Collection

3.2.2 Data Processing

3.2.3 Research Questions

3.2.4 Data Analysis

3.3 Bond Yield Plus Risk Premium Model

3.3.2 Beta and Risk premium

3.4 WACC Estimation

4 Data Analysis

4.1 Results of the Survey

4.2 Calculation of the Historical Results Using the Data from CERFRANCE Isère

4.2.1 Normality of the Datasets

4.2.2 ROE

4.2.2.1 ROE Tested by Time and Specialization

4.2.2.2 ROE for Dairy Producers

4.2.2.3 ROE for Grain Producers

4.2.2.4 ROE for Cattle Ranching

4.2.2.5 ROE for diversified production

4.2.2.6 Results Summary for ROE by Groups of Leverage

4.2.3 ROA

4.2.3.1 ROA Tested by Time and Specialization

4.2.3.2 ROA for Dairy Producers

4.2.3.3 ROA for Grain Producers

4.2.3.4 ROA for Cattle Farming

4.2.3.5 Results Summary for ROA by Groups of Leverage

4.3 Calculation of the Capitalization Rate Using Bond Yields Plus Risk Premium

4.4 Approach of the WACC for Farming in Isère

4.4.1 Calculation of the Betas

4.4.2 Calculation of the WACC for the Different Specialization

5 Discussion

5.1 Survey Results

5.2 ROE and ROA Analysis the Farms of Isère

5.3 Capitalization Using the Bond Yield Plus Risk Premium

5.4 The WACC Methodology

6 Utilization of These Results

6.1 Managerial Implications

6.2 Agricultural Policies Implications

7 Limitations and Further Research Implications

7.1 Survey Construction

7.2 ROE and ROA Analysis

7.3 Bond Yield Plus risk Premium Model

7.4 The WACC Methodology

7.5 Further Research Implications

Conclusion

Bibliography

Appendices