Risk and economic sustainability of crop farming systems

AGRICULTURAL SYSTEMS Agricultural Systems 94 (2007) 541–552 www.elsevier.com/locate/agsy

Risk and economic sustainability of crop farming systems Gudbrand Lien a

a,b,*

, J. Brian Hardaker c, Ola Flaten

a

Norwegian Agricultural Economics Research Institute, P.O. Box 8024 Dep., No. 0030 Oslo, Norway b Eastern Norway Research Institute, Servicebox, No. 2626 Lillehammer, Norway c School of Economics, University of New England, Armidale, NSW 2351, Australia Received 16 May 2006; received in revised form 6 December 2006; accepted 26 January 2007

Abstract In economic terms, resilience in farming has to do with the capacity of a farm business to survive various risks and other shocks. Despite its importance, resilience has seldom been directly considered in evaluations of economic sustainability. A whole-farm stochastic simulation model over a 6-year planning horizon was used to analyse organic and conventional cropping systems using a model of a representative farm in Eastern Norway. The relative economic sustainability of alternative systems under changing assumptions about future technology and price regimes was examined in terms of financial survival to the end of the planning period. The same alternatives were also compared in terms of stochastic efficiency. To model the risk of business failure adequately there is a need to deal with the risk of bankruptcy, and a modification of traditional analysis was used for that purpose. The organic farming system was found to be somewhat less economically sustainable than the conventional system, especially if the organic price premiums and the organic area payments were to be phased out. The results illustrate possible conflicts between pursuit of risk efficiency and economic sustainability. The model developed could be used to support farmers’ choices between farming systems as well as to help policy makers develop more sharply targeted policies. Ó 2007 Elsevier Ltd. All rights reserved. Keywords: Sustainability; Resilience; Risk assessment; Whole-farm stochastic simulation; Stochastic efficiency

1. Introduction Although there is wide agreement that sustainability is a good thing, in agriculture and generally, the term has been given so many meanings that commentators have suggested that the concept is not definable (Pretty, 1995), meaningless (Beckerman, 1992), or at best is context specific (Pannell and Schilizzi, 1999, Zhen et al., 2005). For example, according to Rigby and Ca´ceres (2001), Jacobs (1995) recorded at least 386 definitions of sustainable development. Not surprisingly, therefore, there is a lack of agree-

* Corresponding author. Address: Norwegian Agricultural Economics Research Institute, P.O. Box 8024 Dep., No. 0030 Oslo, Norway. Tel.: +47 2236 7250; fax: +47 2236 7299. E-mail address: [email protected] (G. Lien).

0308-521X/$ - see front matter Ó 2007 Elsevier Ltd. All rights reserved. doi:10.1016/j.agsy.2007.01.006

ment on how to assess sustainability. Different analysts have selected a large number of indicators, with limited consistency across studies. In a farming systems context, a review of the literature also reveals use of a wide range of indicators. This diversity led Pannell and Glenn (2000) to propose a framework for economic evaluation to rationalise the selection of indicators in agriculture. In this paper, we focus on the economic sustainability of agricultural systems which, while not comprehensive, seem relevant to the decision problem of interest. We start from a suggestion by Conway (1985) that ‘sustainability is the ability of a system to maintain productivity in spite of a major disturbance, such as caused by intensive stress or a large perturbation’. Such a definition focuses on the resilience of the system, a concept introduced by Holling (1973). Applying this notion to the choice between alternative farming systems, we view economic sustainability as the ability of the system to continue into the future

542

G. Lien et al. / Agricultural Systems 94 (2007) 541–552

(Hansen and Jones, 1996; Kaine and Tozer, 2005). At the level of the individual farm, we take this to mean primarily that the farm business must remain financially viable while providing an acceptable livelihood for the farm family. Naturally, the ability to survive financially will be compromised if the farming system leads to the degradation of the farm resources, chiefly the land itself. There is a close link, therefore, between financial survival and the trend in total factor productivity, which was proposed by Lynam and Herdt (1989) as the criterion of sustainability most useful for guiding agricultural research. Economic sustainability, as we have chosen to define it, involves future outcomes that cannot be observed at the time that choices are made between farming systems of different types, such as organic versus conventional. As Hansen and Jones (1996) wrote: ‘Since sustainability deals with the future, it cannot be readily observed’. Conway (1994) noted the difficulty in assessing sustainability in terms of the persistence of a system because of the infeasibility of long-term experiments. It is clear that such assessments require modelling the performance of alternative management options into the future (Pandey and Hardaker, 1995; Hansen and Jones, 1996). Moreover, any such models must reflect the uncertainty about what the future might bring (e.g., Payraudeau and van der Werf, 2005). Indeed, the uncertainty about the future, and the possibility of shocks to the system from future adverse events, are the very reasons why it is important to explore the resilience of the system as a key aspect of economic sustainability. Evidently, an investigation of the economic sustainability of a farming system needs to model the stochastic and dynamic nature of the system. That implies that sustainability can be assessed in terms of the probability of persistence to some future moment in time. Moreover, although sustainability is usually argued to be about the long-term future, it is hard to model the inherent uncertainty far into the future because predictions about the distant future are too unreliable. Lynam and Herdt (1989) suggested a time horizon greater than 3–5 years but probably less than 20 for assessing sustainability. In this study, we have chosen to investigate economic sustainability to a relatively near time horizon of 6 years using a whole-farm model which allows the financial performance to be assessed. However, to compensate to some extent for the short time horizon, we can use stochastic simulation to examine each technology evaluated under a range of possible uncertain futures. Some 1.8% of the area of grain and potatoes crops in Norway was under organic management in 2005. The Norwegian government is keen to encourage the expansion of organic farming, partly to supply a developing market for organic produce, but also because of the additional benefits to the society expected from organic farming through less environmental harm, healthier and safer food, a contribution to the rural economy, improved animal welfare and reduced food surpluses (Ministry of Agriculture, 1999). Currently the Norwegian government aims that 15% of food both produced and consumed will be organic

by 2015. Increased organic crop production is perceived as crucially important for the further development of the organic sector because it will soon be required that 100% feed used in organic livestock systems shall be organically produced. Policy measures implemented in Norway to encourage farmers to convert to organic production include direct payments for organic farming, subsidised extension services and price premiums. It seems likely that organic and convention cropping systems will respond differently to variations in weather, implying different impacts on income risk for a farm. For example, restrictions on pesticide and fertiliser use may lead to increased production risk in organic farming compared with conventional farming. Additionally, smaller organic markets may mean greater price fluctuations. Most economic studies comparing cropping systems look exclusively at expected profitability measured by average net farm income (Roberts and Swinton, 1996). Yet expected profitability is an insufficient criterion as it ignores likely differences in the riskiness of net income between cropping systems. One way to assess and compare profit (in)stability is by using stochastic simulation. Mahoney et al. (2004), Smith et al. (2004) and Lien et al. (2006) all used stochastic simulation within a stochastic dominance framework to analyse income risk differences between conventional and organic crop systems. With price premiums included, North American studies indicated higher expected economic returns for organic than for conventional cropping systems (Mahoney et al., 2004; Smith et al., 2004). Smith et al. (2004) reported more variable returns in the organic rotations, while Mahoney et al. (2004) did not find returns in the organic strategies to be more variable than the conventional ones. The Norwegian study found higher but more variable returns in the organic than the conventional cropping system when area payments for organic farming and organic price premiums were included (Lien et al., 2006). As far as we know, the only study that empirically includes both the risk aspect, through stochastic simulation, and the economic sustainability aspect when comparing cropping systems is Hansen et al. (1997). They investigated the economic sustainability of a hillside farm in Colombia. Their results identified cropping system, area under cultivation, consumption requirements and crop prices as important determinants of economic sustainability. The specific objectives in our study are: (i) to compare the economic sustainability and risk efficiency of organic versus conventional cropping systems in Eastern Norway, (ii) to find out whether threats coming from outside the system affect economic sustainability positively or negatively, (iii) to ascertain whether economic sustainability and risk efficiency are complementary or conflicting gaols, and (iv) to take into account the treatment of the risk of bankruptcy, that has largely been ignored in earlier risk efficiency analysis in agriculture, but is essential to model the risk of business failure adequately.

G. Lien et al. / Agricultural Systems 94 (2007) 541–552

2. Material and methods 2.1. Quantifying economic sustainability We first expand on the framework described by Hansen and Jones (1996) for using stochastic simulation models to quantify the economic sustainability of a farming system. Let time to failure, TF, be represented with a cumulative probability distribution, F T F ðtÞ, where t is a time variable and T is a time horizon. For the time period (0, T), economic sustainability, S, is defined as SðT Þ ¼ 1  F T F ðT Þ;

ð1Þ

which is equivalent to the survival function in mortality studies. Following Hansen and Jones, economic sustainability can be estimated by the simulated relative frequency of surviving realisations, or ^ Þ ¼ nðT Þ ; SðT N

ð2Þ

where n(T) is number of non-failures at time T and N is number of iterations in the simulation model. Sensitivity analysis can be used to rank the impacts of different variables on the economic sustainability of the farming system. The steps in the analysis are as follows: (i) simulate the model for the base scenario and record; (ii) increase or decrease the investigated variable i by a small proportion p, e.g., 5%, simulate the model and record; (iii) repeat step (ii) for each variable tested for sensitivity; (iv) use equation (2) to calculate S^ for the base scenario as well as for all investigated changes in variables; and (v) calculate the relative sensitivity, si, for all investigated variables as "

S^i  S^0 si ¼ S^0

#, p;

ð3Þ

where S^0 and S^i are economic sustainability estimates for the base and alternative scenarios. Relative sensitivity, si, is interpreted as an elasticity indicating the percentage change in S^ for a 1% change in the investigated variable. System failure can be measured in financial terms. For example, Hamblin (1992) suggested that agriculture in a cash economy fails to sustain if production falls below the levels necessary for profitability. Lenders may impose threshold debt-to-equity ratios above which they will force foreclosure by recalling loans. Hansen and Jones (1996) used two criteria of failure: a debt-to-equity ratio exceeding 2.0 or a negative net present value (NPV) of future cash flows. We have chosen to identify failure when the farm owner’s equity falls below zero, indicating technical insolvency. While our criterion is less readily triggered than the debt-to-equity ratio of 2.0 used by Hansen and Jones (1996), we believe it is a reasonable approximation of the likely response of many lenders when a family farm faces hard times. Because it is not easy to model all the possible on- and off-farm strategies that a farm family can use to

543

avoid financial ruin, our model may somewhat overestimate the risk of failure, justifying a less stringent setting. The exact setting of the criterion of farm economic sustainability used could readily be tailored to the circumstances of an individual farm decision maker. The above economic sustainability criterion should not be the only criterion used to make a choice between farming systems. The measure focuses only on the lower tail of the distribution. This is also the case with the ‘value-at-risk’ (VaR) criterion, which has been widely used in finance. This risk measure is defined by a quantile of a distribution, and the 5% quantile is commonly called VaR (e.g., Manfredo and Leuthold, 1999). The use of VaR as a measure of risk has been criticised by, e.g., Artzner et al. (1999) and Szego¨ (2005). Considering the lower tail of the distribution may mean that risky prospects with the chance of bad outcomes are rejected, with too little consideration of up-side potential. This may be so even though the up-side of the distribution at one point of time can reduce the probability of being in the down-side distribution later, due to asset dynamics. More generally, however, focussing only on the down-side consequences implies a high degree of risk aversion which may not be justified in reality since, while the failure of a family farm business may be judged to be very undesirable, it does not imply the extinction of the family, the members of which can move on to another life, albeit perhaps a less preferred one. In other words, while the equity invested in the farm business may be lost, the value of wealth not at risk, such as the human capital of the family members, is not lost. 2.2. Comparing farming systems with stochastic efficiency It follows that to evaluate the relative risks of alternative farming systems appropriately we need to consider the whole range of outcomes, good and bad, and their associated probabilities, along with the relative preferences (utilities) of the decision maker(s) for those outcomes. To supplement the outlined economic sustainability criterion, risky outcomes are measured in terms of the probability distribution of current wealth from farming, defined as the net present value of the farm equity at the end of the planning horizon. To assess and compare the risk efficiency of the two farming systems we have used stochastic efficiency with respect to a function (SERF) (Hardaker et al., 2004b). For each risky alternative (in this study organic and conventional cropping systems) and for a chosen form of the utility function, U, the utility for wealth can be calculated depending on the degree of risk aversion, r, and the distribution of the wealth, w: U ðw; rÞ ¼

Z

U ðw; rÞf ðwÞdz 

N X

U ðwn ; rÞP ðwn Þ;

ð4Þ

n¼1

where U is evaluated for selected values of r within lower to upper range rL to rU. The second term represents the continuous case and the third term is the discrete

544

G. Lien et al. / Agricultural Systems 94 (2007) 541–552

approximation for computational purposes. P(wn) is the probability for iteration n in the Monte Carlo simulation. Partial ordering of alternatives by utility values is the same as partial ordering them by certainty equivalents (CEs). CE is defined as the sure sum with the same utility as the expected utility of a risky alternative (e.g., Keeney and Raiffa, 1976). We convert the utilities to CEs by taking the reverse of the utility function: CEðw; rÞ ¼ U 1 ðw; rÞ:

ð5Þ

CEs are readily interpreted because, unlike utility values, they are expressed in money terms. For a risk-averse decision maker (the normal case), the estimated CE is less than the expected money value (EMV). The difference between the EMV and the CE is the risk premium, reflecting the effect of the decision maker’s risk aversion (Hardaker et al., 2004b). The rule for SERF analysis for the given assumptions is that the efficient set contains only those alternatives that have the highest (or equal to highest) CE for some value of r in the relevant range. In SERF, to compute the CEs, any convenient form of the utility function can be used. However, because we assume that farmers prefer more wealth to less and are risk-averse, we are restricted to using a monotonic concave form, i.e., U 0 > 0 and U00 < 0. In this study, where the risky prospect can be large compared to the farmer’s total wealth, we have chosen to use a modified power utility function (Eq. (6)) with the plausible properties of constant relative risk aversion (CRRA) and decreasing absolute risk aversion (DARA). As part of the analysis of risk efficiency it is necessary to decide how to deal with bankruptcy. Venter (1983) has suggested that part of the wealth can reasonably be considered not at risk, since not all capital is resumed by creditors in the event of bankruptcy. Family or personal human capital remains and some other sorts of capital considered necessary for a reasonable existence, such as personal effects, are typically also left untouched, depending on local laws and customs. Support for the ‘‘wealth not at risk’’-view can be found in Quiggin (1992) who points out that the availability of bankruptcy has some similarities with insurance, since it removes the worst down-side risk. In the light of the above discussion, in the SERF analysis we have used a power utility function, modified to take account of wealth not at risk: ( ð1r Þ ~ > 0; ð1  rr Þ1 wF r ; if w U ðwÞ ¼ ; ð6Þ 1 ð1rr Þ ~ 60  ð1  rr Þ w ; if w where rr is defined as the assumed constant value of the relative risk aversion function with respect to wealth, rr(w), ~ is stowF is the wealth of a farmer staying in business, w chastic or risky wealth (represented as the net assets of  is wealth level not at risk. the farming business), and w Note that, with this formulation, a horizontal section of the utility function will imply risk seeking rather than

aversion for some risks which threaten financial viability. This behaviour can be seen to be rational, if not wholly responsible, when a business or a person is nearly bankrupt. In principle, the wealth not at risk incorporates the discounted value of future consumption after the planning horizon. In the case of a family farm, this should (roughly) represent the human capital of the family members ð wH Þ plus any domestic or business assets or personal effects protected under the law from resumption to meet debts in case ¼w H þ w  P , all measured in of bankruptcy ð wP Þ, i.e., w present values at time t = 0. The wealth of a farmer staying ~ þw  P since the value of in business is composed of wF ¼ w human capital for a continuing farm family is presumed to be reflecting in the returns from farming and hence in the present value of terminal equity of the farm business. While the same principles may apply in accounting for the risk of bankruptcy for a corporate business, there are additional complications, such as that most shareholders lose only a small proportion of their wealth while senior executives might face considerable losses of personal assets and future income. To avoid these further complications we make the reasonable assumption that the case farm is a family owned and operated business. 2.3. Whole-farm simulation model To implement the approaches proposed above, a wholefarm stochastic simulation model was developed to compare the economic sustainability and risk efficiency of organic versus conventional farming for a typical arable farm in Eastern Norway. The model used to evaluate the financial performance of the farm business is a recursive budgeting model over a 6-year time horizon using equations linking farm production activities, subsidy schemes, capital transactions, household consumption, financing arrangements and tax obligations. Stochastic features were incorporated by specifying probability distributions for key uncertain variables. The derivation of these distributions is outlined below. One aspect that needs to be considered in stochastic simulation is the stochastic dependency between variables (Hardaker et al., 2004a). To avoid biasing the results, for the stochastic variables we allowed for first-order autocorrelation (i.e., inter-temporal correlation) between years, as well as crosscorrelation. The coefficients of variation of all stochastic variables were assumed to increase linearly by 2% a year over the planning period, reflecting greater (general and policy) uncertainty with time over the planning period. Except for rates of interest (see below), the stochastic variables, X~ qt , were simulated as X~ qt ¼ X qt þ rqt  CSNDqt  Eq ; X qt

ð7Þ rqt

where is the per unit mean of variable q at time t; is the standard deviation per unit of variable q at time t; CSNDqt is a cross- and auto-correlated standard normal deviate for the variable q at time t, and Eq is the variance

G. Lien et al. / Agricultural Systems 94 (2007) 541–552

expansion factor for variable q which accounts for changes (increases) in the relative variability of a stochastic variable over time. Richardson et al. (2000) give more details of how the stochastic dependencies were estimated and simulated. A problem with the modelling procedure arose in cases when terminal equity was positive yet in some of the years between the start and end of the period the equity was negative, and the farmer was therefore technically insolvent. Computed negative equity in some years during the planning period does not necessary mean that the business would cease because the farmer has options to avoid bankruptcy. Partly because these options would vary widely between farms, it was considered too difficult to model them. Instead, an approximate approach was used whereby the interest rate on borrowed funds was increased by 5 percentage points over the stochastic base interest rate for any years in which the equity was negative. This was subjectively judged to reflect the extra (direct and indirect) costs as the solvency of the farm business declined. These costs include the possibility of higher interest rates on loans as creditworthiness falls. With our approach, therefore, any negative equity during the planning period is carried forward to the terminal equity with further accumulation of interest on debts and only at the end of the planning period is account taken of possible bankruptcy. Private consumption was assumed fixed every year in the planning period, independent of bad or good years. In reality, it is to be expected that consumption spending would be increased or reduced depending on the available funds, but refinement along these lines was not judged necessary for this illustrative example. To summarise, the way the overall model fits together is outlined in the flow diagram in Fig. 1. The distribution of terminal equity derived from the simulation results was used to estimate the probability of business survival as in Eq. (2). The distribution of wealth as defined by Eq. (6) was used as input data for the risk SERF analysis using Eq. (4). The lower left loop in Fig. 1, accounting for negative terminal equity, is to represent wealth not at risk for the SERF analysis, as in Eq. (6). The simulation model used was programmed in Excel and simulated using the Excel add-in SimetarÓ (Richardson, 2004). In the simulation Latin hypercube sampling was used with 500 iterations to estimate the performance variables for each specified scenario. For further details about the stochastic budgeting model see Lien (2003). 2.4. Data The input data on crop production activities summarised in Table 1 were obtained from experimental arable cropping system data (1991–1999) from Eastern Norway (Eltun et al., 2002), supplemented with data on prices, area payment schemes and variable and fixed costs from other sources. Two cropping systems were included in the main data set, conventional crop production and organic crop production. Each farmlet in the experiment had eight rota-

545 General assumptions & data Choice of system & rotation

Iteration n = 1...N

Year t = 1...T

Stochastic yields & prices i

Crop yields & prices

Farm annual net cash flow

Stochastic interest rate Increase stochastic interest rate

Financial balance at end of year t

Yes

Equity initially negative? No No

t = T? Yes Set to equity not at risk

Terminal equity

Yes Count n(T)

Terminal equity negative? No Record terminal equity

n = N?

No

Yes Compute S(T ) and CE(w, r )

Fig. 1. Outline of how the whole-farm simulation model fits together.

tion plots and an 8-year crop rotation. All of the crops in each rotation were grown each year. Inspection of the experimental data permitted the combination of some of the crops within a rotation (varieties of the same crops) without significantly compromising the information from the experiment. The consolidation resulted in five crops in the conventional system and six crops in the organic system. The stochastic yield variables were based on the experimental cropping data. Two farm models, one for each farming system, were constructed, each with 40 ha of arable land, a typical size of crop farms in the region. The farms with conventional cropping systems were assumed to grow 15 ha barley, 10 ha oats, 10 ha spring wheat, and 5 ha potatoes. The organic crop systems consisted of 10 ha barley, 5 ha oats, 10 ha spring wheat, 5 ha potatoes, and 10 ha of fertility-

546

G. Lien et al. / Agricultural Systems 94 (2007) 541–552

Table 1 Descriptive yield, product price, area payments and variable costs for each individual crop and cropping systems Variable Conventional Product yield Product price Area payment Variable costs Organic Product yield Product price Area payment Variable costs

Unit

Barley Ia

Barley IIa

Oats

Potatoes

Wheat

Mean kg ha1 CVb kg ha1 Mean NOK kg1 CVb NOK kg1 Det.c NOK ha1 Det.c NOK ha1

5018 0.28 1.64d

5665 0.16 1.64d

5394 0.16 1.41d

3300 6061

3300 6061

3300 5684

30,839 0.23 1.66 0.21 2500 26,505

5903 0.16 2.04 0.09 3300 7302

Mean kg ha1 CVb kg ha1 Mean NOK kg1 CVb NOK kg1 Det.c NOK ha1 Det.c NOK ha1

3165 0.43 2.79d

3823 0.35 2.79d

3415 0.44 2.36d

5800 6322

5800 6322

5800 6143

21,103 0.44 2.49 0.21 5000 26,510

3422 0.18 3.18 0.05 5800 6715

Grass clover

8939 0.23 1.43d 3540e 3926

Year 2004 price level. a Barley I and Barley II represent two different varieties of barley. b CV, coefficient of variation, defined as standard deviation divided on mean for the variable. c Det., deterministic variable. d Product prices for barley, oats and grass clover are specified as deterministic prices. e Organic area payment for grass clover is NOK ha1 550.

building annual grass-clover (for silage). These crop-mix proportions are almost the same as were used in the experiment. The different land allocation in the organic compared to the conventional farming system may imply differences in profitability, probably to the disadvantage of the less intensive organic system. However, the need for a fertility-sustaining rotation in organic farming can be seen as a part of the ‘‘cost’’ of organic farming (Lampkin et al., 2006, Sections 1 and 5). Norwegian crop farmers receive quite substantial producer support, chiefly through import tariffs on crop commodities and governmental area payments for crops. Under the annual agricultural agreement between Norwegian farmers and the government, target prices (maximum average prices) are set for most commodities, including wheat, barley and oats (Norwegian Agricultural Economics Research Institute, 2005, pp. 17–48). Hence, the general level of grain prices can be regarded as non-stochastic in Norway. However, variability in quality parameters causes some unpredictability in the farm-gate price for wheat. These quality parameters were recorded in the experiment and this information was used to model stochastic wheat prices. The potato price has been quite unpredictable and was stochastically modelled. Deflated historical potato prices in NOK per kg for 1991–1999 from the Agricultural Price Reporting Office (Landbrukets Priscentral, 2000) were used to specify the empirical potato price distribution. Based on organic potato price premiums in Norway in 2003/2004 and price premiums for organic potatoes in other European countries, organic potatoes were initially assumed to be sold at prices 50% above conventional prices and, for lack of reliable historical data for empirical estimation, assumed to have the same relative price variability, and hence higher absolute variability. Higher absolute variabil-

ity for organic potato prices was subjectively judged to be appropriate due to the much smaller and hence presumably more volatile market for organic produce. Deterministic product prices (reduced for the yielddependent hauling cost and ensilage cost for clover grass), input prices, prevailing area payment schemes (2003/2004), variable costs and fixed costs were taken from Norwegian Agricultural Economics Research Institute (2004). For more details about the cropping system data used and empirical results of differences in risk between conventional and organic cropping systems in Eastern Norway, see Lien et al. (2006). The following initial mean rates of interest per year were assumed: short-term loan interest rates 6%, long-term loan interest rate 5%, and deposit interest rate 4%. The probability distributions and trends over the planning horizon in the stochastic interest rates on financial assets and liabilities were forecasted with an autoregressive model. The reason for using an autoregressive model and not a simple regression model is that interest rate often has a mean reversion trend, i.e., the interest rate normally reverts to a long-run trend. The forecasting model was estimated using annual average rates on 10-year Government bonds for the period 1985–2003. The interest rates on shortand long-term loans and deposits were all assumed to be perfectly correlated. The forecast values and their standard deviations from the estimated first-order autoregressive model AR(1) equations were used as indexes for the stochastic distribution and stochastic trends of all interest rates used in the budgeting model (Lien, 2003). The present value of the human capital component of  H , was assumed to be roughly NOK wealth not at risk, w (Norwegian kroner, € 1.00  NOK 8.00) 300,000, intended to represent capitalised value of income of the farm family from wages (or other sources such as social security bene-

G. Lien et al. / Agricultural Systems 94 (2007) 541–552

fits) over the longer term. The present value of other wealth  P was initially set to NOK 200,000. not at risk component w Since this estimate for wealth not at risk includes the less tangible benefits of farming (e.g., the value of the farming lifestyle and prospect of capital gain in the value of the farm) and the quantification of these effects is uncertain and varies greatly between farms, a sensitivity analysis has been included, showing how the level of wealth not at risk influences the results (for a fixed degree of risk aversion). The stochastic total wealth from farming (including the ~ , is defined as the present value possibility of bankruptcy), w of terminal equity (net assets), discounted at an annual discount rate of 0.05, chosen to represent the farmer’s opportunity cost of the invested equity (Miller and Bradford, 2001). Of course, since farmers have different opportunities, the discount rate will vary between farmers and so should be adapted to each case. Similarly, the appropriate degree of risk aversion will vary from case to case. The SERF method shows which risky alternative decision makers with different degrees of risk aversion prefer, leaving the final choice to the individual decision maker. The range of risk aversion to be used in the SERF analysis is crucial, and here the analysis is evaluated in the range of relative risk aversion with respect to wealth, rr, between 0 and 4. While Arrow (1965) argued that relative risk aversion for wealth was likely to be about 1.0, values of 0.5 and 4 were selected to represent the plausible lower and upper limits (Hardaker et al., 2004a). The value of zero was included to show the extreme case of indifference to risk. 2.5. Description of scenarios investigated By way of illustration, the model was used to compare the two cropping systems, conventional and organic, under three scenarios. For the first we assumed that the prevailing yield and price levels, the existing subsidy payment system for conventional farming and the current organic price premiums continue to apply. Lien et al. (2006) showed that the current organic area payments and price premiums make organic crop farming significantly more profitable than conventional farming. If organic area payments were removed, the same study showed that the expected economic return from organic cropping was slightly higher than under conventional management. There is a widespread belief, at least among crop farmers, that organic area payments may soon be reduced or removed (Koesling et al., 2004). Hence, in the first scenario we have chosen to illustrate the economic sustainability of the organic system without organic area payments. There is also an expectation that the price premium may decrease with increased supply of organic product as more farmers convert to organic production (Koesling et al., 2004). Hence, in scenario two, we also phased out the organic market price premiums.

547

In scenario three, we opted to explore the effects on the economic sustainability and risk efficiency of the two systems of a shock, represented by total crop failures in one of the 6 years in the planning period. This scenario is intended to explore the economic robustness of the farm systems to big risks (Hardaker, 2006). 3. Illustrative results 3.1. Scenario one: the ‘current’ situation but organic area payments excluded Fig. 2 shows the means and limits of the distributions of the equity of the two farming systems for each year in the planning period. On average the organic farming system, compared to conventional, has somewhat higher equity during the planning period. However, the variability is also greater. The cumulative distribution functions (CDFs) of terminal equity of the two systems are presented in the left part of Fig. 3. This figure shows that, although the economic sustainability of the organic systems appears to be somewhat less than for conventional, for neither of the systems is economic sustainability seriously threatened (less than 0.02 probability of negative terminal equity for the organic system). Which of the two farming systems a farmer would prefer may therefore depend on the comparison in terms of overall risk efficiency. The SERF analysis of the two farming systems is summarised in the right part of Fig. 3. The CE of the organic system is above that of the conventional system over the full range of relative risk aversion modelled. These results imply that the economic sustainability of the conventional system is superior to that of the organic, since the lower part of the cumulative distribution for the conventional systems has slightly higher values of terminal equity for any cumulative probability level below about 0.2. By contrast, Hansen et al. (1997) found significant differences between cropping systems with respect to economic sustainability. In terms of the risk efficiency criterion, any farmer with a coefficient of relative risk aversion within what we assumed to be a plausible range would prefer the organic system to the conventional. Cacho and Simmons (1999) found a similar divergence between economic sustainability and risk efficiency using a different approach. 3.2. Scenario two: effects of gradually reduced organic price premiums In this scenario, it is assumed that organic price premiums follow a yearly linear decreasing trend, so that by 2009 the organic producer receives the same prices as the conventional farmer. Fig. 4 illustrates the relationship between time and economic sustainability of the two farming systems for this scenario. The probability of survival of the organic system begins to decline from 2007 reaching a level

548

G. Lien et al. / Agricultural Systems 94 (2007) 541–552 2400000

Equity (NOK)

1800000 1200000 600000 0 -600000

2004

2005

Average conventional Average organic

2006

2007

2008

Year 0.1st Percentile con 0.1st Percentile org

2009

99.9th Percentile con 99.9th Percentile org

Fig. 2. Simulated mean and range of equity in NOK for the 6-year planning period for conventional (con) and organic (org) farming systems under scenario 1.

1.00 CE of wealth in NOK

1 000 000

Probability

0.75

0.50

0.25

0.00 -1000000

0 1000000 Terminal equity (NOK) Conventional

2000000

750 000

500 000

250 000

0 0

3 2 1 Relative risk aversion, r r Conventional

Organic

4

Organic

Fig. 3. Simulated cumulative distribution functions of terminal equity in NOK (to the left) and certainty equivalent (CE) curves of wealth in NOK evaluated in the range of relative risk aversion with respect to wealth (rr) between 0 and 4 (to the right) for conventional and organic farming systems.

Table 2 Sensitivity elasticities of economic sustainability (si) for the organic farming systems with decreasing organic price premiums

Sustainability

1.00 0.75 0.50 0.25 0.00 2004

2005

2006 Year Conventional

2007

2008

Variable type

si

Yield or product price Variable cost Area payment

0.28 0.74 0.14

2009

Organic

Fig. 4. Relationship between time and economic sustainability of the conventional and organic farming systems with decreasing organic price premiums.

of 86.1% by 2009. The corresponding probability of survival for the conventional system by 2009 is estimated to be 99.8%. A closer investigation of which variables most affect the economic sustainability of the organic farming system under the assumption of declining price premiums is reported in Table 2. (The conventional system is not reported, since for this scenario it was almost 100% economically sustainable.) At the whole-farm level, total vari-

able costs affected the economic sustainability of the organic farming system most, in that economic sustainability decreased by 0.74% for a 1% increase in the variable costs. The second most sensitive exogenous variables were found to be the prices or yields. If the organic price premiums should (partly) erode, could yield improvement over time compensate for the decreased prices? Organic farming may have a potential for increased yields due to rising soil fertility (Ma¨der et al., 2002) and improved management. Productivity could rise as farmers new to this form of production accumulate experience in how to do things better. However, simulation results show that even an unrealistic annual 5% increase in yields in organic farming would not be enough to outweigh the unfavourable impact on economic sustainability of declining organic price premiums.

G. Lien et al. / Agricultural Systems 94 (2007) 541–552

As Figs. 4 and 5 show, for scenario two there is consistency between the farming system choice based on economic sustainability grounds and on the basis of risk efficiency. By both criteria the conventional system is superior to organic.

Sustainability

1.00

3.3. Scenario three: capacities of the systems to respond to crop failures In this scenario, we artificially introduced crop failures in the third year of the 6-year planning horizon. We assumed that third year yield levels are so low that the crops are not worth harvesting, i.e., zero yields. As a consequence, we also reduced the variable costs for all crop enterprises to account for the harvesting and post-harvesting costs saved. Fig. 6 shows the economic sustainability graph. As expected, the economic sustainability of both the conventional and organic system was negatively affected by the assumed crop failures. By the final year, the probability of survival was found to be 87% for the organic system, compared with 92% for the conventional system. In Table 3 the variables at whole-farm level found to most influence the economic sustainability of the two systems are presented. The most sensitive exogenous variables for the conventional system were found to be yields or product prices, in that economic sustainability increased by 0.85% for a 1% increase in product price or yield. The higher elasticities in the organic than in the conventional system for total variable costs may be due, among other thing, to the somewhat higher cost level in the organic system. The higher elasticity in conventional than organic farming systems for area payments occurs because, under our assumptions, the extra organic area payments have been excluded, so that, under the less intensive organic cropping system, the level of payments is less than for the conventional system. In the study of a Colombian hill farm by Hansen et al. (1997), economic sustainability was more sensitive to crop prices than variable costs.

0.25

2005

2006

2007

Year Conventional

2008

2009

Organic

Fig. 6. Relationship between time and the economic sustainability of conventional and organic farming systems in the event of crop failures in the third year.

Table 3 Sensitivity elasticities of economic sustainability (si) for the conventional and organic farming systems in the event of crop failures in the third year Variable type

si

Yield or product price Variable cost Area payment

Conventional

Organic

0.85 0.71 0.36

0.73 0.83 0.28

Fig. 7 shows the CDF graph of terminal equity (to the left) and the SERF graph of NPV (to the right) in this scenario. The left part of the figure shows that the conventional system is more economically sustainable than the organic system. The right part of the figure shows the opposite ranking. Farmers with any degree of risk aversion in the assumed range would prefer organic to conventional. 3.4. Sensitivity analysis of the levels of wealth not at risk In Fig. 8, we have kept the degree of relative risk aver P fixed sion with respect to wealth, rr, fixed at one and w at NOK 200,000, and show the effect on simulated certainty equivalent for each scenario of varying assumed

CE of wealth in NOK

Probability

0.50

1 000 000

0.75

0.50

0.25

750 000

500 000 250 000

0

0.00

0 750000 Terminal equity (NOK)

Conventional

0.75

0.00 2004

1.00

-750000

549

1500000

Organic

0

2 3 1 Relative risk aversion, r r

Conventional

4

Organic

Fig. 5. Simulated cumulative distribution functions of terminal equity in NOK (to the left) and certainty equivalent (CE) curves of wealth in NOK evaluated in the range of relative risk aversion with respect to wealth (rr) between 0 and 4 (to the right) for conventional and organic farming systems, under the assumption of declining organic price premiums.

550

G. Lien et al. / Agricultural Systems 94 (2007) 541–552 1 000 000 CE of wealth in NOK

1.00

Probability

0.75

0.50

0.25

750 000

500 000

250 000

0

0.00 -1000000

0

0

1000000

Terminal equity (NOK)

Conventional

1

2

4

3

Relative risk aversion, r r

Conventional

Organic

Organic

Fig. 7. Simulated cumulative distribution functions of terminal equity in NOK (to the left) and certainty equivalent (CE) curves of wealth in NOK evaluated in the range of relative risk aversion with respect to wealth (rr) between 0 and 4 (to the right) for conventional and organic farming systems, under the assumption of crop failures in the third year.

CE (NOK)

1 000 000 800 000 600 000 400 000 0

300 000

600 000

900 000

Human capital not at risk (NOK) Sc. 1 Conventional

Sc. 1 Organic

CE (NOK)

1 000 000 800 000 600 000 400 000 0

300 000

600 000

900 000

Human capital not at risk (NOK) Sc. 2 Conventional

Sc. 2 Organic

CE (NOK)

1 000 000 800 000 600 000 400 000 0

300 000

600 000

900 000

Human capital not at risk (NOK) Sc. 3 Conventional

Sc. 3 Organic

Fig. 8. Simulated certainty equivalent (CE) curves of wealth in NOK for conventional and organic farming systems, given a degree of relative risk  H for the three scenarios (1, 2, and 3) aversion with respect to wealth equal one, w  P equal to NOK 200,000 and varying levels of human capital not at risk, w described in section 2.5.

 H between NOK zero and human capital not at risk, w NOK 900,000. In scenarios one (the ‘‘current’’ situation) and scenario two (gradually reduced organic price premiums), the conventional system is almost unaffected as human capital

not at risk is varied. Especially in the case with gradually reduced organic price premiums, the organic system is quite sensitive to change in the assumed value of human capital. The reason is that, compared to the conventional system, a larger part of the lower CDF tail is negative

G. Lien et al. / Agricultural Systems 94 (2007) 541–552

(see left part of Fig. 5) so that the bankruptcy provisions built into the model more often come into play. In scenario 3 (capacities of the systems to respond to crop failures), human capital levels not at risk considered (NOK 0, 300,000, 600,000 and 900,000) correspond to points on the upper part of the CDFs of the risky NPV (the expected wealth of farming is just above NOK 500,000, cf., Fig. 7). Consequently, exploiting higher present values of human capital will frequently be a better option than the present value of terminal equity to be got from farming. In other words, for scenarios 2 and 3 the organic system benefits most at increasing levels of wealth not at risk, since increased human capital (as a part of the wealth not at risk) reduces the down-side risk in a similar way as insurance but keeps the up-side potential. As all graphs in Fig. 8 show, no curves cross, and the rankings of the conventional and organic systems for all scenarios are unaffected by varying the level of human capital not at risk.

4. Discussion and conclusion On the basis of our findings, it seems that the organic farming system is somewhat less economically sustainable than the conventional system, under the cases examined with the organic area payment removed. This must be a qualified conclusion for various reasons. The first relates to the unavoidable omission of any long-term difference between the two systems. Critics of conventional highinput farming often argue that this form of production will have adverse long-run consequences. They take this position often in the face of positive historical trends in output and productivity. On the other hand, some studies, including some in Norway, have reported declining soil nutrient concentrations of nitrogen, phosphorus and/or potassium, under organic farming, calling into question the long-term sustainability of these crop farming practices (e.g., Eltun et al., 2002; Gosling and Shepherd, 2005). Second, this analysis has taken no account of differences in externalities of the two systems. It would be possible to extend the study by valuing and including any externalities. Unfortunately, measuring and valuing such externalities is likely to be difficult. As an alternative, it could be possible to account for the assessed risk that negative externalities such as ground or surface water pollution, might lead to legal liabilities, which themselves would threaten the continuation of a farming system. Third, as in the study by Hansen et al. (1997), the present model is confined to fixed farming systems, one for each form of production, as represented in the experimental data. However, in practice, farmers are likely to change cropping plans in the light of evolving expectations about yields and prices relevant to the different modes of production. It might be possible in future work to link the simulation model to a mathematical programming model to compute optimal annual farm plans accounting for a farmer’s individual beliefs, preferences and circumstances.

551

Finally, as in all such studies, the results depend on the assumptions made, some, such as the assumption about the variance of organic potato prices, based on wholly subjective assessments. With more data, these assumptions could be refined, possibly affecting the conclusions reached. With these reservations, we suggest that the model illustrated above could be useful for supporting decisions by farmers on whether or not to shift out of conventional production and into organic farming. In such applications it would be relatively easy to ‘tune’ the model to reflect an individual farm and farmer’s circumstances and expectations. Similarly, use of the above model could be helpful to policy makers. Results such as those presented might be useful to support decisions about what policies would be effective to pursue objectives of increased organic production. For example, the results presented above suggest that the aim may be compromised if the market premiums for organic produce diminish. Seeking to push too many farmers into changing to organic farming could lead to an oversupply of organic produce and a fall in price premiums not dissimilar to that modelled, leading to financial unsustainability for some organic producers. An important point illustrated in the results presented is the difference between the particular measure of economic sustainability used and risk efficiency. While the ranking of the two systems is in the main similar between economic sustainability and risk efficiency criteria, there are differences. Logic suggests that no criterion of sustainability (whether based on a single or on multiple indicators) will yield identical ranking to those indicated using economic efficiency criteria – even when the latter are extended to include changes in values of assets and externalities. Those making recommendations to farmers and policy makers should be aware of the costs to farmers and society of recommending or requiring the uptake of farming methods that may appear technically more sustainable but that are less economically efficient or even less economically sustainable. References Arrow, K.J., 1965. Aspects of the Theory of Risk-Bearing. Academic Bookstore, Helsinki. Artzner, P., Delbaen, F., Eber, J.-M., Heath, D., 1999. Coherent measures of risk. Mathematical Finance 9, 203–228. Beckerman, W., 1992. Economic growth and the environment: whose growth? Whose environment? World Development 20, 481–496. Cacho, O.J., Simmons, P., 1999. A genetic algorithm approach to farm investment. Australian Journal of Agricultural and Resource Economics 43, 305–322. Conway, G.R., 1985. Agroecosystem analysis. Agricultural Administration 20, 31–55. Conway, G.R., 1994. Sustainability and agricultural development: tradeoffs with productivity, stability and equitability. Journal of Farming Systems Research-Extension 4, 1–14. Eltun, R., Korsæth, A., Nordheim, O., 2002. A comparison of environmental, soil fertility, yield, and economic effects in six cropping systems based on an 8-years experiment in Norway. Agriculture, Ecosystems & Environment 90, 155–168.

552

G. Lien et al. / Agricultural Systems 94 (2007) 541–552

Gosling, P., Shepherd, M., 2005. Long-term changes in soil fertility in organic arable farming systems in England, with particular reference to phosphorus and potassium. Agriculture, Ecosystems & Environment 105, 425–432. Hamblin, A., 1992. How do we know when agricultural systems are sustainable? In: A. Hamblin (Ed.), Environmental Indicators for Sustainable Agriculture. Report on a National Workshop, November 28–29, 1991, p. 90, Bureau of Rural Resources, Land and Water Resource Research and Development Corporation, Grains Research Corporation, Canberra, cited by Hansen and Jones (1996). Hansen, J.W., Jones, J.W., 1996. A system framework for characterizing farm sustainability. Agricultural Systems 51, 185–201. Hansen, J.W., Knapp, E.B., Jones, J.W., 1997. Determinants of sustainability of a Columbian hillside farm. Experimental Agriculture 33, 425–448. Hardaker, J.B., 2006. Farm risk management: past, present and prospects. Journal of Farm Management 12, 593–612. Hardaker, J.B., Huirne, R.B.M., Anderson, J.R., Lien, G., 2004a. Coping with Risk in Agriculture, second ed. CABI, Wallingford. Hardaker, J.B., Richardson, J.W., Lien, G., Schumann, K.D., 2004b. Stochastic efficiency analysis with risk aversion bounds: a simplified approach. Australian Journal of Agricultural and Resource Economics 48, 253–270. Holling, C.S., 1973. Resilience and stability of ecological systems. Annual Review of Ecology and Systematics 4, 1–12. Jacobs, M., 1995. Sustainable development – from broad rhetoric to local reality. Conference Proceedings from Agenda 21 in Cheshire, 1 December 1994, Document 493, Cheshire County Council, cited by Rigby and Ca´ceres (2001). Kaine, G.W., Tozer, P.R., 2005. Stability, resilience and sustainability in pasture-based grazing systems. Agricultural Systems 83, 27–48. Keeney, R.L., Raiffa, H., 1976. Decisions with Multiple Objectives: Preferences and Value Tradeoffs. Wiley, New York. Koesling, M., Ebbesvik, M., Lien, G., Flaten, O., Valle, P.S., Arntzen, H., 2004. Risk and risk management in organic and conventional cash crop farming in Norway. Acta Agriculturae Scandinavica C – Food Economics 1, 195–206. Lampkin, N., Measures, M., Padel, S. (Eds), 2006. 2007 Organic Farm Management Handbook, seventh ed. Organic Research Group, Institute of Rural Sciences, University of Wales, Aberystwyth. Landbrukets Priscentral, 2000. Landbrukets Priser: Jordbruk, Hagebruk, Driftsmidler. Landbrukets Priscentral, Oslo (in Norwegian). Lien, G., 2003. Assisting whole-farm decision making through stochastic budgeting. Agricultural Systems 76, 399–413. Lien, G., Flaten, O., Korsaeth, A., Schumann, K.D., Richardson, J.W., Eltun, R., Hardaker, J.B., 2006. Comparison of risk in organic, integrated and conventional cropping systems in Eastern Norway. Journal of Farm Management 12, 385–401. Lynam, J.K., Herdt, R.W., 1989. Sense and sustainability: sustainability as an objective in international agricultural research. Agricultural Economics 3, 381–398. Ma¨der, P., Fliessbach, A., Dubois, D., Gunst, L., Fried, P., Niggli, U., 2002. Soil fertility and biodiversity in organic farming. Science 296, 1694–1697. Mahoney, P.R., Olson, K.D., Porter, P.M., Huggins, D.R., Perrilo, C.A., Crookston, R.K., 2004. Profitability of organic cropping systems in southwestern Minnesota. Renewable Agriculture and Food Systems 19, 35–46.

Manfredo, M.R., Leuthold, R.M., 1999. Value-at-risk analysis: a review and potential for agricultural applications. Review of Agricultural Economics 21, 99–111. Miller, S.E., Bradford, G.L., 2001. Teaching net present value analysis: return-to-assets versus return-to-equity. Review of Agricultural Economics 23, 223–241. Ministry of Agriculture, 1999. St. meld. nr. 19 (1999–2000). Om norsk landbruk og matproduksjon. The Ministry of Agriculture, Oslo (in Norwegian). Norwegian Agricultural Economics Research Institute, 2004. Ha˚ndbok i Driftsplanlegging 2004/2005. Norwegian Agricultural Economics Research Institute, Oslo (in Norwegian). Norwegian Agricultural Economics Research Institute, 2005. Norwegian Agriculture – Status and Trends 2005. Norwegian Agricultural Economics Research Institute, Oslo. Pandey, S., Hardaker, J.B., 1995. The role of modelling in the quest for sustainable farming systems. Agricultural Systems 47, 439– 450. Pannell, D.J., Glenn, N.A., 2000. A framework for the economic evaluation and selection of sustainability indicators in agriculture. Ecological Economics 33, 135–149. Pannell, D.J., Schilizzi, S., 1999. Sustainable agriculture: a question of ecology, equity, economic efficiency or experience? Journal of Sustainable Agriculture 13, 5–66. Payraudeau, S., van der Werf, H.M.G., 2005. Environmental impact assessment for a farming region: a review of methods. Agriculture, Ecosystems & Environment 107, 1–19. Pretty, J.N., 1995. Regenerating Agriculture: Policies and Practices for Sustainability and Self Reliance. Earthscan, London. Quiggin, J., 1992. Some observations on insurance, bankruptcy and input demand. Journal of Economic Behaviour and Organization 18, 101– 110. Richardson, J.W., 2004. Simulation for Applied Risk Management. Department of Agricultural Economics, Texas A&M University, College Station, Texas. Richardson, J.W., Klose, S.L., Gray, A.W., 2000. An applied procedure for estimating and simulating multivariate empirical (MVE) probability distributions in farm-level risk assessment and policy analysis. Journal of Agricultural and Applied Economics 32, 299–315. Rigby, D., Ca´ceres, D., 2001. Organic farming and the sustainability of agricultural systems. Agricultural Systems 68, 21–40. Roberts, W.S., Swinton, S.M., 1996. Economic methods for comparing alternative crop production systems: a review of the literature. American Journal of Alternative Agriculture 11, 10–17. Smith, E.G., Clapperton, M.J., Blackshaw, R.E., 2004. Profitability and risk of organic production systems in the northern Great Plains. Renewable Agriculture and Food Systems 19, 152–158. Szego¨, G., 2005. Measures of risk. European Journal of Operational Research 163, 5–19. Venter, G.G., 1983. Utility with decreasing risk aversion. 1983 Proceedings of the Causality Actuarial Society, vol LXX, No. 133, pp. 144– 155. Zhen, L., Routray, J.K., Zoebisch, M.A., Chen, G., Xie, G., Cheng, S., 2005. Three dimensions of sustainability of farming practices in the North China Plain: a case study from Ningjin County of Shangdong Province, PR China. Agriculture, Ecosystems & Environment 105, 507–522.