Agribusiness Review - Vol. 7 - 1999

Paper 8
ISSN 1442-6951

THE RETURNS FROM SOME REPLANTING STRATEGIES IN THE ORANGE INDUSTRY

K.J. Elton¹ , R. Hutton¹ and Dr. J.D. Mullen²

1. New South Wales Agriculture, Yanco NSW;
2. New South Wales Agriculture, Orange NSW.

Acknowledgements: The assistance of Adam Bialowas is appreciated.

Abstract

Due to a downturn in concentrate juice markets, there has been a trend within the orange industry to reduce the reliance on Valencia oranges. Reworking (where much of the main branch system of existing healthy Valencia trees is removed and budsticks of selected Navel orange clones or other citrus varieties are inserted in cut ends of the main limbs) is an alternative to replanting as it reduces the pre-production yield losses which are associated with establishing a replant site. Using benefit/cost criteria applied to development budgets, a reworking strategy was found to be more profitable than strategies which involved either replanting to Navels or delaying replanting to Valencias.

An important source of risk is the difference in expected price between Valencia and Navel oranges and how this varies through time. Based on stochastic dominance testing, the reworking strategy was dominant for price conditions experienced by the industry over the last thirty years. However, in the future, this will depend on demand and supply conditions in fresh fruit and processing markets.

1. Introduction

One of the most important issues currently affecting the Australian Citrus Industry is the low return to growers for Valencia oranges, which are mainly processed for juice. Large stocks and increasing production in both Brazil and Florida are largely responsible for the decrease in the international price for frozen concentrate orange juice (FCOJ). This downturn in world FCOJ prices has resulted in prices for local oranges for concentrate production falling from around $100 a fresh fruit tonne at the beginning of the 1996/97 season, to a low of $30 a tonne in 1997. Due to the downturn in the concentrate juice market, there has been a tendency to broaden existing varieties to include more Navel orange plantings and reduce reliance on Valencia oranges.

In response to low returns for oranges processed into juice, an estimated 607,000 Valencia orange trees have already been removed from orchards in the three citrus producing regions of southern Australia (Riverland, Murray Valley and Riverina). NSW Riverina, Victoria and SA during the period April 1995 to April 1997. It is expected that a further 500,000 trees are likely to be removed by the end of 1999, making the total removals 1.1 million Valencia trees, or 22% of the Valencia plantings.

In contrast, Navel oranges compete with mandarins as the most profitable sector of the Australian citrus industry and Navel orange exports generate significant export revenue. Australian growers are able to consistently produce good quality Navel oranges. The fruit is in demand on export and local fresh fruit markets.

This paper reports a financial analysis of three strategies available to growers with existing stands of processing Valencias. One strategy is to ‘stick it out' and continue growing Valencias. A second strategy is to replant to Navel oranges for the fresh market. The third strategy is to ‘rework' existing Valencia trees to Navels . Reworking involves removing much of the main branch system of the Valencia tree and inserting Navel budsticks into the ends of the main limbs and is described more fully in an Appendix. It is an attractive alternative to replanting because it makes use of existing healthy rootstocks and hence reduces pre-production yield losses associated with replanting. To be an economically viable alternative to replanting, high early yields must be realised from reworking, with rootstocks remaining healthy for a reasonable period before replanting is necessary. However, there are no current accurate guidelines for long-term productivity from reworking under Australian conditions.

This paper compares these three investment strategies. Benefit/cost analysis is applied to development budgets. The potential impact of price uncertainty is analysed using Monte Carlo simulation techniques within @RISK as well as stochastic dominance analysis. Clearly the outcome of these strategies depends on the relative prices of processing Valencias and fresh Navels in the future. Industry outlook information and historical prices are the basis of the risk analysis below.

2. Market Conditions in the Orange Industry

There are approximately 3,000 citrus growers in Australia. Nationally, oranges make up 77% of all citrus plantings grown on 22,519 hectares. Currently orange plantings comprise 54% Valencias (58% in 1996) and 46% Navels (42% in 1996).

The extent of recent tree pulling in the Australian Valencia industry has already been noted. During this time total citrus plantings have contracted by 4% to 9.3 million trees. The forecast total of the Australian Valencia orange crop for 1998/99 season is 320,000 tonnes, the lowest crop since 1988/89, excluding the 1997/98 season. The likely distribution of the 1998/99 Valencia crop comprises 60,000 tonnes for fresh export, 85,000 tonnes for fresh domestic, 155,000 tonnes for fresh juice, and 20,000 tonnes for FCOJ juice. With continued removal of trees, the 1999 Australian Valencia orange crop could be further reduced.

In contrast plantings of mandarins and summer Navels have increased by 6% and 8% respectively. It is estimated that by the year 2010 new plantings of Navels and mandarins will add $15.4 million to the gross value of citrus production in southern Australia.

In 1980 when structural adjustment pressures in the industry first appeared, the domestic industry had been receiving some tariff protection. Tariffs for the citrus industry in 1979 were 65% ad valorum on imported frozen concentrate juice but were reduced to 44% in 1982. In July 1986 the Federal Government removed all exemptions for countries or commodities under the Developing Country Preference arrangement. This reduced the ad valorum component of this tariff on Brazilian FCOJ to 5%. Over the next 10 years the general tariff was reduced to 5 % .

With the price of concentrate orange juice falling further during 1986, the Industries Assistance Commission reviewed temporary assistance to citrus growers. Prior to this review, the tariff on orange and tangerine juices, including concentrates was reduced. Following the review, this decision was reversed, however, no further tariff assistance was granted. It was also decided that the Federal Government would match funds raised by the states for carry-on loans to citrus producers under Part B of the Rural Assistance Scheme (RAS). During this same period there was widespread recognition that the industry needed to reduce its reliance on fruit for processing and to re-direct itself towards more competitive areas. At present, industry protection (import tariff of 5%) is limited and has been in place since July 1996.

Production of FCOJ has continued to increase in Brazil and Florida. Florida's fresh orange production reached a record 9.5 million tonnes in 1997/98 (USDA statistics) . Normally 95% of Florida's orange production is processed into orange juice. The 1997/98 Brazilian orange crop yielded a record 16.36 million tonnes of which 10.6 million tonnes was processed, being two million tonnes higher than the previous season. Approximately 75% of the Brazilian orange crop is processed. According to the association representing Brazil's citrus juice processors, Brazil produced 420,950 tonnes of FCOJ of which 163,000 tonnes were exported in just two months from the start of the 1997/98 market year (July 1). It is reported that this season's production is 20% down with exports falling by almost 2%. This has given rise to an increase in FCOJ price from lows of less than US$950 (A$30/ton fresh fruit equivalent) in 1997/98 to approximately US$1550 in April 1999 (equivalent to an on-farm price of A$90/ton in Australia). Statistics for July 1999 show that the international FCOJ price has again fallen to US$1000 (A$60/tonne). This demonstrates the price volatility seen on the international market.

The four major processors/exporters of Brazilian FCOJ have all continued to buy into the Florida processing industry. Three processing plants in Florida have announced major expansion plans worth US$32 million. It has been reported (Florida Department of Citrus) that Florida processors during the past year have added 151 million litres of storage capacity, with more planned.

The world's demand for orange juice is not keeping pace with the increased production and hence there is pressure on price. The current lower world FCOJ price is expected to continue in the short term, and will remain significantly lower than prices recorded in the early 1990's. Future FCOJ price variations will continue to be dependant on changing production levels in Brazil and Florida resulting from un-seasonal weather conditions. In 1996, world parity price equated to about $85 Australian per tonne of Valencias, down from $130/tonne in 1995/96. Figure 1 illustrates the volatile fluctuations in real prices received for Australian oranges, especially Valencia orange processing prices which are directly influenced by FCOJ prices. Currently, the world parity FCOJ price (US$1550/tonne) is equivalent to $90/tonne for fresh fruit which is still below the average production cost for Valencia oranges of $243/t in Australia (AHC Citrus Best Practice Groups - Cost of Production 1996/97).

Figure 1 - Real Prices for Valencia and Navel Oranges 1967/68 to 1996/97

Domestic and export market opportunities for fresh oranges are less bleak. Fresh fruit prices on the main local markets at Flemington and Newcastle fluctuate seasonally depending on the size of the Australian crop. As can be seen in Figure 1 and Table 2 average real prices for fresh Navels have also fluctuated quite widely over recent years from a low of $9.45 per box in 1989/90 to $15.13 in 1995/96. These prices represent wholesale prices and equate to an on-farm real price of approximately $270 to $400 per tonne net of packing, transport and marketing charges. Grower returns of $700/tonne have been achieved for premium quality fruit sold in the domestic fresh markets and US export returns have ranged from $1,100 to $1,600/tonne.

An export drive in the past decade has seen Australia's exports of fresh citrus fruit rise from 50,000 tonnes in 1986/87 (5,000 tonnes from Riverina) to 88,000 tonnes with an f.o.b. value of $80 million in 1995/96. At this time, citrus exports from the Riverina totalled nearly 14,000 tonnes. By 1997/98 exports from the Riverina totalled nearly 21,000 tonnes and the national volume increased to 137,000 tons with and f.o.b. value of $141 million.

The expansion in export volumes of citrus from the Riverina has increased dramatically since 1987 and Navel orange exports are clearly the largest single contributor to this change, especially since 1993. Navels constituted more than 75% of citrus exports to Japan from the Riverina in 1996 (5,800 tonnes). More recently, marketing of Riverina Navels to the USA through the Riversun group of packers and exporters (3,850 tonnes) has further increased Navel orange exports from the region.

The average price paid for Australian Navels in the USA over the past 2 seasons has been US$29/carton (A$45) and returns to growers net of packing and marketing costs has been $26/box ($1,430/ton).

In the light of these contrasting developments in fresh and processing sectors it is not surprising that growers are considering replanting or reworking to Navels. There seems little reason to believe that current price relativities are a temporary phenomenon.

3. The Three Replacement Strategies for Comparison

The following analysis compares the three strategies as applied to a hectare of 12 year old Valencia oranges planted at 556 trees per hectare. The assumptions in this analysis reflect current practice in the Riverina. The first strategy is to remain with Valencias, avoid the cost of replacement and hope for favourable price movements. Valencias reach their peak yield at around 18-30 years of age. The average age of Valencia trees in the Riverina is around 50 years of age. We assume that there would be some economic advantage in replacing these trees at 60 years of age. It is estimated that this would cost approximately $8,674 per hectare.

The second strategy is to replant with Navels at a similar cost to that of Valencias. Navels start bearing in the second or third year after planting, however, the fruit is generally harvested in the fourth year of production when yields become commercially significant. We assume that the yields of Navels will also peak around 18 -30 years of age and that the trees will be replanted after 20 years.

For the reworking strategy analysed here, half the tree is reworked in year 1 and the remaining half in year 3 so that Valencias are produced in the first three years. Maximum yields are assumed 15 years after reworking, at which point replanting is required. Reworking costs are $11,857/ha.

The yield profile for Valencia oranges is based on actual data from research trials conducted at Yanco Agricultural Institute. The yield data were generated and extrapolated using a yield prediction model developed by Cullis and Verbyla (1992) . This model is used to predict mean annual yield data based on two-yearly moving average yields recorded over fifteen years from trial plantings.

The yield profile for Navel oranges also has been taken from research trials at Yanco Agricultural Institute. This data is annual yields based on two-yearly moving averages, over fifteen years. As already noted the yield profile for reworking to Navels is uncertain as it is a relatively new technology. Yield projections have been based on canopy development and fruiting potential observed in field trials. Research has shown that following the third year of reworking, reworked trees may be expected to yield similarly to 6 to 7 year old trees. Yield data used in the analysis of the three strategies of the analysis are summarised in Table 1.

Table 1 : Yield Profile for Valencia and Navel Oranges

Price data for fresh Navels were taken from annual monthly average prices received at the Flemington markets over the period 1967/68 - 1996/97 (Table 2) . Monthly prices were transformed into annual average prices to smooth out monthly fluctuations. A total of 30 observations were recorded for wholesale prices of Navel oranges sold in NSW during this period. Data which were not available for the years 1978/79 and 1981/82 were calculated as an average of the price that fell either side of those years. Processing prices for Valencia oranges were taken from the Australian Commodity Statistics 1996 for the period 1973/74 to 1996/97 (Table 2) . Nominal prices were converted to 1997 dollars using the Australian CPI Index. The on farm price of navels on a per tonne basis was estimated by multiplying the real per case price by 50, the number of cases of navels picked that end up at the wholesale market less $350 per tonne which is an allowance for packing costs and a proportion of navels that go to processing. These prices were used in both a deterministic and risk analysis of the strategies. A real price of $100 per tonne, received for processing Valencias in the Riverina in 1997, was also considered in the deterministic analysis.

Table 2 : Nominal and Real Prices for Valencia and Navel Orange

The average on-farm price from these datasets in 1997 dollars for Navels and Valencias were $319/tonne and $217/tonne respectively. We would expect that the processing market would provide a floor for the fresh market and Figure 1 provides some support for this. Since the early ‘80s both markets but particularly the fresh navel market, appear to have become more volatile. The assumptions made in transforming wholesale to on-farm prices of navels are based on a high level of management. The fifty percent premium (approximately) assumed here for navels is quickly reduced as the proportion of navels not suitable for the fresh market increases.

For mature plantings production costs are approximately $3,500 and $4,300 per hectare for processing Valencias and fresh Navels, respectively. Hence the respective gross margins for mature plantings of Valencias and Navels are around $1,800 and $2,500 per hectare. It is important to note that production costs influence the final gross margin and include variables such as harvesting and marketing levies which are dependant on yield profiles.

4. Methodology

Benefit/cost analysis underpins the analyses of the three replacement strategies. The strategies are compared at average real prices using the net present value criterion at an interest rate of eight percent. Then Monte Carlo simulation techniques in @RISK are used to examine the implications of price uncertainty.

Clearly oranges are a perennial crop and the time period over which rotations are run is long. In the first instance this has required the construction of development budgets for each strategy and these are presented in Tables 3 to 5 (in Appendix) . Apart from their use in investment analysis the development budgets give information about the pattern of cash flow, financing requirements and pay back periods. As taxation and borrowings are likely to vary significantly between individual situations, no allowance has been made for these factors in the analysis. Prices and costs in Year 1 are assumed to be in real dollars and hence there is no allowance for inflation in the budgets. Similarly no allowance is made for the likely impact of new technology on yields.

The rotations associated with the different strategies vary in length. To be a valid comparison the Faustman formula was applied to get the present value of an infinite series of each rotation. If the net present value, NPV, of a rotation of A years is M then the NPV of an infinite series of rotations is given by (Pearse, 1990, P.138) where the denominator is the Faustman factor and is one less than the discount factor from a standard discount table.

The procedure for the first strategy, starting with Valencias, was to estimate the present value in year 49 (where at year 48, the Valencia tree is 60 years old) of a complete rotation and then using the Faustman factor estimate the value in year 49 of an infinite series of this rotation. The present value of the infinite series of rotations 60 years in length was discounted back to year 1 and added to the present value in year 1 of the stream of income and costs up to year 49. For the second strategy the present value of an infinite series of Navel rotations 20 years in length was calculated. For the reworking strategy we assumed that the reworked stand had to be replanted in year 16 and then reworked again in year 36. We estimated the present value of a replant - rework rotation in year 16 and applied the Faustman factor to this. The present value of the infinite series of rotations 34 years long was then discounted back to year 1 and added to the present value of the stream of income and costs up to year 16.

5. Risk Analysis

Initially the analysis of the three alternatives was deterministic in nature and based on average prices. However these strategies involve price (and yield) risk. On the price side, while under good management the average price of Navels is higher than that of Valencias, they have been more volatile in recent years and the price basis may change in favour of Valencias. It has already been noted that the yield from reworked Navels under Australian conditions is uncertain in the long-term. These factors may influence decisions by risk averse growers.

If producers are risk averse, then their choice of strategy will depend not just on expected NPV but also on the distribution of NPV. In this analysis Monto Carlo simulation techniques within @RISK have been used to generate a distribution for the NPV for each strategy and these distributions have been compared using stochastic dominance techniques.

Such risk analysis requires the specification of probability distributions for prices and yield. One approach to generating a distribution for expected prices would be to subjectively define a distribution based on econometric studies of supply and demand in the orange industry and on outlook information. It is difficult to prepare accurate outlook information for annual crops let alone perennial crops such as oranges. One difficulty is the lack of recent research in demand and supply conditions in the domestic and international orange markets. On the supply side the most recent Australian study appears to be that by Alston, Freebairn and Quilkey (1980 ).

On the demand side we have been unable to find recent studies of the orange industry. At a more general level, Heifner and Kinoshita (1994) found that real prices for many agricultural commodities have declined over long periods, that price variability has changed for certain commodities, and that differences among commodities in price variability are persistent. Results indicated that several commodities exhibited price variability exceeding 20% during many decades. Included in these commodities were oranges which were shown to exhibit significantly higher than average price variability.

An alternative approach to defining price distributions, which was used here, is to base them on historical price series. The problem with using historical price series is that they reflect not only random exogenous factors but also structural changes in demand and supply. As the price series is extended backwards with a view to capturing the random variability in prices, it becomes less likely that price reflects current demand and supply conditions and those that are likely to apply in the future.

The price data used in the analysis were described above. In determining the distribution functions associated with each price data set, the Bestfit program was applied. Each price series is skewed to the right by occasional high prices and may be best described as a gamma distribution. Our sampling procedure has been to sample prices from the distributions for each simulation run in year 1 and then assume that this real price remained constant over the investment period. In the simulation analysis we assumed that there is no correlation between fresh Navel orange prices and processing Valencia orange prices, as the fresh market is largely driven by internal factors whereas the processing price is largely dependant on external factors. Each simulation exercise consisted of 500 runs.

6. Stochastic Dominance

When producers are risk neutral, investment alternatives can be ranked by net present value or some other benefit/cost criterion. However the general view is that producers are risk averse (Bardsley and Harris (1987) and Anderson and Dillon (1991) ) which means that producers base investment decisions not only on expected returns but on other moments of the distribution of returns such as the variance of returns. If the utility function for a producer is known and is specified in terms of the moments of the distribution of net returns, investments can be ranked by their contribution to utility. Rarely are we in this situation. An alternative risk analysis technique is stochastic dominance testing which can be used to identify efficient and inefficient sets of investment alternatives for producers whose risk preferences fall within specified bounds.

Given two actions A and B defined by their cumulative density function (CDF) of outcomes, A is preferred to B in terms of first-degree stochastic dominance (FDSD) if CDF (A) £ CDF(B). Graphically this means that the CDF of A must always lie below the CDF of B.

Referring to Figure 2 and applying first-degree stochastic dominance, A dominates B, but A and C cannot be ranked, and B and C cannot be ranked. The first-degree stochastically efficient set therefore comprises A and C. First-degree stochastic dominance has limited discriminatory power because it includes all those who prefer more to less.

Figure 2 CDF's and Stochastic

Second-degree stochastic dominance (SSD) has more power because it applies the additional assumption that the decision maker is risk averse (Anderson et al 1988) or that the marginal utility of income declines as income rises. Under these assumptions, a proposal is dominant if the area under the CDF (cumulated from the right) is less than the alternatives with the additional provision that the income from the dominant proposal at low levels of probability (cumulative) is greater than the alternatives. This latter condition is necessary because, under the assumptions of this test, the marginal utility of income at low levels of income is high and hence the area under the CDF test is only unambiguous under these circumstances.

If SSD does not hold, then further testing is required. For the examples in Figure 2 it is clear that A dominates C by SSD but it is difficult to distinguish between B and C because their CDFs intersect in such a way that B cannot dominate C for all values of x . As an aside, we know that A dominates B by FSD and hence the net result is that the efficient set comprises only A.

However the discriminatory power of SSD is still low. Stochastic dominance with respect to a function, SSDF, or generalised stochastic dominance, GSD, (Meyer, 1977) makes the further assumption that the risk preferences of producers fall within certain bounds where these preferences are defined by Pratt's coefficient of absolute risk aversion. By making assumptions about the range of risk preferences, rather than identifying a unique representation of risk preference, stochastic dominance testing sacrifices the pursuit of a unique optimal decision for a search for the most efficient set of decisions (Anderson et al 1988) . The more we restrict the nature of risk preferences, the more powerful stochastic dominance testing becomes and the smaller is the risk efficient set of investment options to choose from. The cost is that the range of risk preferences used represents a smaller proportion of decision makers. Estimates for the coefficient of absolute risk aversion were made by Bardsley and Harris (1987) and Anderson and Dillon (1991) . Following Jones et al (1996) , SDWRF could have been applied in this study using two ranges in r(x), the coefficient of absolute risk aversion, reflecting a more and a less risk averse decision maker. Jones derived these measures for farmers with an asset value of $1m and this level of assets is likely to be typical of citrus growers in the area we are interested in. As can be seen from the results we did not have to apply SDWRF. The program, Generalised Stochastic Dominance, (Cochran and Raskin 1988) is often used to test for stochastic dominance. McCarl (1990) is another important contributor in this area.

7. Results

Starting with a deterministic analysis based on expected prices, the present value for the infinite series for a full rotation of Valencias is $22,921, and is $31,794 for the second strategy, a full rotation of Navels. For the third strategy, the reworking rotation, a present value for the infinite series of $46,750 is projected. A discount rate of 8% is used for all three strategies. Using a real price of $100 per tonne for Valencias the present values for the series of rotations for Strategy 1 and 3 are $-11,926 and $43,118 respectively. Hence, it is evident that the third strategy involving reworking is the preferred strategy when growers are risk neutral.

The results of risk analysis are presented in the form of a cumulative distribution function (CDF) for the discounted sum of the projected cash flows (NPV) in Figure 3 and the actual NPV for selected points within the distributions are presented in Table 6 . The CDF (figure 3) illustrates the probability of a result less than or equal to any value in the range. For example, the CDFs indicate that there is a 20% chance that NPV will be less than or equal to approximately $5,700, $9,300 and $21,900 for the Valencia, Navel and reworking strategies respectively.

Figure 3 - Cumulative Distribution of the three Re-working Strategies

Table 6 : The Distribution of NPVs

The ranking of the strategies by mean and maximum NPVs is the same as for the deterministic analysis. However, while the NPV from the reworking strategy is greater than that from the Navel strategy at every level of probability - the CDF for the reworking strategy is always to the right of that for the Navel strategy - the CDF for Valencias crosses those for the other two strategies at low levels of probability. The minimum NPV for Valencias is a loss of $21,700 whereas the losses from the reworking and Navel strategies are $31,300 and $45,900. Hence while we can say that the reworking strategy is dominant to the Navel strategy at the first degree we cannot say it dominates the Valencia strategy to this degree nor can we say that the Navel strategy dominates the Valencia strategy.

Because the NPV of the reworking strategy is less than that of the Valencia strategy at low levels of probability, technically the strategies should be compared using the SSDF procedure rather than the simpler SSD test based on areas under the CDFs. However in this case the CDFs crossed at a probability level of about 0.08% which says that there is only a probability of about 0.08% that the NPV from the Valencia strategy will exceed that from the reworking strategy. Hence we have made the judgement that the reworking strategy is dominant to the Valencia strategy for all but a few extremely risk averse growers. We did not proceed with the SSDF test.

8. Conclusions

This paper has compared three alternative investment options specific to replanting and reworking strategies in the orange industry. Initially benefit-cost analysis using NPV as a decision criteria was applied in a deterministic framework. The NPVs for this analysis indicated that a reworking strategy was preferred to that of a full rotation of either Valencias or Navels. Navels were preferred to Valencias because of higher average prices and the reworking strategy had the additional attraction that yield losses during re-establishment were lower. The level of management of navels is an important determinant of the extent of the price basis between Navels and Valencias. Implications of price uncertainty were explored using risk analysis in conjunction with stochastic dominance testing. For all but extremely risk averse growers the reworking strategy was the preferred option.

The results of this paper have depended on assumptions about prices and yields. It has been noted that there is a lack of information concerning the outlook for FCOJ, which is an area for future research. Further, the results of this analysis have depended critically on assumptions about yields from reworking Valencia trees to Navels. An additional area of future research may include risk analysis incorporating yield uncertainty.

In this paper we have compared three strategies among many for replanting in the orange industry. Clearly no claim can be made other than that the reworking strategy we identified is dominant to the other two. Further research could examine optimal replacement policies for orchards , which is a similar issue to that of optimal machinery replacement (Burt, 1965) .

References

Alfons , W., Walker, M., Swanton, C., and Shaw, J. (1992), ‘Economic comparison of alternative tillage systems under risk', Canadian Journal of Agricultural Economics 40, 199-217.

Alston , J.M., Freebairn, J.W., Quilkey, J.J. (1980), ‘A model of supply response in the Australian Orange Growing Industry', Journal of Agricultural Economics 24(3), 248-267.

Anderson , J.R. (1975), ‘Programming for efficient planning against non-normal risk', Australian Journal of Agricultural Economics 19(2), 94-107.

Anderson, J.R. and Dillon, J.L. (1991), Guidelines on Incorporating Risk in Farming Systems Analysis for the Development of Dryland Areas , FAO of the United Nations, Rome.

Antle , J.M. (1993), ‘Incorporating risk in production analysis', American Journal of Agricultural Economics 65(5), 1099-1106.

ABARE (1996), Australian Commodity Statistics , Canberra.

Bardsley , P. and Harris, M. (1987), ‘An approach to the econometric estimation of attitudes to risk in agriculture', Australian Journal of Agricultural Economics 31(2), 112-126

Bevington , K., Sanderson, G. and Barkley, P. (1991), ‘Topworking of citrus trees', NSW Agriculture.

Burt, O.R. (1965), ‘Optimal Replacement Under Risk', Journal of Farm Economics 47, 324-346.

Button , J. (1975), ‘The topworking of citrus', Citrus and Sub-tropical Fruit Journal 495, 5-8.

Cullis , B.R. and Verbyla, A.P. (1992), ‘Nonlinear regression modelling and time dependent covariants in repeated measures experiments, Australian Journal of Statistics 34(2), 145-160.

Eaves , Daryl. (1992), ‘Improving vineyard profitability with better irrigation scheduling: an example from the Riverland', The Australian Grapegrower and Wineries Annual Technical Issue.

Elton , K.J. (1997), Farm Budget Handbook - NSW Citrus, NSW Agriculture.

Etherington , D.M. (1997), ‘A stochastic model for the optimal replacement of rubber trees', Australian Journal of Agricultural Economics 21(1), 40-59.

Flemington Market Reporting Service . (1997), Flemington Wholesale Market Prices 1967/68-1996/97, NSW Agriculture .

Florida Department of Citrus . Link to Agriculture Statistics Page ( www.ultimatecitrus.com ).

Garner , R.J. (1979), Grafting Established Trees , Fourth Edition, BAS Printers Ltd, UK.

Hardaker , B., Pandey, S. and Patten, H. (1991), ‘Farm planning under uncertainty: a review of alternative programming models', Review of Marketing and Agricultural Economics , 59(1), 9-22.

Heifner , R and Kinoshita, R. (1994), ‘Differences among commodities in real price variability and drift', Journal of Agricultural Economics Review 45(3), 10-20.

Jones , R. (1997), ‘A Review of Risk and Stochastic Methodologies for Agricultural Research and Policy Analysis', Research Report.

Jones, R., Wall, L., and Marshall, G. (1996), ‘A stochastic economic analysis of on-farm drainage recirculation in the Berriquin Irrigation District', Agricultural Economics Bulletin, No. 13, NSW Agriculture, 1996.

McCarl , B.A. (1990),'Generalised stochastic dominance: An empirical examination', Southern Journal of Agricultural Economics , 22(2):49-55.

Meyer , J. (1977), ‘Choice among distributions', Journal of Economic Theory 14, 326-36.

Palisade Corporation (1997), @RISK: Risk Analysis and Modelling, New York.

Pannell, D.J., Malcolm, B. and Kingwell, R.R. (1995), ‘Are we risking too much? Perspective's on risk in farm modelling and farm management' Paper presented to the 39 th Annual Conference of the Australian Agricultural Economics Society, University of Western Australia, February.

Pearse's , P.H. (1990), Introduction to Forestry Economics, UBC Press.

USDA statistics . http://www.usda.gov/nass/aggraphs/citrus.html

Appendix: Technical Aspects of Reworking

There are several techniques of reworking.

Budding Budding may be done into strong regrowth shoots forced by cutting back the tree or directly into the scaffold branches of young trees. Budding may be carried out in either spring or autumn.

Limb grafting Limb grafting is usually done in the spring as soon as bark slips. The tree is cut back to 4-6 main scaffold limbs at a convenient working height and grafted using 2-3 bark grafts per limb. Budsticks are typically 5-7mm in diameter and 10-15cm long with 3-4 healthy buds. The tree is cut back on the eastern side and only 1-3 limbs grafted. The remaining canopy on the western side provides protection for the newly developing shoots and maintains cash flow during the first two years after grafting. Improved fruit size is generally observed on the old top following cutting back.

Crown grafting Crown grafting is carried out by cutting off the top of the tree just below the main scaffold branches. The method of grafting is the same used with limb grafting. Depending on the diameter of the stump, 3-4 grafts are equally spaced around the trunk.

Reworking is an attractive alternative to replanting because it makes use of already productive rootstock and hence reduces the losses associated with replanting.

Reworking is probably suited to relatively young healthy orchards and has the potential to give rapid change over fro m one variety to another at a reasonably low cost, assuming there are no adverse virus infections, trees are healthy and existing row directions and tree spacings are those needed for an efficient, highly productive orchard. Reworking allows the rapid growth of the new scion variety, resulting in the loss of only one or two crops from those trees. Furthermore, alternate trees can be left productive until the reworked ones produce adequate crops. The remainder can then be reworked if so desired. With older orchards, reworking small areas may be an effective means of generating cash flow but should not be considered as a long-term alternative to a replanting program.

Production outcomes of reworking Valencias in Australia are not known. Further direct comparisons in Australia have been made between the different methods of reworking. Most reports suggest that a reasonable crop could be expected in the third year after reworking, and good crops by the fourth or fifth years (NSW Agriculture). In Spain, trees reworked by budding are reported to reach 70% of former production levels by the third year and 90% by the fourth year after reworking. In contrast to these reports much slower recovery has been observed with trees reworked to the late Navel selections by limb grafting in Australia.

Table 3 . Strategy 1 - Full Rotation of Valencias

PV for infinite series of Rotation 1 $22,921

Table 4 : Strategy 2 – Full Rotation of Navels

PV for infinite series of Rotation 1 $31,794

Table 5 . Strategy 3 - Reworking Rotation