Forthcoming publication in the Proceedings of the ISHS XIIIth International Symposium on Horticultural Economics, August 1996.

Hoy Carman and Roberta Cook

Department of Agricultural Economics, University of California, Davis, CA 95616

1.2. Abstract

Lifting the long-standing ban on imports of avocados produced in Mexico will have economic impacts on California producers. Increased imports will almost certainly occur, given the size and cost structure of the Mexican industry, and any increase will decrease both prices and total crop revenues for U.S. producers. Over time, the reduction in average avocado prices, revenues per hectare, and profit expectations will lead to a smaller domestic industry as new plantings decrease and tree removals continue. A recursive model of bearing area adjustments to alternative levels of imports ranging from a base of 22.7 to a high of 226.8 grams per capita is used to simulate changes in average prices and bearing area to the year 2010, given a set of assumptions concerning changes in determinants of demand. Under specified conditions, imports of 226.8 grams (.50 pounds) per capita decreased average prices by almost 17 percent below the base and bearing hectares by over 18 percent below the base in 2010.

1.3. Additional Index Words

Avocado supply response, demand characteristics, simulation model.

1.4. Introduction

Mexican avocados have been banned from the United States since 1914 because of the presence of pests, such as the seed weevil, which could devastate the California avocado industry. The U.S. Department of Agriculture is currently considering a controversial proposed rule that would allow Mexican avocados to be imported into 19 Northeastern states during a period from November to February. A decision on the proposed rule is expected soon.

While public hearings on the proposed rule have focused on the potential impact of introducing new pests into the California avocado industry, opening the U.S. market to Mexican avocado imports will have both short- and long-run economic impacts on U.S. producers, even without the spread of Mexican avocado pests. Avocados are a specialty good with few close substitutes that account for a very small proportion of the average consumersí budget. Given these characteristics, it is not surprising that demand is inelastic at the producer level. Thus, increasing the quantity of avocados placed on the U.S. market will reduce U.S. producer prices and total crop revenue. Consumers, on the other hand, could benefit from potentially lower retail prices and improved quality in the winter months.

This study illustrates the dynamic nature of avocado industry adjustments to changing economic variables in a competitive industry. Building on previous estimates of demand and supply relationships, this analysis focuses on numerical evaluation of possible economic impacts of Mexican avocado imports on the California industry under alternative scenarios. Inverse demand and acreage response functions for California avocados are estimated and combined into a recursive model that is used to simulate the impact of various levels of imports on California avocado planted area and revenues over time.

1.4.1. U.S. and Mexico Avocado Production

California typically produces 85 to 90 percent of the U.S. avocado crop with Florida accounting for the remainder. Avocados are a high value crop, with annual sales revenue ranking well within the top ten California fruit and nut crops. Because of weather constraints, California avocado production tends to be concentrated near the coast in Southern California, with San Diego county accounting for about one-half of total hectares. The California avocado industry has experienced significant growth in planted area, production and demand since the 1960ís. In 1960, California had just over 8,100 bearing hectares with annual production of about 32,205 metric tons, while in 1987/88 bearing trees peaked at 30,880 hectares with production totaling just over 162,386 metric tons (California Avocado Commission). Recent estimates indicate that the bearing surface decreased about 6,770 hectares from the 1987/88 peak to 24,110 hectares in 1995/96 as a result of low prices during the 1980's, increased operating costs (especially high water costs since 1990), and urban pressures. The Census of Agriculture (1992) reported that 5,973 California farms harvested 27,320 hectares of avocados that year, resulting in an average of 4.57 hectares per farm. More than half (58.65%) of the farms harvested fewer than 2.02 hectares (5 acres) of avocados and these farms accounted for only 10 percent of total harvested acres. At the other end of the distribution, 215 farms (3.6 %) with more than 20.23 hectares (50 acres) of avocados accounted for 42.8 percent of total acreage.

The U.S. demand for avocados has increased over time as a result of increased consumer income, population growth and industry advertising and promotion. The total value of the crop increased from $9.9 million in 1960-61 to over $251 million in 1993/94. U.S. per capita consumption of avocados increased from less than 0.20 kilograms in the early 1960's to over 1.0 kilogram in 1987 and again in 1993. California growers supported industry research and promotion under a state marketing order from 1961 through 1977 and have operated under the California Avocado Commission Law since 1978. Recent research indicates that producer financed advertising and promotion totaling over $110 million since 1961 increased avocado demand, prices, and hectares over time (Carman and Green, 1993).

Mexico is the worldís largest avocado producer and consumer. Production totaled over 724,500 metric tons in 1992 from just over 87,500 hectares of bearing trees (SARH, 1993). Per capita consumption of avocados in Mexico increased from over 3 kilograms in 1970 to over 8 kilograms in the late 1980ís (Paz-Vega, 1989). While avocados are grown throughout Mexico, production is concentrated in the state of Michoacan, which accounted for 599,268 tons (83%) of total 1992 Mexican production. Michoacan had 74,487 hectares of avocados in 1992, with 70,340 bearing and 4,147 nonbearing. Michoacan average yields were 7.61 metric tons per hectare in 1992 while California yields were 4.99 tons per hectare. In addition to Michoacanís sizable yield advantage over the California avocado industry, it also has substantially lower water and labor costs. Interviews with trade sources estimate landed costs for Michoacan avocados at the U.S. border (Laredo, Texas) at $.80 to $.89 per kilo, exclusive of duty. This compares to f.o.b. costs for packed fruit in California ranging from $1.12 to $1.61 per kilo. In California, significant variation in water costs, farm size and yields contribute to this large variation in costs.

1.4.2. U.S. Avocado Imports

U.S. avocado imports are typically quite small. For example, during the period from 1962 through 1990, imports on the average accounted for just over one percent of annual U.S. avocado consumption. Increased levels of imports occurred from 1991 through 1994 with the largest amount of imports (24,145 metric tons) occurring in 1992 when both California and Florida had small crops. Imports are seasonal with most entering from September through December, when California production is the lowest. Chile is the main source of imports, with the Dominican Republic typically the second largest supplier.

Mexico exports 3 to 5 percent of its annual crop and less than 15 percent of recent production is judged to be of export quality (Paz-Vega, 1989). Industry sources, however, indicate that 15 to 30 percent of its crop could be export quality given sufficient economic incentives. Currently, demand for Mexican fresh avocados in the export market is limited by (1) competition from Spain and Israel, and (2) lack of access to the U.S. market.

1.5. Analytical Framework

Specification of an empirical model to estimate the possible impacts of increased imports on the California avocado industry requires estimated relationships for California acreage response (supply adjustments) and the demand facing producers. A previously estimated model of the California avocado industry that was used to evaluate the impacts of advertising and promotion programs (Carman and Green, 1993) is revised and updated for this study.

1.5.1. Avocado Supply

Annual avocado production is the product of bearing area and average yield. While average yields vary significantly from year-to-year as a result of weather and other factors, bearing area tends to trend up or down over time as a result of planting and removal decisions that depend on profits and profit expectations. The theoretical framework for models of perennial crop producer supply response has been developed and tested for several crops. A summary of empirical studies, modeling approaches and estimated relationships is presented in Alston, et al. (pp. 16-22). An example for avocados, which outlines the determinants of the plantings and removals relationships used in this study, is found in Carman and Green (1993). Following is a brief summary of the major components of the bearing area response model.

The bearing area of avocados at any point in time is based on previous plantings and removals decisions. This relationship can be expressed as:

BAt = BAt-1 + NPt-k - Rt-1 (1)

where BA is bearing area, the subscript t or time designates the year, k is a lag of k years required from the time a tree is planted until it reaches bearing age, NP is area of new plantings and R is area removed. There are two common approaches for estimating bearing area in year t, depending on the nature and quality of the data available. If detailed and accurate planting and removal data are available, one can directly estimate the new plantings and removal relationships. If accurate planting and removal data are not available, as is the case with avocados, it is possible to focus on estimating the annual change in bearing area, and indirectly account for plantings and removals by specification of lagged variables related to each. The relationship to estimate is derived by subtracting BAt-1 from both sides of equation (1), which yields

ÆBAt = NPt-k - Rt-1 (2)

where the annual change in bearing area, BAt - BAt-1, is defined as ÆBAt .

Estimation of equation (2) involves the separate specification of variables related to plantings and removals and time lags that determine their impact on the change in bearing area. The procedure used is to specify a plantings equation and a removals equation and then substitute the independent variables with time lags for NP and R in equation (2).

Plantings of new avocado trees during any year is based on the expected profitability of avocado production over the life of the trees. Since expectations cannot be observed, estimation of a plantings equation requires specification of a set of observable (proxy) variables related to expectations. The new plantings equation for avocados was:

NPt = f(TRAt-1,m, TAX) (3)

where TRA is a lagged moving average of farm level total revenue per acre for avocados deflated by the index of prices paid by farmers for all commodities, services, interest, taxes and wage rates (with the length of the moving average m to be based on the data) and TAX is a zero-one variable to capture the impact of income tax law changes. Given that profit expectations are based on recent experience, new plantings were expected to increase as average real returns increased. TAX is a zero-one variable that captures the effect of income tax rules that reduced the after-tax cost of avocado orchard development and encouraged new plantings during the period from 1970 through 1976. Attempts to specify other proxy variables related to the new planting decision have not been successful.

Avocado trees are removed in response to low profit expectations, which may be due to disease, tree age, low prices and yields, increased costs, and opportunities to convert the land to higher value uses. Because of unreliable estimates of annual removals, most studies have experienced difficulty in isolating statistically significant variables related to removals and have used a constant percentage of acres or a measure of profitability at the time the removal decision was made. The removals function was specified as:

Rt = f(TRAt-1,m) (4)

where the variables are as defined above. One expects removals to increase when returns are decreasing and to decrease when returns are increasing.

Substituting the explanatory variables for planting (NP) and removals (R) with appropriate time lags into equation (2) yields the change in bearing area relationship to be estimated, which was:

ÆBAt = f(TRAt-4,m, TAX, TRAt-1,m) (5)

The major time lag in the specification is for the time required from making a planting decision to classification of the tree planted as bearing. While the time required to reach bearing varies by variety, three years is typical for the varieties that account for over 85 percent of acreage. Since it is not unusual to have another year delay between the planting decision and actual planting as a result of land preparation and acquisition of nursery stock, the usual lag between the decision to plant an avocado tree and the time it reaches bearing age can be four years (k=4).

Average annual yields and bearing area combine to determine total production during any year. The pattern of average yields over time was examined but no trend was evident. In addition, there is very little that producers can do to alter average yields during a given year. Thus, either actual or average yields can be used for projections.

1.5.2. Avocado Demand

Avocados are an annual crop which are largely consumed fresh and not stored from one crop year to the next. Thus, production is essentially predetermined for a given year and the inverse demand model with price as the dependent variable is appropriate. In general, the demand for avocados is expected to be of the form:

Pct = f(Qcat, Qot, Yt, At) (6)

where the real price of California avocados in year t (Pt) is a function of total sales of California avocados (Qcat), total sales of other avocados including Florida production and imports (Qot), real consumer income (Yt) and real advertising expenditures (At), with each of the variables on the right hand side of the equation expressed in per capita terms. Close substitutes or complements for avocados have not been identified.

1.6. Estimation Results

The change in bearing hectares equation (5) and the price equation (6) were each estimated using data for the 33-year period 1961-62 through 1993-94. Both equations were estimated used customary U.S. weights and measures (price per pound and acres) and the results were converted to metrics.

1.6.1. Avocado Supply

The change in bearing acreage equation was estimated using lagged averages of deflated total revenue per acre that ranged from two to five years as the variable related to avocado plantings and removals. The five-year average for new plantings and the two year average for removals provided the best statistical results. The estimated change in bearing acreage equation is:

ÆBAt = -4044.94 + 1.98 TRAt-4,8 + 1598.99 TAXt-6 - 0.29 TRAt-1,2 R2 = .89 (7)

(-5.52) (7.66) (2.23) (-1.73) D.W. = 1.61

where the values in parentheses are the t-statistics for each coefficient. The results are comparable in magnitude to previous results (Carman and Green, p. 613), except for the negative coefficient for the removals part of the expression. The lagged average of deflated (by the index of prices paid by farmers, 1992 = 100) total revenue per acre [TRAt-4,8 = {TRAt-4 + TRAt-5 + TRAt-6 + TRAt-7 + TRAt-8}/5] has the hypothesized positive impact on plantings and is statistically significant. The estimated coefficient on the TAX variable indicates that the shift of investor interest (due to restrictions on citrus and almonds) increased total avocado plantings by 9,594 acres for the six year period 1970-71 through 1975-76 (almost 1,600 acres each year) relative to the "no tax law" period. The impact of taxes on change in bearing acreage was lagged four years because of the delays noted above. The estimated coefficient for the removals part of the expression [TRAt-1,2= {TRAt-1 + TRAt-2}/2] had an unexpected negative sign but was not statistically significant. As in previous studies, a portion of annual removals are probably included in the constant term. Unexplained variation in the change in bearing acreage equation is relatively small, especially in light of the simple nature of the expression and the data problems associated with measuring annual plantings and removals of avocados.

1.6.2. Avocado Demand

The inverse demand function specified in equation (6) could be estimated using any of several functional forms. While most studies of the impact of advertising on demand have utilized either a linear or double-log model, there was no a priori reason to select one over the other. This study estimated the demand function using the Box-Cox model, and then performed nested tests of the double-log and linear models as special cases of the Box-Cox model. The transformed equation estimated is

  (8)

where the starred variables are the Box-Cox transformations, i.e.,  and the other variables are transformed similarly. If l=1, the model is linear, while if l=0, the model is the double-logarithmic specification. Parameters for the equation were estimated using the Shazam computer program (White and Bui, 1988). The estimated equation, based on 33 annual observations extending from the 1961-62 through 1993-94 crop years was:

P*ct= 1.8285 -.5653Q*ct+ 0.00018Y*t+ 0.0151A*t l= -0.24 (9)

(16.12) (-13.29) (6.27) (2.20) R2<2 = .87,

where the figures in parentheses are the t-statistics for each coefficient. The dependent variable Pct is the average annual grower price (cents per pound) for California avocados in year t deflated by the consumer price index for all items (1975=100). Each of the estimated coefficients for the independent variables had the expected sign and each was significantly different than zero at a 95 percent or greater level. There is an inverse relationship between price and per capita sales of all avocados (Qt), as dictated by the law of demand, and the positive relationship between price and U.S. per capita disposable income deflated by the consumer price index (Y)t)) indicates that avocados are a normal good. The positive relationship between deflated advertising and promotion expenditures and price indicates that California Avocado Commission marketing expenditures have increased the demand and real price for avocados.

Since the estimated demand relationship treats quantity as predetermined for a given crop year, it is appropriate to consider price flexibilities of demand (the percent change in price divided by the percent change in quantity) rather than price elasticity of demand (the percent change in quantity divided by the percent change in price). The estimated price flexibility of demand with respect to quantity, calculated at average values for each of the variables, is -1.53. This indicates that producer prices are very responsive to changes in production, i.e., a 1.0 percent increase in quantity produced results in a 1.53 percent decrease in producer prices. It also indicates that an above average crop in terms of quantity per capita (other factors equal) will result in both lower prices and decreased total revenues. Using the inverse of the price flexibility as an estimate of the price elasticity of demand indicates that demand is inelastic ( -0.65) at the producer level. Note that the price flexibility with respect to advertising was 0.28 at average values. Thus, a 10 percent increase in advertising would shift demand enough to increase prices 2.8 percent for a given crop at average values of price and advertising.

A likelihood ratio test was used to test the non-linear Box-Cox specification against the double-log and linear alternatives. The linear model was rejected in favor of the Box-Cox specification but the null hypothesis that the model was a double-log specification was accepted. We use the Box-Cox specification for the simulation model since it provides superior statistical results. The Box-Cox and the double-log specification yielded similar results. The estimated constant flexibilities of demand for the price dependent log-linear equation with respect to each of the variables were: quantity, -1.51; income, 2.33; and, advertising, .30.

1.6.3 Avocado Supply Response

The price per pound and change in bearing acreage equations estimated above were combined into a recursive model of supply response. Once initial values for the exogenous variables were entered, the model generated values for annual bearing acreage, average price per pound, and total revenue per acre for California avocados. Model results were then converted to metric measures.

A comparison of actual and simulated values for bearing hectares shows that simulated hectares were above actual hectares from 1967 through 1986 and below actual hectares after that. Simulated bearing hectares reached a peak sooner than and below actual bearing hectares. The actual peak for bearing hectares was 30,880 in 1988, while the simulated peak of 29,694 hectares occurred two years earlier in 1986. Overall, the model performed well in tracing adjustments in bearing hectares which occurred during the 1961-62 through 1993-94 period.

To derive estimates of the impact of imports on prices and production of California avocados over time, the simulation model was run for a base scenario and varying levels of avocado imports for a 16 year period from 1993-94 through 2009-2010 under an assumed set of conditions. Assumptions used for the projections were (1) real per capita consumer income increases 1 percent annually, (2) population increases 1 percent annually, (3) Florida avocado production increases from 45.36 grams per capita in 1995 to 90.72 grams per capita in 2000 and then remains constant at 90.72 grams per capita (4) the Consumer Price Index increases 2.5 percent annually (5) the Index of Prices Paid by Farmers increases 1 percent annually (6) California avocado advertising is maintained at a real annual level of $7.09 per 1,000 population ($19.50 per thousand in 1993-94 nominal terms), and (7) average yields per hectare are the same as those that occurred from 1980 through 1995.

Both actual and simulated bearing hectares were declining at the end of the observation period in 1994-95, with actual hectares dropping below simulated hectares. Because of recent prices and returns, further decreases in bearing hectares are expected. Using the estimated coefficients in equation (7), changes in bearing hectares are expected to be zero when lagged real total revenue is $5,913 per hectare; values below $5,913 per hectare lead to reductions in bearing hectares while the industry is expected to expand bearing hectares for real total revenue above $5,913 per hectare (1992 prices). Actual annual real total revenues averaged less than $4,942 per hectare for the period from 1981-82 through 1991-92 and then increased to $5,886 per hectare in 1992-93 and $5,491 per hectare in 1993-94. Further reductions in bearing hectares are also confirmed by recent reported new plantings and nonbearing hectares that are much lower than at any time since 1961-62.

The simulation model was used to examine California avocado prices, bearing hectares and real total revenues per hectare for five alternative import scenarios ranging from a base of 22.7 grams per capita to imports of 45.4, 90.7, 136.1 and 226.8 grams per capita. Because of population growth, total imports increase through time for each scenario. For the base alternative of 22.7 grams per capita, imports increase from 5,987 metric tons in 1995 to 6,940 metric tons in 2010; for the 226.8 grams per capita scenario, imports increase from 59,783 metric tons in 1995 to 69,400 metric tons in 2010.

Using the base simulation, bearing hectares of California avocados decreased from 25,528 hectares in 1995 to a low of 23,538 hectares in 2002 (7.80% decrease) and then increased to 24,641 hectares in 2010 (an overall 3.84% decrease from 1995). Changes in bearing hectares for each of the five scenarios are summarized in Table 1. The negative impact of imports on bearing hectares and total crop returns increases as imports increase. For the scenarios with imports of 136.1 grams and 226.8 grams per capita, lagged domestic real total revenues were below the bearing hectares stabilizing level in 2010 and hectares continued to decrease, even with the demand enhancing effects of population and income growth. Comparing the simulation of 226.8 grams per capita imports with 22.7 grams per capita imports, we find that simulated real total revenue per hectare was 22 percent lower and bearing hectares were 14.8 percent lower in 2010 with larger imports. Nominal prices increased under each of the scenarios, with the amount of increase being an inverse function of the amount of imports. Nominal price per kilogram in 2010 was 24 percent higher with the base simulation of 22.7 grams of imports per capita than with 226.8 grams of imports per capita. Overall, simulated real total U.S. crop revenue in 2010 was 34 percent lower with 226.8 grams than with 22.7 grams per capita imports.

To evaluate the impact of differing levels of imports on average prices over time, the model was run using average yields for the period from 1961-62 through 1993-94 (5.99 tons per hectare) for each year rather than the actual pattern of yields for the most recent 16 years. Note that the constant yield assumption used for Table 2 resulted in a slightly greater reduction in bearing hectares in 2010 than did varying yields shown in Table 1. The simulated impact of imports on average real price varies with the level of imports. Real avocado prices increased 3.07 percent under the base import scenario of 22.7 grams per capita and almost one percent for imports of 45.4 grams per capita. Higher levels of imports resulted in decreased real prices, with the 2010 decrease estimated at almost 17 percent for imports 226.8 grams per capita.

1.7. Concluding Comments

Lifting the U.S. ban on imports of Mexican avocados will adversely affect domestic crop returns and the size of the California industry. Our analysis indicates that the demand for avocados at the producer level is relatively inelastic, which means that for a given crop year, increased sales from imports will decrease both average prices and total crop revenues. We also found that planted area responds to crop revenues, with relatively low average returns over the past decade being associated with decreasing hectares of avocados. Given the existing industry structure and the absence of significant imports, one would expect the continuing reduction in California bearing area to result in increased prices over time, which would stabilize and then increase planted area. With increased imports, however, downward pressures on prices and real total revenues will encourage continued reductions in the size of the California industry, and all producers can expect lower per hectare revenues.

Recent decreases in California avocado bearing area associated with sharply higher water costs may signal a structural change not included in the model. Future adjustments to water costs must be considered in applying the model results. If avocado trees continue to be removed due to increased water costs, then areas may decline below the levels simulated by the model. If the adjustment to higher water costs is completed, then the simulation results should be closer to actual changes.

Seasonal patterns of avocado production vary between California, Florida, and Mexico. More research is needed on the importance of seasonality in estimated demand relationships to better assess economic impacts that may vary by month. Concentrating imports during the winter months when California production is seasonally low, for example, may reduce the impacts on the California industry while placing strong downward price pressure on the Florida industry.

1.8. Acknowledgments

Helpful comments provided by Tom Bellamore, California Avocado Commission, and the anonymous referees are appreciated. The authors are solely responsible for any errors or omissions.

1.9. References

Alston, J.M., H.F. Carman, J.E. Christian, J. Dorfman, J.R. Murua, and R.J. Sexton. Optimal Reserve and Export Policies for the California Almond Industry: Theory, Econometrics and Simulations. Berkeley: University of California Agricultural Experiment Station, Giannini Foundation Monograph No. 42, February 1995.

Calif Agric Stat Ser. California Fruit and Nut Acreage. Sacramento, CA, annual issues.

Calif Agric Stat Ser. California Fruit and Nut Statistics. Sacramento, CA, annual issues.

California Avocado Commission. Annual Reports. Irvine, California.

Carman, H. F. and R. D. Green. "Commodity Supply Response to a Producer Financed Advertising Program: The California Avocado Industry." Agribusiness, 9(1993): 605-621.

Paz-Vega, Ramon. ìMexican Avocados: Threat or Opportunity for California?î California Avocado Society Yearbook, 73(1989): 87-106.

SARH. Nuario Estadistico de la Produccion, Gricola de los Estados Unidos Mexicanos, Septiembre 1993.

U.S. Council of Economic Advisers. Economic Indicators. Monthly issues.

U.S. Dept. of Agriculture, E.R.S. Fruit and Tree Nuts Situation and Outlook Reports. Annual yearbook issues.

U.S. Dept. of Commerce. 1992 Census of Agriculture. Vol.1, part 5, Oct. 1994.

White, Kenneth J. and Linda T.M. Bui. Basic Econometrics: A Computer Handbook Using Shazam. New York: McGraw-Hill, 1988.

1.10. Attachments