HD1761 
 G& 
 
 no. 344 
 
 St ack 5 
 
 UNIVERSITY OF CALIFORN A, DAVIS 
 
 75 02305 8830 
 
 r IANNINI ' FOUNDATIO N 
 
 OF AGRICULTURAL 
 ECONOMICS 
 
 UNIVERSITY OF 
 CALIFORNIA 
 
 The California Prune Board's 
 Promotion Program: 
 An Evaluation '^^a 
 
 Julian M. Alston, Hoy F. Carman, James A. Chalfant, John M. Crespi, 
 Richard J. Sexton and Raymond J. Venner 
 
 ^iannini Foundation Research Report Number 344 
 
 March 1998 
 
 DIVISION OF AGRICULTURE AND NATURAL RESOURCES 
 CALIFORNIA AGRICULTURE EXPERIMENT STATION 
 OAKLAND, CALIFORNIA 
 
THE CALIFORNIA PRUNE BOARD'S PROMOTION PROGRAM: 
 
 AN EVALUATION 
 
 The Authors: Julian Alston, Hoy Carman, James Chalfant, and Richard Sexton 
 are Professors in the Department of Agricultural and Resource Economics at 
 U.C. Davis. John Crespi and Raymond J. Venner are graduate students in the 
 U.C. Davis Department of Agricultural and Resource Economics. The authors 
 are Usted alphabetically, with no order of priority implied. 
 
 ACKNOWLEDGMENTS 
 
 The authors are grateful to the staff of the California Prune Board for contributing data to the study, and to Bart 
 Minor and Richard Peterson for insightful discussions on the economics of the prune industry. Sunsweet Grow- 
 ers also contributed key data to the study, and the cooperation of Sunsweet staffers Vivian Chase, Erin Hull, and 
 Bob Swensen is much appreciated. Thanks also to Jennifer James for comments on the manuscript, and to 
 Claudette Oriol, who prepared the final manuscript. 
 
EXECUTIVE SUMMARY 
 
 California is the world leader in prune production, 
 accounting for about 99 percent of U.S. production and 
 70 percent of the world's supply. The industry, through 
 the California Prune Board (CPB) and its various pack- 
 ers, especially Sunsweet Growers, the largest marketer 
 of California prunes, has invested substantially in the 
 promotion of prunes to consumers. This study ana- 
 lyzes the effectiveness of these expenditures in increas- 
 ing consumer demand for prunes and, thereby, in rais- 
 ing industry revenues. The results from this project 
 are useful for decision makers in the California prune 
 industry as well as to researchers studying the effects 
 of promotion on market demand. The analysis used 
 to derive the results is also pertinent to other Califor- 
 rua commodity groups, in light of increased scrutiny 
 surrounding generic promotion programs. The study 
 was conducted under an agreement between the CPB 
 and the University of California, and was carried out 
 by a research team of faculty and graduate students in 
 the Department of Agricultural and Resource Econom- 
 ics at the University of California, Davis. 
 
 The study involved econometric analyses of U.S. 
 domestic demand for California prunes. Economic 
 theory implies that, to be effective, expenditures on 
 promotion must increase consumers' demand for the 
 product being promoted. Other factors generally con- 
 sidered to influence demand, and which need to be 
 incorporated into a demand study, include the price 
 of prunes, the prices of close substitutes or comple- 
 ments, measures of consumers' purchasing power, and 
 factors to account for any time trends or seasonality in 
 demand. 
 
 Three data sets were assembled to study prune de- 
 mand. The main data set consisted of 51 observations 
 on retail prune consumption and prices in the United 
 States, reported in monthly intervals for the period 
 September 1992 to July 1996. Expenditures on promo- 
 tion by the California Prune Board and by Sunsweet 
 Growers were closely matched to the four-week ob- 
 servations on sales for this period. A second data set 
 consisted of annual observations on domestic prune 
 shipments and prices for the period 1949 to 1995. The 
 measure of promotion in the annual model consisted 
 of annual real expenditures by the CPB and Sunsweet 
 on all types of domestic promotion. A third data set 
 consisted of the results of a test market analysis of tele- 
 vision advertising for prunes conducted in six U.S. cit- 
 ies. 
 
 Results from analysis of the monthly data indicate 
 that prune promotion has increased the demand for 
 prunes. Across several alternative model specifications 
 examined and reported in part 3, prune promotion 
 
 consistently had a statistically significant, positive 
 impact on retail prune sales. For the various models 
 estimated using ordinary least squares (OLS), the elas- 
 ticity of sales with respect to promotion generally 
 ranged from 0.17 to 0.22, while the promotion elastic- 
 ity in the model estimated using 2SLS was 0.21. This 
 means that a 10 percent increase in expenditures on 
 promotion would have increased sales about 2 per- 
 cent, holding price and other explanatory variables 
 constant. 
 
 The models based on the annual data series did 
 not perform as well. Promotion, measured in this case 
 by annual real expenditures by the CPB and Sunsweet 
 on all types of domestic promotion, generally did not 
 have a statistically significant effect on demand. Such 
 results were not believable, however, in light of diag- 
 nostic tests that we performed to evaluate our specifi- 
 cation of the structure of these annual demand mod- 
 els. The tests led us to conclude that — either because 
 of poor or missing data or an incorrect model form — 
 the models were not specified correctly. Thus, we were 
 unable to use the annual data in any meaningful way. 
 
 The television advertising test-market campaign 
 was conducted for 12 weeks in Fall 1990, with three 
 cities selected as test markets, and three used as con- 
 trols. The advertisements featured generic advertis- 
 ing of dried prunes. Our analysis of the test-market 
 data indicates that the television advertisements had 
 a positive and statistically significant effect on prune 
 demand both during the period of the advertising cam- 
 paign and during the post-test period. The model we 
 developed indicated that in-store displays, by them- 
 selves, had no impact on prune sales. 
 
 A simulation approach was used to translate the 
 effects of promotion on prune demand into estimates 
 of the resulting marginal benefits (the increase in in- 
 dustry revenues from an incremental increase in pro- 
 motional expenditures) to prune growers. Because of 
 our greater faith in the data underpinning the monthly 
 analysis of demand, the superior statistical perfor- 
 mance of models estimated using the monthly data, 
 and the congruence of these model results with the 
 results from the test-market analysis, we based our 
 simulation analysis on results from models estimated 
 from the monthly data. Because the statistical analy- 
 sis was restricted to demand modeUng, while the simu- 
 lation analysis required a complete model of the in- 
 dustry, including supply response, it was necessary to 
 construct a synthetic supply model and conduct simu- 
 lations for a variety of alternative supply specifications. 
 
 The marginal benefit-cost ratio for promotion of 
 California prunes was calculated. This ratio refers to 
 
 iii 
 
the net revenues generated from incremental expen- 
 diture on promotion, and hinges importantly on the 
 value of the price elasticity of supply, and on whether 
 growers bear the entire burden of funding the expen- 
 ditures or some of the burden is shifted to consumers 
 in the form of higher prices. Returns to growers from 
 allocating expenditures to promotion would be maxi- 
 mized by expanding expenditures until the marginal 
 (last) dollar spent on promotion yields just a dollar 
 back in revenues. The analysis suggests that the in- 
 dustry stopped short of this optimizing condition dur- 
 ing the 1992-1996 period covered by the monthly data. 
 The calculated marginal benefit of an additional dol- 
 lar spent on promotion, given the amounts actually 
 expended, ranged from $2.65 to almost $30.00, sug- 
 gesting that additional promotion expenditures would 
 have generated positive net revenues to producers. 
 
 Only when producers are (implausibly) assumed to 
 bear the entire cost of the promotion is it possible to 
 derive average benefit-cost ratios less than 1:1, and to 
 do so requires an elasticity of supply of 1 .0 or more, 
 which is only likely to be relevant for longer-run 
 changes. 
 
 We conclude that promotion of California prunes 
 conducted by the CPB has increased the demand for 
 prunes and returns to producers of prunes. Over the 
 four-year period analyzed in the monthly model, in- 
 vestments by prune growers in promotion yielded 
 them marginal returns of at least $2.65 for every dol- 
 lar spent. Moreover, marginal benefit-cost ratios in 
 the range of 2.7:1 and higher indicate that the indus- 
 try could have profitably invested even more in pro- 
 motion during this period. 
 
 iv 
 
TABLE OF CONTENTS 
 
 1. INTRODUCTION 1 
 
 2. THE CALIFORNIA PRUNE INDUSTRY — 1949 TO PRESENT 2 
 
 2.1 Prune Production in California 2 
 
 2.2 Marketing California Prunes 4 
 
 2.3 Marketing Institutions for California Prunes 7 
 
 2.4 Trends in Factors Associated With Prune Demand 11 
 
 2.5 Concluding Comments on Supply and Demand Trends 13 
 
 3. ACCOUNTING FOR CHANGES IN AGGREGATE U.S. PRUNE CONSUMPTION 14 
 
 3.1 Aggregate Domestic Demand Models, Theoretical Considerations 14 
 
 3.2 Aggregate Monthly Domestic Demand Models 16 
 
 3.3 Aggregate Annual Demand Models, 1949-1995 29 
 
 4. SIMULATION MODEL AND BENEHT-COST ANALYSIS 33 
 
 4.1 Approaches for Evaluating the Benefits from Promotion 33 
 
 4.2 An Approximation Using Elasticities 35 
 
 4.3 Simulation Model 36 
 
 5. ANALYSIS OF TEST-MARKET STUDIES 41 
 
 5.1 Six-City Study of Television Advertising 41 
 
 5.2 In-Store Promotion 45 
 
 5.3 Conclusions from the Test-Market Analysis 47 
 
 6. CONCLUSION 48 
 
 APPENDIX TABLES 49 
 
 REFERENCES 66 
 
 V 
 
LIST OF FIGURES 
 
 Figure 2.1. California Prune Acreage and Production, 1949-95 2 
 
 Figure 2.2. Yields of California Prunes, 1 949-95 3 
 
 Figure 2.3. Grower Price of California Prunes, 1949-95 4 
 
 Figure 2.4. Value of Production of California Prunes, 1949-95 5 
 
 Figure 2.5. Exports and Domestic Shipment of California Prunes, 1949-95 6 
 
 Figure 2.6. Domestic Shipments of California Prunes, 1949-95 6 
 
 Figure 2.7. U.S. Per Capita Consumption of California Prunes, 1949-95 7 
 
 Figure 2.8. California Prune Board Promotional Expenditures, 1949-95 10 
 
 Figure 2.9. California Prune Board Expenditures on Domestic Promotion, 1949-95 12 
 
 Figure 3.1 . U.S. Per Capita, Monthly Prune Consumption (Retail), September 1992 to July 1996 16 
 
 Figure 3.2. U.S. Per Capita, Monthly Prune Consumption (Retail) - Actual versus Fitted Values, September 
 
 1992 to July 1996 26 
 
 Figure 4. 1 . Conceptual Supply and Demand Model 33 
 
 Figure 4.2. Incidence of Assessments 35 
 
 Figure 5.1. Percentage Change in Prune Sales in Test and Control Markets 43 
 
 Figure 5.2. Share of Prunes Sold with a Deal 43 
 
 vi 
 
LIST OF TABLES 
 
 Table 2.1. Prune Tree Sales in California 3 
 
 Table 2.2. U.S. Per Capita Consumption of Dried Fruit (lbs. per year) 8 
 
 Table 2.3. U.S. Per Capita Consumption of Fruit Juices (gallons per year) 9 
 
 Table 2.4. U.S. Per Capita Consumption of Dried Prunes in 1986, by Age Group 11 
 
 Table 3.1. Description of Variables in the Monthly Demand Model 18 
 
 Table 3.2. Summary Statistics of the Variables in the Monthly Demand Model 19 
 
 Table 3.3. Coefficient and Elasticity Estimates from the Monthly Demand Models 22 
 
 Table 3.4. OLS Estimates Comparing Different Specifications of Promotional Expenditures in the Preferred 
 
 Monthly Model 25 
 
 Table 3.5. 2SLS Estimates Comparing Different Specifications of Promotional Expenditures in the Preferred 
 
 Monthly Model 28 
 
 Table 3.6. Description of Variables in the Annual Demand Model 30 
 
 Table 3.7. Summary Statistics of the Variables in the Annual Demand Model 30 
 
 Table 3.8. Coefficient and Elasticity Estimates from Annual Prune Demand Models 32 
 
 Table 4.1. Approximation of /i, the Benefit from a Marginal Increase in Promotion Expenditure 36 
 
 Table 4.2. Marginal Benefit-Cost Ratios for Prune Promotion: A Comparison of Estimates from Four 
 
 Regressions using both OLS and 2SLS 38 
 
 Table 4.3. Marginal Benefit-Cost Ratio for a Supply Elasticity of One 39 
 
 Table 5.1. Test and Control Markets in the Television Ad Test Campaign 41 
 
 Table 5.2. Time Frame of the Television Advertising Test 41 
 
 Table 5.3. Overall Effect of Television Advertisements on Prune Sales 42 
 
 Table 5.4. Effect of Television Advertisements on Prune Sales in Individual Markets 42 
 
 Table 5.5. Description of Variables in the Television Ad Demand Model 44 
 
 Table 5.6. Summary Statistics of the Variables in the Television Ad Demand Model 45 
 
 Table 5.7. Econometric Results: Estimated Effects of Television Ads on Prune Sales 46 
 
 Table 5.8. Test Design of New PICS at Safeway Stores 47 
 
 vii 
 
LIST OF APPENDIX TABLES 
 
 Table A2.1. Selected Fruit Juices: U.S. Per Capita Consumption (in gallons) 49 
 
 Table A2.2 Selected Commercial Fruits and Vegetables (farm weight): U.S. Per Capita Consumption (in 
 
 pounds) 50 
 
 Table A2.3. Fresh Fruits (retail-weight equivalent): U.S. Per Capita Consumption (in pounds) 52 
 
 Table A2.4. Dried Fruits: U.S. Per Capita Consumption (in pounds) 55 
 
 Table A2.5. Domestic Shipments of California Prunes 56 
 
 Table A2.6. Expenditures by the California Prune Board 58 
 
 Table A3.1. U.S. Dried Prune Data Used in the Monthly Models (1992-96) 60 
 
 Table A3.2. U.S. Dried Prune Data Used in the Annual Models (1949-95) 62 
 
 Table A4.1. Marginal Benefit-Cost Ratios for Prune Promotion: A Comparison of Means from Simulations 
 
 Based on Four Regressions Using Both OLS and 2SLS 63 
 
 Table A4.2. Marginal Benefit-Cost Ratios for Prune Promotion: A Comparison of the Lower 95% Boundaries 
 of Four OLS and 2SLS Regressions 64 
 
 Table A4.3. Marginal Benefit-Cost Ratios for Prune Promotion: A Comparison of the Upper 95% Boundaries 
 of Four OLS and 2SLS Regressions 65 
 
 viii 
 
1. INTRODUCTION 
 
 The objective of this study is to evaluate the eco- 
 nomic impacts of industry-financed market develop- 
 ment and promotion activities for California prunes. 
 Using a state marketing order program first estab- 
 lished in 1952, all California prune producers pay man- 
 datory assessments that are administered by the Cali- 
 fornia Prune Board (CPB) to promote and increase the 
 demand for California prunes. The central questions 
 addressed by this study are: (a) how have CPB mar- 
 keting expenditures affected the demand for and sales 
 of California prunes, and (b) have net revenues to pro- 
 ducers from CPB marketing programs increased 
 enough to offset program costs? Answering these 
 questions requires the development of an economet- 
 ric model of the demand for prunes, using time-series 
 data on economic variables that have important ef- 
 fects on the consumption of prunes. 
 
 The report is organized in line with the steps taken 
 to answer the research questions. The first step is to 
 document developments with economic implications 
 for the California prune industry, and construct a data 
 base for the analysis. This is done in the next section 
 of the report (part 2), with documentation and dis- 
 cussion of the California prune industry in the post- 
 World War II era. Supply factors, including changing 
 patterns of production, planted area, yields, and vari- 
 
 eties are discussed first, followed by demand factors, 
 including changing domestic and international mar- 
 kets, prices, and patterns of consumption. Part 3 re- 
 ports the results of modeling the monthly and annual 
 per capita demands for prunes. The models are speci- 
 fied, the variables are defined and described, the mod- 
 els are estimated, diagnostic tests are performed, and 
 the results are interpreted. 
 
 The monthly econometric estimates are used as the 
 basis for a benefit-cost evaluation of the Prune Board's 
 marketing programs in part 4. Three approaches are 
 used to estimate marginal benefit-cost ratios for grower 
 expenditures on prune promotion. First, the elastici- 
 ties estimated in the monthly demand models are used 
 to approximate the ratios. Second, the estimated mod- 
 els are used to calculate marginal returns from pro- 
 motion, using a range of supply elasticities. Third, 
 Monte-Carlo simulations are performed, to obtain con- 
 fidence intervals on the benefit-cost ratios for each of 
 the estimated monthly models. An annual model is 
 also estimated, but is not used in the benefit-cost analy- 
 sis. Part 5 presents and discusses the results of some 
 test-market studies of the demand for prunes. Finally, 
 part 6 summarizes the main findings, interprets the 
 results, and presents conclusions. 
 
 1 
 
2. THE CALIFORNIA PRUNE INDUSTRY — 1949 TO PRESENT 
 
 Economic and crop statistics on developments and 
 changes in the California prune industry, especially 
 since 1949, provide the institutional background and 
 data required to develop the econometric models of 
 prune demand. 
 
 2.1 Prune Production in California 
 
 The California prune is a descendent of La Petite 
 d'Agen, a prune plum native to Southwest France. 
 Prune trees were introduced to the United States in 
 1856, when Pierre PelHer brought prune cuttings from 
 France to the Santa Clara Valley in California. Prunes 
 thrived under California climatic conditions, and the 
 industry was well-established and growing by the 
 early 1900s. The earliest California Department of Ag- 
 riculture acreage statistics for prunes recorded total 
 1919 acreage of 154,279 acres— 106,880 bearing and 
 47,399 non-bearing. Even though current acreage is 
 substantially less, California now accounts for about 
 99 percent of annual U.S. prune production, and an 
 average of about 70 percent of the world's annual sup- 
 ply of prunes (CPB 1996). 
 
 Trends in Acreage 
 
 California prune acreage has varied significantly 
 through time, but has remained rather stable over the 
 
 past decade (1985-1994). Initially concentrated in the 
 Santa Clara, Sonoma, and Napa Valleys, bearing acre- 
 age of California prunes peaked at 174,050 acres in 
 1930, then decHned steadily to 101,377 acres in 1950, 
 reached a temporary low of 80,122 acres in 1960, ex- 
 panded to 97,560 acres in 1969, decreased again to the 
 lowest recorded acreage of 65,400 acres in 1981, and 
 then expanded from 1982 to 1990 (figure 2.1). Since 
 1990, bearing acreage has varied from 77,588 to 80,100 
 acres, but it appears that acreage is increasing as a re- 
 sult of recent new plantings. Sales of prune trees by 
 California nurseries increased from 300,000 in 1989 to 
 over 1.5 million in 1994, and continued at high rates in 
 1995 and 1996 (table 2.1). At the same time, non-bear- 
 ing acreage of California prunes increased from an 
 estimated 6,000 acres in 1990 to 17,000 acres in 1993 
 (CPB, January 1996). 
 
 The location of planted acreage has also changed 
 over time. Most California prune orchards were lo- 
 cated in the Coastal Valleys before 1950, but popula- 
 tion pressures and more profitable alternative crops 
 led to the removal of the majority of that acreage. At 
 the same time, when prune trees were being removed 
 in the Coastal Valleys, new prune orchards were be- 
 ing planted in the Central Valley (Dale et al. 1988). 
 Now, prune acreage (and production) is concentrated 
 in the Sacramento Valley. Counties with more than 
 
 Figure 2.1. California Prune Acreage and Production, 1949-95 
 
 -S 
 
 to 
 
 § 250 
 
 Source: USDA, Agricultural Statistics. 
 
 2 
 
Table 2.1. Prune Tree Sales in California 
 
 Year 
 
 Number of Prune Trees 
 (1,000) 
 
 1989 
 
 300 
 
 1990 
 
 450 
 
 1991 
 
 550 
 
 1992 
 
 610 
 
 1993 
 
 900 
 
 1994 
 
 1,504 
 
 1995 
 
 985 
 
 1996 
 
 650 
 
 Source: California Prune News: Annual Report, 
 
 January 1996. 
 Notes: 1996 value is a projection. 
 
 3,000 acres of prune trees in 1992 included Butte 
 (10,439), Colusa (3,825), Glenn (7,561), Sutter (21,364), 
 Tehama (9,138), Tulare (5,571), and Yuba (10,857). The 
 1992 acreage in counties where prune production was 
 concentrated in 1950 includes Napa (93), Santa Clara 
 (1,937), and Sonoma (1,252). 
 
 Trends in Yields and Production 
 
 Data on average per acre yields of prunes during 
 the period since 1949 show an upward trend, with sig- 
 nificant year-to-year variation (figure 2.2). The annual 
 yield variability is primarily the result of weather con- 
 ditions and the alternate-bearing tendencies of tree 
 
 crops. Average yields of 2.21 tons per acre for the 5- 
 year period 1991-1995 were 38 percent higher than the 
 average of 1 .60 tons per acre for the 5-year period 1950- 
 1954. Possible reasons for the increase in average yields 
 include cultural methods (improvements in pruning 
 and tree spacing) and the changing location of pro- 
 duction (yields are higher in the Central Valley than 
 in the Coastal Valleys) (Dale et al. 1988). Variety im- 
 provements have resulted in improved yields for other 
 tree crops, but do not appear to have been an impor- 
 tant factor for prunes. The California prune industry 
 has relied almost exclusively on the French and Im- 
 proved French varieties, which have recently ac- 
 counted for about 97 percent of bearing acreage and 
 continue to account for the majority of non-bearing 
 acreage (1995 California Prune Acreage Survey). These 
 varieties have the desirable feature for mechanical 
 harvesting that the fruit remains on the tree when ripe. 
 Other commercial prune varieties, which account for 
 the remaining two to three percent of acreage, include 
 Friedman, Imperial, Robes de Sergeant, Moyer, Victor 
 LG, and 707. The CPB is funding research to support 
 development of new prune varieties that are earlier- 
 maturing than Improved French, yet retain its desir- 
 able product qualities. If successful, such a variety 
 will provide significant economic advantages from im- 
 proved utilization of harvest labor, equipment, and 
 processing capacity. 
 
 Significant year-to-year variation in total produc- 
 tion of California prunes results from the variability 
 of average yields (figures 2.1 and 2.2). The increased 
 average yields per acre during the 1950-1995 period 
 more than offset the observed reduction in bearing 
 
 Source: USDA, Agricultural Statistics. 
 
 3 
 
acreage. Total production increased from an annual 
 average of 157,200 tons during the 1950-1954 period 
 to an annual average of 172,600 tons during the 1991- 
 1995 period. A simple linear trend regression equa- 
 tion estimated for yields over the period 1949 through 
 1995 resulted in a trend coefficient of 0.017 (t- value = 
 4.29), indicating that average California prune yields 
 increased by 0.017 tons (34 pounds) per acre per year. 
 
 Trends in Prices and Value of Production 
 
 Average annual prices received by California prune 
 growers are shown in figure 2.3. These prices, in dol- 
 lars per ton of prunes in natural condition, are pre- 
 sented in both nominal and real terms. The real prices, 
 in 1995 dollars, were obtained by deflating nominal 
 prices by the Consumer Price Index (1995 = 1.00). 
 
 Variations in annual crop size have contributed to 
 fluctuations in the price per ton for prunes, with the 
 expected inverse relationship between average price 
 and annual quantity. For example, small prune crops 
 in 1958 and 1972 were associated with high prices. Real 
 prune prices generally declined until 1989, but have 
 increased slightly since then. The total value of Cali- 
 fornia prune production, in real terms, generally de- 
 clined until 1986, and has increased since 1987 (figure 
 2.4). 
 
 2.2 Marketing California Prunes 
 
 The marketing process for prunes has changed over 
 time, but the changes have been gradual rather than 
 
 dramatic. This section describes some prune industry 
 changes, outlines trends in prune consumption, and 
 discusses trends in some possible determinants of 
 prune demand. 
 
 Harvest and Processing 
 
 California prunes are typically machine-harvested 
 from mid- August to mid-September, with the harvest 
 date for a given orchard determined by a combination 
 of fruit firmness and sugar content. While prunes were 
 once harvested by hand in three or four "pickings" of 
 an orchard, today most California prunes are harvested 
 by machine, with a single pass through the orchard. 
 Immediately after harvesting, the ripe prunes are 
 washed, placed on large wooden trays, and dehy- 
 drated to about 21 percent moisture content. The typi- 
 cal conversion rate in processing is three pounds of 
 fresh fruit to one pound of dried prunes. After dehy- 
 dration, prunes are graded for size, inspected, and put 
 into storage in bulk containers; these dehydrated 
 prunes are stored in what is known as "natural condi- 
 tion." When the processor receives orders, the fruit is 
 removed from storage and partially rehydrated to 24- 
 30 percent moisture to make the prunes softer and 
 more pliable. The partially rehydrated prunes are then 
 sterilized, inspected, packaged, and shipped. The only 
 preservative used in processing prunes is potassium 
 sorbate (the potassium salt of sorbic acid), which is 
 used to protect against mold and yeast spoilage (CPB 
 1996). 
 
 Source: USDA, Agricultural Statistics. 
 
 4 
 
Figure 2.4. Value of Production of California Prunes, 1949-95 
 
 Trends in Consumption, Imports, and Exports 
 
 Domestic shipments and exports of California 
 prunes during the 1949-1994 period are shown in fig- 
 ure 2.5. Since 1949, domestic shipments of California 
 prunes have generally declined, while exports of Cali- 
 fornia prunes have tended to increase, in both abso- 
 lute and percentage terms. Currently, about 60 per- 
 cent of California's annual prune shipments are con- 
 sumed in the United States and about 40 percent are 
 exported, as compared with average shares of 83 per- 
 cent domestic and 17 percent export during the years 
 from 1949 through 1953. Domestic prune shipments 
 dipped to 85,000 tons in 1978, and remained below 
 100,000 tons annually until 1987, when they reached 
 103,000 tons. After peaking at 117,700 tons in 1989, 
 domestic prune shipments decreased to a little over 
 94,000 tons in 1995. California prune exports reached 
 71,000 tons in 1989 and peaked at almost 79,500 tons 
 in 1990. Exports have remained over 70,000 tons an- 
 nually since 1989, except during 1993, when exports 
 dipped to just under 61,000 tons. Germany and Japan 
 have recently been the leading importers of U.S. 
 prunes, followed by Italy, the United Kingdom, and 
 Canada. U.S. imports of prune products increased in 
 the early 1990s, but still remain very small relative to 
 California production. During 1984-92, U.S. imports 
 as a percentage of U.S. exports of dried prunes ranged 
 from 0.7 percent in 1987 to 2.4 percent in 1992 (USDA 
 Agricultural Statistics). 
 
 Prune Shipments by Product Category 
 
 The four major product categories for prunes, with 
 their 1995 shares of total domestic shipments, are: dried 
 pitted (50%), dried with pits (10%), juice and concen- 
 trate (38%), and canned prunes (2%). While the pro- 
 portion of the annual crop in each category has 
 changed over time, the most dramatic change has been 
 from dried prunes with pits to dried pitted prunes. 
 The pitted prunes share of total dried prunes remained 
 under two percent until 1961, increased to 12 percent 
 by 1965, and then increased rather steadily through 
 time to 85 percent in 1994 and 83 percent in 1995. This 
 increase in the market share of pitted prunes is the 
 result of improvements in pitting technology, which 
 permit removal of the prune pit with minimal skin 
 break and very little change in the shape of the prune. 
 Given the improved quahty of pitted prunes, consum- 
 ers in the United States and worldwide have been 
 switching from prunes with pits to pitted prunes. Do- 
 mestic shipments of California prunes by product cat- 
 egory are shown in figure 2.6. Since 1986, the Califor- 
 nia prune industry has shipped more pitted prunes 
 than prunes with pits. 
 
 Domestic shipments of prune juice and concentrate 
 have declined since the early 1960s. Prune juice is pre- 
 pared from a water extract of dried prunes and con- 
 tains not less than 18.5 percent by weight of water- 
 soluble solids extracted from dried prunes. Prune juice 
 may contain one or more optional acidifying ingredi- 
 
 5 
 
Figure 2.5. Exports and Domestic Shipment of California Prunes, 1949-95 
 
 Domestic 
 
 Source: California Prune Board, Annual Reports. 
 
 Figure 2.6. Domestic Shipments of California Prunes, 1949-95 
 
 90,000 
 
 Source: California Prune Board, Annual Reports. 
 
 6 
 
ents: lemon or lime juice or citric acid. Prune juice 
 may also contain honey, in a quantity not less than 
 two percent nor more than three percent by weight, 
 and may contain Vitamin C, not less than 30 milli- 
 grams nor more than 50 milligrams per 6 oz. serving 
 (CPB 1996). Prune juice concentrate is a viscous form 
 of prune juice, packed at 70° Brix (soluble soUds) mini- 
 mum, with higher Brix packs for export shipments or 
 on special orders. No preservatives are added to prune 
 juice concentrate, as the 70° Brix concentrate is self- 
 preserving (CPB 1996). These standards contribute to 
 a uniform product. 
 
 In the United States, relatively few prunes are eaten 
 fresh or canned. There are three standard types of 
 canned prunes. Regular canned prunes are fully 
 cooked in water, syrup, or their own juice. Nectar- 
 style canned prunes are distinguished by their high 
 drained weight, since they contain about one-third 
 more prunes per unit volume. Moist-pack canned 
 prunes are processed to 35 to 42 percent moisture and 
 sealed in the can, with no liquid and no preservatives 
 (CPB 1996). 
 
 U.S. per capita consumption of California prunes, 
 by product category, is shown in figure 2.7. In gen- 
 eral, U.S. per capita consumption of both dried prunes 
 and prune juice has decUned over time. U.S. consump- 
 tion of dried prunes declined from over 1.1 pounds 
 per person in 1949 to less than 0.4 pounds per person 
 in 1978, and has remained at about 0.4 pounds per 
 person since then, except for an increase in the late 
 1980s. As shown in table 2.2, U.S. per capita consump- 
 tion of all dried fruits varies from year to year but it 
 did increase through the 1980s and early 1990s, largely 
 
 as a result of increased raisin consumption. Per capita 
 consumption of fruit juices has generally increased 
 over time, with most of the increase occurring in the 
 noncitrus category, and more specifically, in apple 
 juice. Prune juice consumption has decreased over 
 time, and it now accounts for a minuscule portion of 
 total fruit juice consumption (table 2.3 and appendix 
 table A2.1). 
 
 2.3 Marketing Institutions for California Prunes 
 
 The California prune industry has developed a 
 rather unique set of marketing institutions designed 
 to improve producer returns. These institutions in- 
 clude federal and state marketing orders, as well as, 
 marketing and bargaining cooperatives. Following is 
 a short description of each of these institutions. 
 
 Prune Marketing Committee 
 
 California prune producers approved a federal 
 marketing order for dried prunes in 1949 that contin- 
 ues today. This order, which includes provisions for 
 mandatory minimum grade and size standards, mar- 
 ket allocation, reserve pools, and research, is adminis- 
 tered by the Prune Marketing Committee (PMC), with 
 funding provided from a mandatory assessment on 
 all handlers. While it has no control over the acreage 
 planted to prune trees, the PMC has some control over 
 the quantity of prunes marketed, through use of a mini- 
 mum size standard and a surplus set-aside. The mini- 
 mum size standard is used each year, but the set-aside 
 provision has not been used since 1974. The minimum 
 size regulation states that a prune that falls through a 
 
 Figure 2.7. U.S. Per Capita Consumption of California Prunes, 1949-95 
 
 B 
 c 
 
 .S 
 
 
 
 
 
 Dried 
 
 
 
 
 
 Juice and 
 
 
 
 
 
 concentrate 
 
 ^ _ 
 
 \/ 
 
 ^ \ 
 
 \ 
 
 v' 
 
 V A ^ \ 
 
 
 
 
 
 ON 
 
 <^ '7: 
 
 ID 
 
 ON ON 
 
 On 
 
 ON ON 
 
 On 
 
 00 
 On 
 
 Source: Domestic Shipments from California Prune Board, Annual Reports. U.S. population from Bureau of the 
 Census, Statistical Abstract of the United States. 
 
 7 
 
Table 2.2. U.S. Per Capita Consumption of Dried Fruit (lbs. per year) 
 
 Year 
 
 Prunes 
 
 Raisins 
 
 Other 
 Dried Fruits 
 
 All 
 Dried Fruits 
 
 1971 
 
 0.58 
 
 1.43 
 
 0.71 
 
 2.72 
 
 1972 
 
 0.49 
 
 1.04 
 
 1.07 
 
 2.60 
 
 1973 
 
 0.55 
 
 1.38 
 
 0.13 
 
 2.06 
 
 1974 
 
 0.51 
 
 1.29 
 
 0.85 
 
 2.65 
 
 1975 
 
 0.60 
 
 1.29 
 
 0.49 
 
 2.38 
 
 1976 
 
 0.53 
 
 1.28 
 
 0.79 
 
 2.60 
 
 1977 
 
 0.49 
 
 1.25 
 
 0.79 
 
 2.53 
 
 1978 
 
 0.43 
 
 1.10 
 
 0.94 
 
 2.47 
 
 1979 
 
 0.38 
 
 1.31 
 
 0.53 
 
 2.22 
 
 1980 
 
 0.43 
 
 1.46 
 
 0.45 
 
 2.34 
 
 1981 
 
 0.46 
 
 1.54 
 
 0.31 
 
 2.31 
 
 1982 
 
 0.42 
 
 1.52 
 
 0.56 
 
 2.50 
 
 1983 
 
 0.46 
 
 1.58 
 
 0.52 
 
 2.56 
 
 1984 
 
 0.39 
 
 1.90 
 
 0.43 
 
 2.72 
 
 1985 
 
 0.47 
 
 1.92 
 
 0.65 
 
 3.04 
 
 1986 
 
 0.44 
 
 1.83 
 
 0.69 
 
 2.96 
 
 1987 
 
 0.62 
 
 1.88 
 
 0.26 
 
 2.76 
 
 1988 
 
 0.58 
 
 2.07 
 
 0.43 
 
 3.08 
 
 1989 
 
 0.63 
 
 1.92 
 
 U./4 
 
 1Q 
 j.Z.y 
 
 1990 
 
 0.97 
 
 1.80 
 
 0.43 
 
 3.20 
 
 1991 
 
 0.73 
 
 1.78 
 
 0.88 
 
 3.39 
 
 1992 
 
 0.58 
 
 1.62 
 
 0.89 
 
 3.09 
 
 1993 
 
 0.68 
 
 1.86 
 
 0.26 
 
 2.80 
 
 1994 
 
 0.71 
 
 1.72 
 
 0.82 
 
 3.25 
 
 Source: Adapted from USDA Economic Research Service. Food Consumption, Prices and Expenditures, 1996. 
 
 23/32 inch screen cannot be sold for human consump- 
 tion. The effect of the "23" screen varies with crop 
 size. In years with small harvests, the prunes are large, 
 and relatively few prunes fall through the "23" screen, 
 perhaps less than 3 percent. In years with large crops, 
 the prunes are smaller, and the "23" screen may re- 
 move over 6 percent of the crop (Lindauer 1993). The 
 "23" screen provides an incentive for growers to thin 
 their prune orchards in years of abundant fruit set, and 
 therefore reduces variation in yields. 
 
 The 100 count rule, which refers to the number of 
 prunes in one pound of dried prunes, ensures that all 
 prunes marketed as whole prunes are of a minimum 
 size; smaller prunes are diverted into processed prune 
 products. Under the 100 count rule, prune samples 
 that weigh more than 1 pound per 100 prunes can be 
 sold as whole dried prunes while those that weigh less 
 than 1 pound per 100 prunes are processed into a prune 
 
 product, such as prune juice, concentrate, or puree. 
 
 The California Prune Board 
 
 A state marketing order for prunes with provisions 
 for promotion and research was approved by Califor- 
 nia prune producers in 1952. The objective of the Cali- 
 fornia Prune Board (CPB), which is the administrative 
 committee for the marketing order, is to increase 
 worldwide demand for California prunes. The CPB 
 administers domestic and international generic 
 (nonbranded) programs that encompass advertising, 
 sales promotion, public relations, and educational ac- 
 tivities. The CPB also funds production and process- 
 ing research. 
 
 Promotion Strategies and Policies: About 50 per- 
 cent of the CPB's annual domestic budget is spent on 
 television advertising (CPB, Annual Reports). Adver- 
 tising can be made more effective by targeting certain 
 
 8 
 
Table 2.3. U.S. Per Capita Consumption of Fruit Juices (gallons per year) 
 
 Year 
 
 Prune 
 
 Total Noncitrus 
 
 Total Citrus 
 
 Total Juice 
 
 1971 
 
 0.12 
 
 1.13 
 
 4.59 
 
 5.71 
 
 1972 
 
 0.11 
 
 1.25 
 
 4.96 
 
 6.21 
 
 1973 
 
 0.07 
 
 0.96 
 
 5.07 
 
 6.03 
 
 1974 
 
 0.10 
 
 0.93 
 
 5.10 
 
 6.03 
 
 1975 
 
 0.08 
 
 1.00 
 
 5.60 
 
 6.61 
 
 1976 
 
 0.09 
 
 1.10 
 
 5.84 
 
 6.93 
 
 1977 
 
 0.11 
 
 1.06 
 
 5.94 
 
 6.99 
 
 1978 
 
 0.09 
 
 1.15 
 
 5.29 
 
 6.44 
 
 1979 
 
 0.10 
 
 1.44 
 
 5.32 
 
 6.77 
 
 1980 
 
 0.09 
 
 1.49 
 
 5.66 
 
 7.15 
 
 1981 
 
 0.09 
 
 1.73 
 
 5.69 
 
 7.42 
 
 1982 
 
 0.10 
 
 1.58 
 
 5.18 
 
 6.75 
 
 1983 
 
 0.08 
 
 1.82 
 
 6.56 
 
 8.38 
 
 1984 
 
 0.06 
 
 1.99 
 
 5.28 
 
 7.27 
 
 1985 
 
 0.07 
 
 2.16 
 
 5.57 
 
 7.72 
 
 1986 
 
 0.07 
 
 2.17 
 
 5.77 
 
 7.94 
 
 1987 
 
 0.07 
 
 2.19 
 
 5.98 
 
 8.17 
 
 1988 
 
 0.06 
 
 2.40 
 
 5.80 
 
 8.21 
 
 1989 
 
 0.07 
 
 2.35 
 
 5.34 
 
 7.69 
 
 1990 
 
 0.04 
 
 2.23 
 
 4.63 
 
 6.86 
 
 1991 
 
 0.04 
 
 2.53 
 
 5.36 
 
 7.89 
 
 1992 
 
 0.03 
 
 2.40 
 
 4.87 
 
 7.27 
 
 1993 
 
 0.04 
 
 2.45 
 
 5.91 
 
 8.37 
 
 1994 
 
 0.04 
 
 2.59 
 
 6.00 
 
 8.60 
 
 Source: USDA/ Economic Research Service. 
 
 groups, and CPB advertising has recently targeted 
 women aged 45 or older who are light or non-recent 
 users of prunes. 
 
 Sales promotional activities feature supermarket 
 display contests in the United States and Canada. The 
 display contests offer retailers travel and merchandise 
 prizes for building winning displays. Generic point- 
 of-sale display cards are provided, which tie in with 
 the television advertising theme for prunes or with 
 nutritional attributes of prune juice. The CPB also ties 
 in with the National Cancer Institute's "5-a-Day" pro- 
 gram, which recommends eating five servings of fruits 
 and vegetables every day (Peterson 1994). 
 
 California public relations activities support the 
 advertising themes and promote the use of pureed 
 prunes as a substitute for butter, oil, or margarine in 
 baking. For instance, a direct mail campaign to food- 
 service operators offered low-fat brownies made with 
 
 prunes, and brochures with low-fat recipes. The CPB 
 also exhibits at trade shows and advertises in trade 
 publications. Other public relations activities include 
 recipe releases to newspaper food editors and super- 
 market consumer affairs directors, and visits to maga- 
 zine food editors. The CPB also distributes one-ounce 
 packages of pitted prunes at fitness events and in ce- 
 reals. 
 
 The CPB also conducts international promotional 
 activities, such as in-store demonstrations, publicity, 
 trade education, and advertising. Support for inter- 
 national promotion increased sharply in 1987, and has 
 increased in real terms since then. Approximately 50 
 percent of the funding for international promotional 
 activities is provided by the USDA Market Promotion 
 Program (MPP). The rise in export sales coincided with 
 the participation of the California Prune Board with 
 the MPP and its predecessors. In 1992, the CPB pro- 
 
 9 
 
Figure 2.8. California Prune Board Promotional Expenditures, 1949-95 
 
 !,000 
 
 7,000 
 
 (a) Current Dollars 
 
 6,000 
 
 -Domestic Public Relations 
 
 Domestic Advertising 
 
 5,000 
 
 o 
 Q 
 
 -a 
 c 
 
 -Export Promotion 
 - Industry Research 
 
 (b) Constant (1995) Dollars 
 
 8,000 
 
 7,000 
 
 S 6,000 
 o 
 D 
 in 
 
 R 5,000 
 
 4,000 
 3,000 
 
 -Domestic Public Relations 
 
 Domestic Advertising 
 
 -A Export Promotion 
 
 -X Industry Research 
 
 Source: California Prune Board, Annual Reports. Consumer price index from the Bureau of the Census. 
 
 10 
 
moted the consumption of California prunes in 13 na- 
 tions, but funding reductions in the MPP led the CPB 
 to eliminate promotional programs in several of them. 
 In 1994, the CPB conducted promotional campaigns 
 in Germany, Italy, the United Kingdom, Japan, and 
 Mexico (Peterson 1994). 
 
 Statistical Overview of Promotional Expenditures: 
 CPB expenditures by category are shown in nominal 
 dollars in figure 2.8a and in real 1995 dollars in figure 
 2.8b. CPB funding, in real dollars, for production re- 
 search has remained roughly constant. In the 1990s, 
 about 4 percent of the CPB budget was invested in 
 production research. All CPB generic advertising and 
 promotion support for California prunes was discon- 
 tinued during 1975-78, when Sunsweet's management 
 voted to eliminate the assessment for generic promo- 
 tion, believing they could better promote Sunsweet's 
 products on their own. The CPB reinstated a public 
 relations program in 1979 and resumed generic adver- 
 tising in 1980. Since the early 1980s, expenditures by 
 the CPB in real dollars declined for advertising and 
 increased for other domestic promotional activities. 
 CPB expenditures on total promotional activities are 
 shown in real and nominal dollars in figure 2.9. In 
 both real and nominal dollars, total CPB expenditures 
 on promotion have increased since the early 1980s. 
 
 While the Prune Marketing Committee and the 
 California Prune Board are separate administrative 
 bodies, established under different enabling legisla- 
 tion, they cooperate by using common office facilities 
 and staff, and they also share industry data. The Dried 
 Fruit Association of California inspects all prunes pro- 
 duced in California. This inspection service certifies 
 the salable weight that the packer uses to pay the pro- 
 ducer, determines the size, off-grade, and undersize 
 of each lot, and collects data used by the industry to 
 establish crop tonnage, inventory composition, and as- 
 sessment fees. 
 
 Marketing Firms 
 
 The processing and marketing of California's an- 
 nual prune crop is almost evenly divided between 
 Sunsweet Growers, the industry's only marketing co- 
 operative, and a group of independent growers and 
 packers. Sunsweet members deliver their entire pro- 
 duction to the cooperative, and their returns are based 
 on the selling price of the processed fruit. Indepen- 
 dent growers sell to independent packers or handlers 
 on a contractual basis. All California prune produc- 
 ers and handlers pay mandatory assessments to sup- 
 port the generic promotional efforts of the California 
 Prune Board. Members of Sunsweet also support sub- 
 stantial brand advertising of Sunsweet prune prod- 
 ucts. Under the prune marketing order, there is no 
 assessment offset: if a packer such as Sunsweet pro- 
 motes its own brand, there is no reimbursement of any 
 
 of the assessment that the packer pays toward the ge- 
 neric program. 
 
 The Prune Bargaining Association 
 
 The Prune Bargaining Association (PBA) is a vol- 
 untary organization that represents about 40 percent 
 of the independent growers (Giacolini 1993). Each 
 year, the PBA negotiates with the independent pro- 
 cessors to establish a selling price for its members. The 
 price negotiated by the PBA influences the price re- 
 ceived by all independent growers. 
 
 2.4 Trends in Factors Associated With Prune 
 Demand 
 
 Several factors associated with prune consumption 
 and the demand for prunes have been changing over 
 time. These factors will be discussed briefly here, and 
 their quantitative impacts on the demand for prunes 
 will be examined in detail in the following sections of 
 this report. 
 
 Age of the Population 
 
 The U.S. population is aging: the share of U.S. citi- 
 zens 65 years old or older increased from 8.0 percent 
 in 1949 to 13.0 percent in 1995 (Bureau of the Census 
 1996). This may or may not bode well for the prune 
 industry. Currently available data indicate that older 
 people consume more prunes per capita than younger 
 people do (table 2.4). 
 
 Table 2.4. U.S. Per Capita Consumption of Dried 
 Prunes in 1986, by Age Group 
 
 Age group 
 (years) 
 
 Per capita consumption 
 (pounds) 
 
 18-24 
 
 0.67 
 
 25-44 
 
 0.62 
 
 45-54 
 
 1.00 
 
 55-64 
 
 1.52 
 
 >65 
 
 1.95 
 
 Source: Dale et al. 1988 
 
 This pattern of consumption may be attributable to 
 either an age effect or to a cohort effect. If it is an age 
 effect, a person's preference for prunes increases as the 
 person ages, and as today's young people become 
 older, their prune consumption will increase. If it is a 
 cohort effect, older people today consume more prunes 
 perhaps because as children they ate more dried fruit 
 relative to fresh fruit (since seasonal fresh fruit avail- 
 ability was limited, while dried fruit was available 
 year-round). Therefore, having established patterns 
 
 11 
 
Figure 2.9. California Prune Board Expenditures on Domestic Promotion, 1949-95 
 
 Source: Expenditxires from California Prune Board, Annual Reports. Consumer Price Index from Bureau of the 
 Census. 
 
 of dried fruit consumption as children, older Ameri- 
 cans still consume relatively more dried fruit. If the 
 cohort effect explains much of the higher consump- 
 tion of prunes by older people, then, holding other 
 influences constant, per capita prune consumption can 
 be expected to decline over time, since those who are 
 young today are not developing similar preferences 
 for dried fruit. 
 
 Changes in Household Structure 
 
 As recently as the 1950s, the predominant type of 
 household included two adults and two children, with 
 only the male head of the household working outside 
 the home.' In the 1990s, single-person households are 
 much more important, female heads of households are 
 common, and nonworking spouses are now the ex- 
 ception more than the norm. These factors, combined 
 with rising incomes and changes in technology avail- 
 able to households (such as microwaves and home 
 freezers) and changes in food products available (in- 
 cluding pre-prepared foods for home serving, and fast- 
 food restaurants), have contributed to major changes 
 
 in the way people live and, in particular, eat. 
 
 Importantly, the growth in per capita incomes can 
 be expected to have led to an increase in the demand 
 for food quality and services associated with food. The 
 increased labor-force participation of women can be 
 expected to have led to an increased demand for con- 
 venience in food, and for food with low preparation 
 time (since the opportunity cost of working women's 
 time is higher). These two factors can account for much 
 of the major changes in food purchase patterns: a 
 higher proportion of meals away from home, and a 
 higher proportion of pre-prepared meals. Since prunes 
 are ready-to-eat without additional preparation, the 
 increased demand for convenience in food could in- 
 crease the demand for prunes. 
 
 Fruit in the Diet 
 
 The average diet in the United States has been 
 slowly changing to include leaner meats and more 
 fruits and vegetables, as recommended by public 
 health organizations. Per capita consumption of fresh 
 fruit exhibited a steady declining trend from 1939 un- 
 til the mid-1960s, when it began a gradual rise. Much 
 
 1 . Alston et al. (1997) documented some of the changes in household structure and consumption patterns in the United 
 States. This section draws heavily on that discussion. See also Blaylock and Smallwood (1986). 
 
 12 
 
of the decline and subsequent turnaround was attrib- 
 utable to the consumption of citrus fruits, particularly 
 oranges, although noncitrus fruits, such as peaches and 
 grapes, exhibited similar trends. The overall increase 
 in per capita consumption of fresh fruits and veg- 
 etables continued slowly during the 1970s, and at a 
 faster rate during the 1980s. Total per capita consump- 
 tion of commercially produced fruits and vegetables 
 was estimated at 678 pounds in 1994 (farm- weight 
 basis), an increase of 20 percent over the quantity in 
 1970. 
 
 Overall, U.S. per capita consumption of processed 
 fruit increased by 18.6 percent and per capita consump- 
 tion of fresh fruit increased by 25.2 percent during the 
 period from 1970 through 1994. There were, however, 
 significant deviations by product category. Per capita 
 consumption of fresh citrus, for example, decreased 
 slightly, while per capita consumption of fresh 
 noncitrus fruits increased by 41 percent (appendix 
 table A2.3). Since fresh noncitrus fruits are believed 
 to substitute for dried fruit (including prunes), the 
 higher sales of fresh noncitrus fruit since 1970 may be 
 associated with reduced purchases of dried fruit. On 
 the other hand, increased demand for all fruits is ex- 
 pected to increase the demand for dried fruit. 
 
 U.S. per capita consumption of dried fruit declined 
 by one-half from 1920 to 1980 (USDA 1979). In the 
 1920s, fruit was consumed fresh mainly during the 
 harvest period, since fresh fruit was typically not avail- 
 able, and when available, it was prohibitively expen- 
 sive during the off-season. Technical advances in va- 
 rieties, production, storage, shipping, and the devel- 
 opment of new areas of production now make fresh 
 fruit available throughout the year at reasonable prices. 
 The increased supply of fresh fruit on a year-round 
 basis may have reduced the demand for dried fruit. 
 Despite the increased consumption of fresh fruits, how- 
 ever, U.S. per capita consumption of all dried fruit has 
 also increased since the late 1970s (appendix table 
 A2.4). 
 
 Baking Use of Prunes 
 
 Research on new uses for prunes has found that 
 prunes can be used as a fat substitute in baked goods. 
 The use of prune puree, for example, allows consum- 
 ers to reduce the fat in baked goods by 60 to 90 per- 
 cent. Other benefits include reduced cholesterol, im- 
 proved preservation, and improved nutrition. Prune 
 puree, prune paste, and diced prunes are now being 
 marketed as cooking ingredients in baked goods, and 
 are promoted as fat substitutes. The share of food tech- 
 nologists who use prunes in baking increased from 10 
 percent in 1992 to about 27 percent in 1996 (CPB Sep- 
 tember 1996). Given the current emphasis on reduc- 
 ing fat consumption in human diets, this could pro- 
 vide an important boost to the demand for prunes. 
 
 Prune puree, however, still constitutes less than 2 per- 
 cent of California domestic prune shipments (appen- 
 dix table A2.5). 
 
 2.5 Concluding Comments on Supply and De- 
 mand Trends 
 
 The California prune industry has experienced sig- 
 nificant changes in both supply and demand over the 
 past 50 years. The bearing acreage of prunes has 
 trended down, but this was largely offset by increas- 
 ing average yields. As a result, total California prune 
 production has varied significantly from year to year, 
 but without either an upward or a downward trend. 
 On the demand side, there have been sigruficant trends 
 in crop utilization, product form, and per capita con- 
 sumption. The amount and share of the annual crop 
 shipped to domestic markets decreased over time, 
 while exports increased to over 40 percent of the crop. 
 During the same time frame, the proportion of dried 
 prunes sold in the pitted form increased from less than 
 1 percent to over 80 percent, as a result of new and 
 improved technology for pitting. Accompanying these 
 changes were significant reductions in U.S. per capita 
 consumption of dried prunes and prune juice. 
 
 A number of factors may have affected the demand 
 for prunes over time. Those mentioned include the 
 advertising and promotion programs conducted by 
 the CPB, as well as the traditional demand shifters of 
 income, population, and prices of competing products. 
 Other potentially important factors include (1) chang- 
 ing demographics, especially the increasing average 
 age of the U.S. population and the changing house- 
 hold structure, (2) health and diet concerns that have 
 increased the demand for fruits and vegetables, (3) 
 the year-round availability of fresh fruit, and (4) new 
 and improved products and uses (pitted prunes and 
 prunes as a fat substitute in baking). However, since 
 many of the factors have been changing at the same 
 time, it is difficult to isolate cause and effect. The fol- 
 lowing sections detail the specification and estimation 
 of aggregate demand relationships for prunes. 
 
 13 
 
3. ACCOUNTING FOR CHANGES IN AGGREGATE U.S. PRUNE 
 
 CONSUMPTION 
 
 In this section, we specify demand models for Cali- 
 fornia prunes and report the results from estimating 
 these models using monthly data covering the period 
 from September 1992 through July 1996, and annual 
 data for 1949 through 1995. These econometric mod- 
 els use per capita quantities of dried prunes in the 
 United States. Important related distinctions concern 
 the market level (e.g., farm versus retail) and whether 
 the quantity is all prunes for all destinations (as farm- 
 level quantities are) or for domestic consumption only 
 (as retail quantities are). Section 3.1 covers general 
 theoretical aspects, section 3.2 covers the monthly 
 models, and section 3.3 covers armual models 
 
 3.1 Aggregate Domestic Demand Models, Theo- 
 retical Considerations 
 
 Aggregate per capita demand models provide pa- 
 rameters that can be used to estimate gross and net 
 benefits to the industry from promotion. Once the 
 model is estimated econometrically, we use the esti- 
 mates from to analyze the effect that a change in pro- 
 motion would be likely to have on per capita consump- 
 tion. Economic theory is used as a guide in the speci- 
 fication of the model, in the identification of variables 
 that are used to explain changes in consumption pat- 
 terns, and in the interpretation of the results from esti- 
 mation. 
 
 A Consumer Demand Model 
 
 Suppose we use Q, to represent the per capita quan- 
 tity of prunes (of uniform quality) demanded by a rep- 
 resentative consumer during a particular year, t. The 
 theory of consumer demand suggests a model in which 
 the quantity demanded, Q,, depends on the corre- 
 sponding price of prunes, PP,, the prices of all other 
 goods that are substitutes or complements for prunes 
 (such as other dried fruits or laxatives, and fresh fruits, 
 in particular), PS^, and total money income or expen- 
 diture on all goods, EXP^. This model can be expressed 
 as 
 
 Q,^f{PP^,PS,,EXP). (3.1) 
 
 To make this model operational, one must specify 
 a particular functional form for/(.) — for instance, a lin- 
 ear functional form, which we use later. In this model, 
 we would expect the own-price effect to be negative 
 
 (a negative coefficient on PP). The cross-price effects 
 (the coefficients on other prices, PS) can be positive 
 or negative, but are expected to be predominantly 
 positive, especially for close substitutes, and the in- 
 come effect (the coefficient on EXP,) is probably posi- 
 tive and in the range for a normal good, correspond- 
 ing to an income elasticity of demand for prunes be- 
 tween 0 and 1 . In other words, an increase in the price 
 of prunes would lead to a decrease in prune consump- 
 tion, while an increase in the price of a substitute or in 
 total money income would lead to an increase in prune 
 consumption. 
 
 In addition, the theory of consumer demand im- 
 plies that the demand equation should be homoge- 
 neous of degree zero in money income and prices — 
 doubling money income and all prices should leave 
 consumption unaffected, since nothing real has 
 changed. This homogeneity condition is commonly 
 imposed by dividing all of the prices and income by a 
 general price index, such as the consumer price index 
 (CPI), thereby expressing all of the monetary variables 
 in the demand equation in real terms (denoted i?PP,, 
 RPS,, and REXP). The resulting model is 
 
 Q, = / (RPP,, RPS,, REXP) . (3.2) 
 
 Both of these demand equations (equations 3.1 and 
 3.2) implicitly assume constant tastes and preferences 
 for prunes. In order to accommodate changes in pref- 
 erences arising from promotion or anything else that 
 may affect demand (such as demographic character- 
 istics of the consumer), the model can be augmented 
 with other demand shift variables? Clearly, promotion 
 is one such variable. To obtain reliable estimates of 
 the influence of the factors that are of most importance 
 for the present study — in particular, the responsive- 
 ness of demand to price and promotion — it is neces- 
 sary to take into account the influence of other demand 
 shift variables as well. Otherwise, there is a risk that 
 the effects of omitted shifters will be attributed falsely 
 to the variables included in the model. 
 
 In a model of consumer demand for California 
 prunes, appropriate shift variables can be included to 
 represent the effects of such things as (a) increased 
 consumer health consciousness and a rising consumer 
 interest in natural foods; (b) other demographic 
 changes, such as changes in the age structure of the 
 population (likely to be especially important for 
 
 2. Blaylock and Smallwood (1986) document some of the general trends in consumer demand for food that may be 
 reflected in shift variables of these types. 
 
 14 
 
prunes, since they are relatively heavily consumed by 
 older people), a higher rate of labor-force participa- 
 tion by women, changes in the ethnic composition of 
 the population, and the fact that more meals are eaten 
 away from home; (c) generic promotion by the Cali- 
 fornia Prune Board, brand promotion by Sunsweet and 
 others, and other changes in merchandising expendi- 
 tures, and (d) changes in the quality of California 
 prunes. 
 
 To deal with all of these individual variables ex- 
 plicitly in a model is impossible, given our limited 
 dataset and the difficulty of identifying their individual 
 effects, when many variables change smoothly, to- 
 gether, over time. Instead, we focus on those shift vari- 
 ables for which we think the effects are likely to be the 
 most important. Thus, we include four types of shift 
 variables including (a) where possible, a variable to 
 represent the changing age structure of the popula- 
 tion; (b) the quantity of promotion, represented by the 
 total — not per capita — promotional expenditures' of 
 both the California Prune Board and Sunsweet, ex- 
 pressed in real terms by dividing expenditures by the 
 CPI, RPROCPB, and RPROSUN,, respectively; (c) a lin- 
 ear time trend variable, TIME,, included to represent 
 the effects of other trends, as described above, that are 
 not being modeled explicitly; and (d) quarterly inter- 
 cept dummy variables (SPR,, WIN,, and SUM,), to re- 
 flect seasonal shifts in demand in the monthly demand 
 models. Note that Fall is the base season — the esti- 
 mated coefficients for SPR,, WIN,, and SUM, show the 
 change in demand from the Fall base season. 
 
 Incorporating the shift variables leads to an aug- 
 mented model of demand, as follows 
 
 Q,=f(RPG,,RPS,, REXP,, RPROCPB,, RPROSUN,, 
 
 SUM,, WIN,, SPR,, TIME,) (3.3) 
 
 The effects of the demand shift variables are not as 
 easy to predict as those of the more conventional ones. 
 Promotion is expected to have an unambiguously posi- 
 tive effect on demand, but, even then, only if it has 
 been successful in increasing demand for prunes; oth- 
 erwise, this variable would have no effect on demand. 
 The effect of the time trend is likely to be negative, 
 reflecting a general shift of consumer demand away 
 from prunes over time. It is expected that per capita 
 consumption of prunes increases with increases in the 
 fraction of the population in older age categories, but 
 it is likely to be difficult to separate age effects from 
 cohort effects (indeed, as noted above, an important 
 question is whether declining per capita consumption 
 
 is an age effect, reversible as the population becomes 
 older again, or a cohort effect, and not reversible). 
 
 Horizontal and Vertical Market Linkages and their 
 Implications 
 
 The above models refer to final consumer demand. 
 Often, however, in empirical work, we use data that 
 relates to the derived demand at the farm or wholesale 
 level, which is derived from the final consumer de- 
 mand and the economic and technological character- 
 istics of the intermediate functions between the final 
 consumer and the market level being studied (e.g., the 
 marketing, processing, and transportation functions). 
 Therefore, in addition to variables indicated by the 
 theory of consumer demand, derived demand func- 
 tions may also include variables representing process- 
 ing costs, labor costs, and so on. In the models pre- 
 sented here, the consumer price index plays a dual role 
 as a general index of the prices of substitutes (other 
 than those identified in the model below) in consump- 
 tion for prunes, and as an index of the prices of mar- 
 keting inputs. Derived demand equations are gener- 
 ally expected to be less price-responsive (less elastic) 
 than retail demand equations. 
 
 Another aspect of market structure we must con- 
 sider is one of horizontal linkages. These include link- 
 ages between the U.S. market and international mar- 
 kets for U.S. prunes, and linkages among markets for 
 alternative end-uses for U.S. prunes. When we study 
 the market for dried prunes, we are considering only 
 a subset of the total market for U.S. prunes; when we 
 study the domestic market we are studying only a sub- 
 set of the total world market for dried prunes. This 
 can be thought of in terms of residual supply and de- 
 mand. In practice, we can study the domestic demand 
 for dried U.S. prunes independently of the markets 
 for other uses of U.S. prunes without experiencing any 
 econometric problems. However, when we want to 
 simulate changes in domestic demand for U.S. dried 
 prunes, we must recognize that the markets are linked, 
 and a significant element of the market response to an 
 advertising-induced increase in domestic consumption 
 of dried prunes will be a reallocation of prunes from 
 other uses; either other domestic uses or exports. These 
 horizontal linkages are captured through a modifica- 
 tion of the effective supply of dried prunes to the U.S. 
 market, which is the residual from total supply and 
 demand for other uses. The issues of residual supply 
 and demand and derived demand are discussed at 
 greater length by Alston et al. (1995). 
 
 3. The choice of whether to include promotion in per capita terms or in total was discussed by Alston et al. (1997). 
 
 15 
 
Figure 3.1. U.S. Per Capita, Monthly Prune Consumption (Retail), September 1992 to July 1996 
 
 0.025 
 
 3 
 O 
 
 0.005 
 
 Q 
 
 CO cn 
 
 CO cn 
 
 CD 
 
 - ofc 2 ofc ^ 
 
 9 9 
 
 < -T- < O 
 ^ JI^ ^ ' ' 
 
 Source: Infoscan, IRI. 
 
 3.2 Aggregate Monthly Domestic Demand Mod- 
 els* 
 
 This section contains a summary of the regression 
 models used to estimate the effects of promotion on 
 monthly per capita prune consumption in the United 
 States. It should be noted that the models in section 
 3.2 concern retail sales only. The models studied here 
 indicate that promotional expenditures have a statis- 
 tically significant, positive effect on monthly prune 
 consumption at the retail level. Before presenting the 
 demand models and the empirical results, we give an 
 overview of the data and data sources for these mod- 
 els. 
 
 Data 
 
 The time period for this part of the study was from 
 September 1992 through July 1996, comprising 51 ob- 
 servations with lengths of four weeks each. The 
 monthly regressions presented here are based on data 
 collected from four sources, as described below. 
 
 Quantities : The data describing the quantities of 
 dried prunes consumed in the United States come from 
 Infoscan IRI retail market profiles representing U.S. 
 consumption based on 64 cities/ regions in the United 
 States. These data were provided by Sunsweet. We 
 used "total" prune consumption over this period, 
 which was the total of all brands sold in retail stores. 
 
 During this period, retail sales accounted for approxi- 
 mately 30 percent of all prune sales (CPB 1997). The 
 quantities were converted into pounds per capita us- 
 ing an estimate of the U.S. midyear population (Inter- 
 national Monetary Fund). We converted this popula- 
 tion statistic into population per observation period 
 using a growth formula provided by the U.S. Census 
 Bureau. Monthly per-capita consumption of prunes, 
 shown for the observation period in figure 3.1, aver- 
 aged 0.015 pounds with a standard deviation of 0.002. 
 
 Prices: The prices used in the demand models were 
 the average retail price ($/ pound) of prunes purchased 
 and the average price of dried fruit other than prunes 
 ($/pound) purchased from retail food stores.^ These 
 prices also came from the Infoscan IRI data and were 
 converted into 1996 dollars using the consumer price 
 index (International Monetary Fund). Deflation by the 
 CPI has the effect of treating an aggregate of all other 
 goods as a general substitute for prunes. It would be 
 desirable to represent the effects of close substitutes 
 explicitly, by including a separate variable for each one, 
 but data constraints and likely statistical problems 
 (such as multicollinearity) mean that we cannot in- 
 clude too many other prices; probably no more than 
 one. While it is difficult to identify a particular index 
 to represent those goods that are close substitutes for 
 prunes, the models presented in this section use the 
 real price of an aggregate of other dried fruit as RPS,. 
 
 4. Throughout this section, we use the term "monthly" to refer to the length of the observation period, where in reality, 
 the observations are four weeks long. 
 
 5. Prices of dried fruit excluding raisins, and prices of raisins alone were also considered, but neither is included in any 
 of the models in this section. 
 
 16 
 
It must be stressed that there was very little price 
 movement for prunes and other dried fruit during the 
 observation period used for the monthly demand mod- 
 els. Deflating the prices by the CPI dampened this 
 movement even further. Over the observation period, 
 the average real price of prunes was $2.18 per pound 
 per month (it ranged from $2.05 to $2.26 per month), 
 and the average real price of other dried fruit was $2.39 
 per pound per month. The standard deviations were 
 0.045 and 0.042 respectively. Thus, we are attempting 
 to correlate a consumption variable with price vari- 
 ables that simply did not move very much. A data set 
 covering a longer time period, with more variation in 
 prices, might be expected to provide improved esti- 
 mates of price effects on demand. 
 
 Promotion : Promotion includes generic promo- 
 tional expenditures by the California Prune Board and 
 brand promotion by Sunsweet Growers, the largest 
 private promoter of dried prunes. There is little or no 
 branded promotion (besides display units and other 
 trade promotions like "buy one get one free") by other 
 packers, and Sunsweet is the only packer to advertise 
 its prunes nationally. The aggregate monthly promo- 
 tion variables (PROCPB and PROSUN) each include 
 monthly sales promotion and monthly television ad- 
 vertising. 
 
 PROMOCPB is monthly sales promotion expendi- 
 ture by the California Prune Board for prunes. Spe- 
 cifically PROMOCPB is the sum of the following bud- 
 getary items: Coupon Program, Public Relations, Sam- 
 pling, and Merchandising /Sales Promotion (Source: 
 California Prune Board Generic Monthly Program Evalu- 
 ation for crop years 92/93-95/96). Throughout this 
 section, "Sales Promotion" refers to the sum of these 
 budgetary items. All figures were converted from 
 monthly into monthly data using daily averages for 
 each month, in order to match the Infoscan IRI quan- 
 tity and price series. PROMOSUN is sales promotion 
 expenditures by Sunsweet Growers for prunes. 
 PROMOSUN consists of monthly total dried fruit pro- 
 motion converted from monthly data into monthly 
 data using average daily expenditures for each month 
 (Source: Sunsweet Growers Monthly Advertising and 
 Merchandising Expenses for crop years 92/93-95/96). 
 
 ADCPB is television advertising expenditure for the 
 specific observation point, as billed to the CPB. Un- 
 like the sales promotion expenditures, which came 
 from the crop-year budgets, these data accurately re- 
 flect the timing of television advertisements. The fig- 
 ures were converted from monthly expenditures to 
 monthly expenditures using daily averages (Source: 
 CPB). ADS LIN is television advertising expenditures 
 as billed to Sunsweet. These data, which were reported 
 on a weekly basis, were summed to provide monthly 
 figures for the observation period (Source: Sunsweet 
 Growers). 
 
 Finally, aggregate promotion variables were con- 
 structed for CPB and Sunsweet which include both 
 sales promotion and television advertising: PROCPB 
 = ADCPB + PROMOCPB and PROSUN = ADSUN + 
 PROMOSUN, expressed in millions of dollars. 
 
 The timing of sales promotion expenditures may 
 involve problems. As discussed above, only the two 
 television advertising variables ADCPB and ADSUN 
 reflect accurately when advertising was seen by view- 
 ers. The two sales promotion variables are off by an 
 indeterminate amount, since they came directly from 
 the CPB's and Sunsweet's accounting records. The 
 problem with these two variables is that they aggre- 
 gate a variety of important sales promotion expendi- 
 tures (coupon payments, public relations, sampling, 
 and merchandising/sales promotion) whose timing 
 varies. Thus, while the AD- variables reflect dollar 
 figures corresponding to the actual timing of promo- 
 tion (e.g., television commercials), the PROMO- vari- 
 ables do not. Since some promotional expenses are 
 paid when the promotion occurs, while others are paid 
 ahead of time, and still others are paid after the pro- 
 motion is over, it is not clear whether these variables 
 should be lagged or not in the regressions. Ultimately, 
 it was decided to use the variables as presented in the 
 budgets. 
 
 Promotional expenditures were expressed in real 
 (or quantity) terms to reflect the view that, if the CPB's 
 and Sunsweet's budgets doubled, and the cost of pro- 
 motion doubled as well, there would be no real change 
 in the amount of promotion undertaken. Ideally, to 
 do this, we would deflate promotional expenditures 
 by an index of the unit cost of promotion. Lacking 
 such an index, we assume that promotion costs rise 
 with prices generally, and use the CPI to deflate pro- 
 motional expenditures. 
 
 Other Variables : A total expenditure variable was 
 also included. As monthly data were not available, 
 the expenditure variable was based on quarterly data 
 on total private domestic consumption expenditures 
 (billions of dollars) for the United States, taken from 
 the International Financial Statistics published by the 
 International Monetary Fund. These figures were then 
 deflated by a quarterly CPI (August 1996 = 1.00), to 
 put them in real terms, and divided by a quarterly 
 estimate of U.S. population, to put them in per capita 
 terms. The annual average of this per-capita income 
 variable was $18,690.98 and its standard deviation was 
 $494.60. 
 
 A time-trend variable was also used in some of the 
 demand models. Over the period of study, there has 
 been a slight decrease in per capita prune consump- 
 tion, so we might expect the coefficient on the trend 
 variable to be negative, if this decline in consumption 
 is not attributable to the other variables in the model. 
 Conversations with CPB staff suggested that prune 
 
 17 
 
Table 3.1. Description of Variables in the Monthly Demand Model 
 
 Variable 
 
 Definition 
 
 Units 
 
 Data Source 
 
 TIME. 
 
 RPP. 
 
 "Month" 
 
 "Monthly" per capita 
 consumption of dried prunes in 
 the U.S. from the week ending 
 September 6, 1992 to July 7, 1996. 
 Recall, one "month" here is 
 exactly four weeks long. 
 
 Real average retail price of 
 prunes. 
 
 One observation is 
 four weeks long. 
 
 pounds of dried 
 prunes per person 
 per month 
 
 real dollars (August 
 1996=1) per pound 
 of processed prunes 
 
 Prune Consumption 
 from Infoscan IRI 
 retail market profiles. 
 U.S. population from 
 International 
 Financial Statistics 
 
 Prices came from 
 Infoscan IRI and were 
 deflated by CPl data 
 from International 
 Financial Statistics 
 
 RPS, 
 
 Real average retail price of all 
 other dried fruit. 
 
 real dollars per 
 pound 
 
 Prices came from 
 Infoscan IRI and were 
 deflated by CPl data 
 from International 
 Financial Statistics 
 
 i?£XP, Real, average, quarterly per-capita 
 
 private domestic consumption 
 expenditures. 
 
 RPROCPB^ Aggregate real promotional 
 
 expenditures by the California 
 Prune Board for dried prunes in 
 the U.S. market. 
 
 real dollars per 
 person per quarter 
 
 real doUars per 
 month 
 
 International 
 Monetary Fund 
 
 California Prune 
 Board 
 
 RPROSUN^ Aggregate real promotional 
 expenditures by Sunsweet for 
 dried prunes in the U.S. market. 
 
 real dollars per 
 month 
 
 Sunsweet Growers 
 
 WIN, SPR, Winter, Spring and Summer 
 SUM dummy variables respectively. 
 
 Notes: RPS was broken down into its components (raisins, RPR, and other dried fruit, RPF) which were used as added 
 instruments in the 2SLS model estimated in this report. 
 
 18 
 
Table 3.2. Summary Statistics of the Variables in the Monthly Demand Model 
 
 Variable 
 
 N 
 
 Mean 
 
 Standard 
 Deviation 
 
 Minimum 
 
 Maximum 
 
 Q 
 
 51 
 
 0.015 
 
 0.002 
 
 0.012 
 
 0.020 
 
 RPP 
 
 51 
 
 2.179 
 
 0.045 
 
 2.054 
 
 2.264 
 
 PDC 
 
 Dl 
 
 T "20/1 
 
 U.U4Z 
 
 Z.Md 
 
 l.oZ/ 
 
 RPR 
 
 51 
 
 2.010 
 
 0.041 
 
 1.916 
 
 2.067 
 
 RPF 
 
 51 
 
 4.599 
 
 0.363 
 
 3.984 
 
 5.258 
 
 REXP 
 
 51 
 
 18,691 
 
 499.52 
 
 17,785 
 
 19,483 
 
 RPROMO 
 
 51 
 
 1.512 
 
 0.483 
 
 0.516 
 
 2.397 
 
 TIME 
 
 51 
 
 26 
 
 14.866 
 
 1 
 
 51 
 
 Notes: In this table, RPROMO = RPROCPB + RPROSUN represents total monthly promotion in millions of dollars (aggre- 
 gated to maintain Sunsweet's confidentiality). RPR and RPF are the real prices of raisins and other dried fruit, respectively. 
 These were used in the 2SLS model. 
 
 consumption varies seasonally. To investigate this pos- 
 sibility, three quarterly dummy variables (WIN = Win- 
 ter, SPR = Spring, and SUM - Summer) were speci- 
 fied, with the Fall quarter (October, November, Decem- 
 ber) used as the base season. The seasonal variables 
 are equal to one for each month during a given season 
 and zero otherwise. When the four-week period for a 
 particular observation spans two seasons, it is assigned 
 to the season that applies for the majority of days in 
 the 4-week period. 
 
 Table 3.1 summarizes the definitions of the variables 
 used in the monthly demand models. A complete list- 
 ing of the data, as used in the regressions, is provided 
 in appendix table A3.1. Summary statistics for the 
 monthly demand variables are included in table 3.2. 
 
 Estimation Results and Selection of the Preferred Monthly 
 Models 
 
 In this section, we present regression equations that 
 represent the monthly demand for prunes in the United 
 States for the period spanning the months September 
 1992 through July 1996. Other regression equations 
 are also included for comparison, and as indicators of 
 the robustness of certain aspects of the preferred mod- 
 els. Initially, the models were estimated by ordinary 
 
 least squares (OLS). 
 
 Promotion Variables and Lags : To investigate the 
 timing issues with the sales promotion variable, sev- 
 eral models that included lagged promotion variables 
 with a variety of polynomial restrictions on the lag 
 structure were tested. These models were all unsatis- 
 factory, and they are not presented here. Further, we 
 decided to use the aggregate PROCPB and PROSUN 
 variables instead of disaggregating each of them into 
 AD and PRO.'' 
 
 Functional Form Choice : The functional form for 
 the demand equation must be specified, in order to 
 estimate an econometric demand model. The choice 
 of the functional form for the demand equation, which 
 is guided by diagnostic tests, is important because it 
 can influence the results of the econometric estima- 
 tion (e.g., see Chalfant and Alston, 1988; Alston and 
 Chalfant, 1991). In what follows, we focus on the re- 
 sults from a demand equation that is linear in all the 
 variables, except that we include the square root of 
 promotion instead of the level of promotion; the square- 
 root model. This model allows diminishing marginal 
 returns to promotion.^ Models of this form were used 
 to study table grape promotion by Alston et al. (1997). 
 
 6. When AD and PRO were included in a linear model, their coefficients were essentially equal. Thus, there was no 
 information gained by separating PROCPB or PROSUN into their components and we would lose degrees of free- 
 dom in doing so. 
 
 7. A consequence of including the square root of promotion, rather than the quantity of promotion, is that this transfor- 
 mation imposes diminishing marginal returns on the demand response for promotion; the linear model is character- 
 ized by constant marginal returns. The marginal return to promotion refers to the incremental benefit from increas- 
 ing promotional effort by a small amount, say one dollar Diminishing marginal returns means that each incremental 
 dollar spent on promotion brings forth a smaller benefit than the last. It is preferable to have a structure that imposes 
 (or at least permits) diminishing returns for two related reasons. First, it would be uneconomic to choose quantities 
 of promotion in a range of constant or increasing marginal returns. Second, in order to solve for optimal promotion, 
 we require a model with diminishing returns. Similar conditions do not apply to the other variables in the model, 
 since they are not chosen by the prune industry. 
 
 19 
 
Unlike Alston et al. (1997), we have more than one elude the square root of all promotion. Model 3.4a 
 category of promotion expenditure, and, in particu- allows the marginal effects of CPB and Sunsweet pro- 
 lar, we wish to include separate variables represent- motion expenditures to differ, while model 3.4b sets 
 ing generic promotion by CPB and brand promotion the promotion parameters in equation 3.4a equal, so 
 by Sunsweet. Accordingly, we consider two types of that the marginal effects of CPB and Sunsweet pro- 
 models that include promotion variables in square-root motion expenditures are forced to be the same, 
 form. The first type consists of two models that in- 
 
 Q = b + b RPP + b RPS + b REXP + (b RPROCPB + b RPROSUN P'- + b SPR + b WIN + 
 
 t 0 PP I PS t EXP t CPB t SUN I SPR I WIN I 
 
 b SUM +b TIME + e . (3.4a) 
 
 SUM IT II 
 
 Q =b +b RPP +b RPS +b REXP +b (RPROCPB + RPROSUN P'- + b SPR +b WIN +b SUM 
 
 I 0 PP I PS t EXP I PRO I I SPR I WIN I SUM t 
 
 + b TIME + e . (3.4b) 
 
 T I I 
 
 We also consider two other models based on sepa- tion 3.4c, the marginal effects of CPB and Sunsweet 
 
 rate square roots of promotion expenditures. In equa- promotion expenditures are allowed to differ: 
 
 Q =b +b RPP +b RPS +b REXP +b (RPROCPB yi'- + b (RPROSUN + b SPR+b WIN + 
 
 t 0 PP I PS I EXP I CPB I SUN I SPR I WIN I 
 
 b SUM + b TIME + e . (3.4c) 
 
 SUM IT It 
 
 A special case of equation 3.4c is given by assum- Note that equation 3.4d can also be seen as a special 
 
 ing that Sunsweet promotion has no effect on the total case of equation 3.4a by setting b^^^ = 0. 
 market for prunes. This is shown by equation 3.4d. 
 
 Q = b + b RPP + b RPS + b REXP + b (RPROCPB j''^ + 1, SPR+b WIN + b SUM + b TIME 
 
 10 PP I PS I EXP I CPB t SPR t WIN I SUM I T I 
 
 + e . (3 Ad) 
 
 t 
 
 We can test models 3.4b and 3.4d as special cases 
 of model 3.4a, and model 3.4d can also be tested 
 against 3.4c. Models 3.4a and 3.4c, however, are not 
 nested as special cases, and cannot be tested against 
 one another using conventional nested tests. 
 
 In all of these models, the b coefficients are multi- 
 pliers that, holding the other independent variables 
 constant, translate changes in the prices and other 
 right-hand-side variables into changes in quantities 
 consumed. For the promotion variables, RPROCPB 
 and RPROSUN, however, the multipliers translate 
 changes in the square-root transformations of the un- 
 derlying variables into changes in quantities con- 
 sumed. Thus, the multipliers here do not represent 
 partial derivatives for the promotion variables, e, rep- 
 resents residual changes in per capita quantities con- 
 sumed that are not accounted for by changes in the 
 right-hand-side variables, is sometimes referred to 
 as the "error" term, since it can be thought of as the 
 
 error in predicting using only the right-hand-side 
 variables. These residuals are typically assumed to be 
 normally distributed random variables, with an ex- 
 pected value of zero and a constant variance. 
 
 Evaluating the Structure of the Model Holding Promotion 
 Constant 
 
 The strategy for estimation was first to evaluate the 
 structure of the model holding constant the specifica- 
 tion of the promotion variables. To do this, we used a 
 single aggregate promotion variable (i.e., the model in 
 equation 3.4b). This model, which performed well, was 
 not rejected by subsequent tests. The second step was 
 to explore the effects of different specifications of pro- 
 motion variables, holding constant the rest of the 
 model. 
 
 We estimated the model in equation 3.4b and then 
 exanuned the estimated coefficients to see whether they 
 satisfied our expectations, based on the theory laid out 
 
 20 
 
above, and, at the same time, examined the residuals 
 to see whether their behavior was consistent with the 
 conventional econometric assumptions. In addition, 
 diagnostic tests were applied to see whether the va- 
 lidity of the model and its parameters could be rejected 
 by the results from alternative models, using alterna- 
 tive functional forms, and making different assump- 
 tions about whether prices and promotion are statisti- 
 cally exogenous.* Only if a model passes all of these 
 tests — that is, it is consistent with economic theory and 
 our expectations about the signs and sizes of the coef- 
 ficients, has well-behaved residuals, and is not rejected 
 by an alternative specification — can we confidently 
 take the next step and use the estimated model to simu- 
 late alternative market scenarios. 
 
 The results from estimating the complete model, 
 as specified in equation 3.4b, are shown in the first 
 column (column 1) of table 3.3. The other five col- 
 umns (columns 2 through 6) report the results of OLS 
 regressions, based on equation 3.4b, that were derived 
 by restricting various combinations of the coefficients 
 on the price of substitutes, total expenditure, and the 
 time trend to be zero. Note that it was difficult to de- 
 rive estimates for all of the hypothesized economic re- 
 lationships. The effects of price movements (and esti- 
 mates of the own-price elasticity of demand) were hard 
 to determine, because there was very little price change 
 for prunes and other dried fruit during the time pe- 
 riod under consideration. In addition, income effects 
 were difficult to distinguish from the effects of other 
 variables characterized by smooth trends over time. 
 These patterns in the exogenous variables meant that 
 several of the estimated coefficients were not statisti- 
 cally significant. 
 
 Restricting the insignificant coefficients to zero (by 
 dropping the corresponding variables from the regres- 
 sion) could have undesirable consequences for the in- 
 terpretation of the remaining coefficients. For instance 
 (as will be seen in the models presented later), if the 
 time-trend variable is left out of the regression, the 
 price variable "picks up" the declining trend in prune 
 consumption and, thus, too much importance is given 
 to the price variable (which can be seen in the elastici- 
 ties greater than one in absolute value when trend is 
 dropped from the regression). Since it is not clear that 
 other dried fruits substitute for prunes, and other dried 
 fruit prices showed little movement during this pe- 
 riod, models were tried excluding these "substitute" 
 
 prices. Further, income or total expenditure was not 
 expected to have much effect on prune consumption 
 because of the short time-period covered, and only en- 
 tered the model in a statistically significant way when 
 trend was not included in the regression. Since the 
 income effect was negative if a trend variable was left 
 out of the model, the strong suspicion is that REXP 
 simply took the place of the trend variable. Thus, 
 REXP, too, was left out of several regressions. 
 
 Because of scaling, the coefficients on the promo- 
 tion variable appear very small (the dependent vari- 
 able, per capita consumption per month, averaged 
 0.015 pounds while the square root of total expendi- 
 tures on promotion averaged $1,212.60). Therefore, a 
 more useful statistic is the elasticity measure, which 
 accounts for the relative size of the variables by using 
 percentage changes. Each of the elasticity coefficients 
 is calculated as the percentage change in per capita 
 prune consumption from a one percent change in the 
 independent variable of interest. Thus, an elasticity 
 of 0.2 means that a 10 percent increase in the indepen- 
 dent variable leads to a 2 percent increase in prune 
 consumption. 
 
 Regression results for variations on model 3.4b are 
 displayed in table 3.3. In this table, the elasticities are 
 displayed in brackets beneath the coefficient estimates, 
 and have been calculated at the means of the economic 
 variables. The figures in parentheses are t-statistics. 
 The Adjusted R^ statistic indicates the proportion of 
 the variation in consumption that has been accounted 
 for by the independent variables included in the equa- 
 tion. The R^ statistics range from 69.2 percent in model 
 2 to 81.1 percent in model 1 in table 3.3. Finally, the 
 Durbin-Watson statistic is used to test for the presence 
 of first-order autocorrelation, which was statistically 
 significant only in model 2 in table 3.3. 
 
 As mentioned above, the period of study was one 
 of very little price movement. Not surprisingly, it was 
 difficult to find significant effects on quantity de- 
 manded from changes in prices. For all the models 
 presented in table 3.3, only the model 2 had a statisti- 
 cally significant estimate of the demand response to 
 price.'* In model 2, the own-price elasticity of demand 
 for prunes is about -1 .8 at the mean of the sample data, 
 a surprisingly large estimate of the elasticity. Com- 
 paratively, a study of the U.K. market suggested a price 
 elasticity of -0.7 for the total market and -1.23 for 
 Sunsweet (Thorogood 1994). Model 2, however, does 
 
 8. If either promotion or price is endogenous, in the sense that their values are affected by changes in quantities con- 
 sumed, as well as causing changes in quantity consumed, the econometric model may suffer from simultaneous- 
 equations bias. Such bias, if it exists, results from correlation of an explanatory variable with the error term, and may 
 lead to a misstatement of the demand response to changes in price or promotion. The direction of such bias is hard to 
 predict in the absence of a specific alternative model in which these variables are simultaneously determined. 
 
 9. Unless otherwise noted, all statements regarding statistical significance are based on a 95 percent confidence level. 
 
 21 
 
Table 3.3. Coefficient and Elasticity Estimates from the Monthly Demand Models 
 
 Independent Variables 
 
 (1) 
 
 (2) 
 
 (3) 
 
 (4) 
 
 (5) 
 
 (6) 
 
 Constant 
 
 0.015 
 
 0.037 
 
 0.068 
 
 0.019 
 
 0.049 
 
 -0.028 
 
 
 (0.283) 
 
 (4.148) 
 
 (6.470) 
 
 (2.340) 
 
 (6.209) 
 
 (-0.551) 
 
 RPP 
 
 -0.003 
 
 -0.013 
 
 -0.005 
 
 -0.004 
 
 -0.006 
 
 -0.003 
 
 
 (-0.794) 
 
 (-3.185) 
 
 (-1.382) 
 
 (-1.213) 
 
 (-1.568) 
 
 (-0.743) 
 
 
 [-0.437] 
 
 [-1.824] 
 
 [-0.687] 
 
 [-0.642] 
 
 [-0.822] 
 
 [-0.428] 
 
 RPS 
 
 -0.009 
 
 
 -0.010 
 
 
 
 
 
 (-2.228) 
 
 
 (-2.541) 
 
 
 
 
 
 [-1.433] 
 
 
 [-1.591] 
 
 
 
 
 (RPROMOy' 
 
 0.535E-2 
 
 0.421E-2 
 
 0.539E-2 
 
 0.523E-2 
 
 0.522E-2 
 
 0.519E-2 
 
 
 (6.339) 
 
 (4.046) 
 
 (6.384) 
 
 (5.969) 
 
 (5.848) 
 
 (5.901) 
 
 
 [0.214] 
 
 [0.168] 
 
 [0.215] 
 
 [0.209] 
 
 [0.209] 
 
 [0.207] 
 
 REXP 
 
 0.129E-5 
 
 
 -0.136E-5 
 
 
 -0.153E-5 
 
 0.247E-5 
 
 
 (0.505) 
 
 
 (-4.248) 
 
 
 (-4.585) 
 
 (0.946) 
 
 
 [1.585] 
 
 
 [-1.676] 
 
 
 [-1.879] 
 
 [3.036] 
 
 WIN 
 
 0.002 
 
 0.001 
 
 0.002 
 
 0.001 
 
 0.002 
 
 0.001 
 
 
 (3.797) 
 
 (2.620) 
 
 (4.376) 
 
 (3.358) 
 
 (3.446) 
 
 (2.990) 
 
 SPR 
 
 0.002 
 
 0.002 
 
 0.002 
 
 0.002 
 
 0.002 
 
 0.002 
 
 
 (3.862) 
 
 r\i o\ 
 
 (3.013) 
 
 (4.510) 
 
 (4.400) 
 
 (4.494) 
 
 (3.699) 
 
 SUM 
 
 -0 0005 
 
 -0.001 
 
 -0 0003 
 
 -0 0005 
 
 -0.0004 
 
 -0 0006 
 
 
 (-1.105) 
 
 (-1.872) 
 
 (-0.826) 
 
 (-1.188) 
 
 (-1.106) 
 
 (-1.418) 
 
 TIME 
 
 -0.0001 
 
 
 
 -0.535E-4 
 
 
 -0.0001 
 
 
 (-1.047) 
 
 
 
 (-4.819) 
 
 
 (-1.544) 
 
 Adjusted 
 
 0.811 
 
 0.692 
 
 0.810 
 
 0.794 
 
 0.787 
 
 0.793 
 
 Durbin-Watson 
 
 2.114 
 
 1.546 
 
 2.115 
 
 2.365 
 
 2.390 
 
 2.291 
 
 Notes: t-statistics are in parentheses, elasticities (at means) are in brackets. Elasticities in the row (RPROMOy- are the 
 elasticities of demand with respect to RPROMO. Model 1 is equation 3.4b in the text. 
 
 22 
 
not include a time trend, a price of substitutes, or an 
 income variable. We suspect this is a misspecification 
 and causes the regression to attribute a declining trend 
 in the consumption of prunes to exaggerated price ef- 
 fects. When a trend is included, as in models 4 and 5 
 of table 3.3, a much smaller own-price elasticity is ob- 
 tained. 
 
 Notice that the coefficient on the price of substi- 
 tutes, included in models 1 and 3 of table 3.3, is statis- 
 tically significant but negative, which is contrary to 
 expectations. Other dried fruit prices and prune prices 
 moved closely together during this period. A mea- 
 sirre of income, or total expenditure, appears in mod- 
 els 1, 3, 5, and 6. As discussed above, when trend is 
 not included but income is, the income variable may 
 be acting like the trend variable; hence the negative 
 sign on the income coefficient in models 3 and 5. In 
 models 1 and 6, which include trend, the income ef- 
 fect is not statistically significant. While it may be the 
 case that income effects are important in determining 
 prune demand (especially in a monthly demand 
 model), it is more likely that income and trend are col- 
 linear, as both are increasing over the observation pe- 
 riod, and it is difficult for the regression estimation 
 procedure to distinguish the separate effects of the two 
 variables. 
 
 Seasonality is present in all of the estimated 
 monthly demand models. Using the Fall season as 
 the base, we see that the coefficient estimates for Win- 
 ter and Spring are significantly positive, whereas the 
 coefficient for Summer is insignificant. From these re- 
 sults, holding price and other factors constant, one 
 concludes that prune demand is significantly greater 
 in Winter and Spring than in Summer or Fall. 
 
 An examination of prune consumption over the 
 1990s shows a decline in demand, so it was expected 
 that the time-trend coefficient would be negative. In 
 fact, the trend coefficient was always negative. For 
 the models presented in table 3.3, trend is significant 
 in model 4 and insignificant in models 1 and 6, with 
 the latter results probably due to the high correlation 
 of trend with income. 
 
 The promotion variable was statistically significant 
 at the 99 percent confidence level in each of the mod- 
 els in table 3.3. In addition, except for model 2 (which 
 excluded the price of substitutes, income, and trend), 
 the estimated coefficients on promotion and corre- 
 sponding elasticities were very stable across model 
 specifications. 
 
 Model 1 is preferred over each of the other models 
 in table 3.3. In model 1, the price coefficient is nega- 
 tive, and of plausible magnitude (although not statis- 
 tically significantly different from zero), with an elas- 
 ticity at the mean of -0.4. The estimated coefficient on 
 total per capita expenditure is positive, but not statis- 
 tically significant. The corresponding elasticity of de- 
 
 mand with respect to income is 1.6, which is larger 
 than expected. The statistically significant cross-price 
 elasticity is -1 .4. This suggests that other dried fruit is 
 a complement rather than a substitute for prunes, or 
 that the price of other dried fruit may be acting as a 
 proxy for other variables. The elasticity of demand 
 with respect to promotion in the preferred model is 
 0.21 . (This elasticity was also 0.21 in all of the other 
 models in table 3.3 except model 3, which was the least 
 acceptable of all.) 
 
 Diagnostic Tests 
 
 As noted above, differences between the predic- 
 tions from the model and actual per capita consump- 
 tion, e^, are referred to as errors or residuals. Diagnos- 
 tic tests can be used to evaluate the properties of the 
 residuals. Evidence that the residuals do not satisfy 
 certain theoretical properties may be interpreted as an 
 indication of model misspecification, such as having 
 omitted relevant explanatory variables or having used 
 the wrong functional form. The DIAGNOSTIC proce- 
 dure in SHAZAM was used to perform a range of tests 
 for heteroskedasticity and omitted variables, and the 
 RESET test (Ramsey 1969; Maddala, 1992; White et al, 
 1990) for misspecification. 
 
 Missing Variables : In a misspecified model, or a 
 model with significant missing variables, the effects 
 of missing variables are relegated to the error term, 
 and this can cause several estimation problems. We 
 tried three variants of Ramsey's specification error test 
 (RESET), in which predictions from the model, Q, 
 were added to three additional regressions of the de- 
 pendent variable on the independent variables. First, 
 the model is re-estimated with added^ then with 
 both Q^and Q ' added, and, finally, with Q^, Q', and 
 Q"* added. In each case, the statistical significance of 
 the added regressors was tested. Passing the RESET 
 test requires an insignificant test statistic for all three 
 tests: no evidence of misspecification. Failing the RE- 
 SET test would suggest that the model should be re- 
 jected, but the test itself would not imply any particu- 
 lar alternative. In model 1 in table 3.3, these test sta- 
 tistics were not statistically significant; hence, there is 
 no statistical evidence of misspecification. 
 
 Heteroskedasticity : Ordinary least squares (OLS) 
 presumes that the variance in the error term is con- 
 stant across observations. If this is not the case (i.e., if 
 heteroskedasticity is present), then OLS yields biased 
 estimates of the standard errors for the coefficients. 
 Although the coefficient estimates themselves would 
 be unbiased, the t-ratios discussed above and pre- 
 sented in table 3.3 would be biased, invalidating our 
 hypothesis tests. Several tests for heteroskedasticity 
 were run on model 1 in table 3.3. Of seven tests per- 
 formed, the conclusions were mixed: the hypothesis 
 of no heteroskedasticity rejected in four of the seven 
 
 23 
 
tests. 
 
 Taking the results of these tests as evidence of 
 heteroskedasticity, we re-estimated model 1 with an 
 alternative set of standard errors obtained using 
 White's (1980) heteroskedasticity-consistent covari- 
 ance matrix. For large enough samples, these estimates 
 allow confidence in our hypothesis tests, as the stan- 
 dard errors of the coefficients that are estimated in this 
 manner provide consistent estimates of the true stan- 
 dard errors. Thus, asymptotically, tests of hypotheses 
 using these estimates are not biased by ignoring 
 heteroskedasticity. Estimating the standard errors for 
 the estimated coefficients in model 1 in table 3.3 using 
 White's heteroskedasticity-consistent covariance ma- 
 trix shows no significant difference in t-ratios relative 
 to the original OLS results. 
 
 Alternative Specifications of Promotion Variables 
 
 We now turn to a consideration of alternative speci- 
 fications of promotion variables, as shown in equa- 
 tions 3.4a-d. The results from estimating these alter- 
 native models are shown in table 3.4. Equations 3.4a- 
 d are denoted models 1-4 in table 3.4. Model 2 in table 
 3.4 is a special case of model 1 in that table, in which 
 CPB and Sunsweet promotion are aggregated into a 
 single promotion variable. This restriction is not re- 
 jected. Model 4 in table 3.4 is also a special case of 
 model 1 (as well as model 3), and the implied restric- 
 tion (that Sunsweet promotion has no effect on the 
 market) is rejected conclusively. Models 1, 2 and 3 of 
 table 3.4 are all acceptable models, but model 4 is not. 
 These three models all say essentially the same things 
 about the demand for prunes. The own-price elastic- 
 ity of demand is inelastic, falling in a range of -0.3 to - 
 0.5, the income elasticity of demand is about 1 .2 to 1 .6, 
 the cross-price elasticity of demand with respect to the 
 price of substitutes is about -1.4, and the elasticities of 
 demand with respect to promotion are virtually iden- 
 tical among the three models. It should be noted that 
 a negative cross-price elasticity could imply that the 
 "substitute" good is, in fact, a complement. While it 
 may be the case that dried fruit consumption moves 
 in a complementary fashion with prune consumption, 
 we suspect that the negative cross-price elasticity is 
 capturing some of the downward trend in prune con- 
 sumption that is not otherwise reflected in the trend 
 variable. The elasticity of demand with respect to 
 Prune Board promotion is estimated as between 0.048 
 and 0.052, while the elasticity of demand with respect 
 to Sunsweet promotion ranges from 0.159 to 0.168. It 
 is important to note that these elasticities refer to the 
 effect of Prune Board and Sunsweet promotion on to- 
 tal sales of dried prunes. In particular the Sunsweet 
 elasticity cannot be used to evaluate the effectiveness 
 or profitability of Sunsweet promotion. To do so would 
 require an elasticity of demand for Sunszueet prunes 
 
 with respect to Sunsweet promotion. The elasticity 
 estimates are all plausible (as discussed above in de- 
 tail in relation to model 2 of table 3.4). The models 
 explain a high proportion of the variation in prune 
 consumption and appear to have generally acceptable 
 statistical properties. For the models in table 3.4, model 
 4 failed one of the three RESET tests, indicating that 
 omission of Sunsweet' s promotion does result in a 
 misspecified model. 
 
 Within-Sample Goodness of Fit : Models 1-3 in table 
 3.4 fit the data generally well, and the variables in- 
 cluded explain about 81 percent of the variation in 
 consumption of prunes. The close correspondence 
 between actual per capita consumption and the esti- 
 mated (fitted) values for model 2 in table 3.4 over the 
 sample period is shown in figure 3.2a. The lower por- 
 tion of the figure (3.2b) also includes a plot for model 
 2, showing the fraction of the fitted values accounted 
 for by all of the variables other than promotion. In 
 other words, it is the fitted values net of the estimated 
 effects of promotion (calculated by subtracting 
 0.535xW-^x(RPROCPB^ + RPROSUN)° '^ from the fitted 
 value in each month). The importance of promotion 
 in the overall demand for prunes is clearly illustrated 
 in figure 3.2b. 
 
 Autocorrelation : When the error term from one 
 period is correlated with the error term in the next pe- 
 riod, OLS standard errors are biased. A test for 
 autocorrelation of the error terms was performed us- 
 ing the Durbin-Watson test for first-order 
 autocorrelation. The Durbin-Watson statistics for 
 models 1-3 in table 3.4 are around 2.1, suggesting an 
 absence of first-order autocorrelation in the residuals. 
 
 Simultaneity and Endogeneity : If the price of 
 prunes and promotional expenditures are statistically 
 endogenous, then OLS is an incorrect procedure and 
 we must use some other procedure (e.g., two-stage 
 least squares) to account for the endogeneity. 
 Endogeneity means that one or more of our indepen- 
 dent variables is correlated with the error term, which 
 violates an assumption in OLS. For instance, when 
 promotion causes demand to increase, prune sales in- 
 crease, which might plausibly lead to further expen- 
 ditures on promotion. Such a feedback from consump- 
 tion to promotion means that promotion expenditures 
 are not exogenous. The important question is whether 
 such feedback appears to be statistically significant in 
 our model; in particular, feedback from consumption 
 to either promotion or prices of prunes. 
 
 Prices and promotional expenditures are statisti- 
 cally exogenous, as we use the term, if we do not ap- 
 pear to bias the estimated coefficients by making the 
 assumption that price and promotion are predeter- 
 mined. In order to evaluate this question, Hausman 
 tests for exogeneity were applied. We performed three 
 tests. First we tested whether prices alone were exog- 
 
 24 
 
Table 3.4. OLS Estimates Comparing Different Specifications of Promotional Expenditures in the Preferred 
 Monthly Model 
 
 Independent Variables 
 
 (1) 
 
 (2) 
 
 
 
 
 n m Q 
 
 O.KJio 
 
 O.Ui9 
 
 0.083 
 
 
 \V.jjU) 
 
 \\J.Z.OO) 
 
 (V.jW) 
 
 (1.367) 
 
 
 -0.291E-2 
 
 -0.305E-2 
 
 -0.229E-2 
 
 -0.153E-2 
 
 
 (-V./'iV) 
 
 
 \-v.D/y) 
 
 (-0.346) 
 
 
 
 t-U.4J/ J 
 
 l-U.J2yj 
 
 1-0.22UJ 
 
 RPS 
 
 -0.904E-2 
 
 -0.910E-2 
 
 -0.953E-2 
 
 -0.81 7E-2 
 
 
 (-ZAoV) 
 
 (-2.228; 
 
 (-2.324) 
 
 (-1.787) 
 
 
 
 t-i.4ooj 
 
 L-i.Dul J 
 
 l-1.2o7J 
 
 RPRuCPB 
 
 0.305E-4 
 
 0.535E-2 
 
 0.280E-2 
 
 0.499E-2 
 
 
 
 \o.ooy) 
 
 (2.4yo; 
 
 (4.824) 
 
 
 LU.UDZJ 
 
 LU.U4QJ 
 
 LU.UaUJ 
 
 rn nQoi 
 [U.Uoyj 
 
 RPRuSUN 
 
 0.270E-4 
 
 0.535E-2 
 
 0.449E-2 
 
 
 
 
 Ko.ooy) 
 
 
 
 
 [0.161] 
 
 [0.168] 
 
 [0.159] 
 
 
 At Ai 
 
 U. WZn-D 
 
 U.12yr,-D 
 
 0.y99b-6 
 
 -0.264b-5 
 
 
 (0.34) 
 
 (0.505) 
 
 (0.343) 
 
 (-0 866) 
 
 
 [1.248] 
 
 [1.585] 
 
 [1.230] 
 
 [-3.244] 
 
 WIN 
 
 0.182E-2 
 
 0.181E-2 
 
 0 183E-2 
 
 0.219E-2 
 
 
 (3.76) 
 
 (3.797) 
 
 (3.823) 
 
 (4.176) 
 
 SPR 
 
 0.179E-2 
 
 0.173E-2 
 
 0.178E-2 
 
 0.301E-2 
 
 
 (3.11) 
 
 (3.062) 
 
 (3.152) 
 
 (6.169) 
 
 SUM 
 
 -0.405E-3 
 
 -0.452E-3 
 
 -0.380E-3 
 
 0.220E-3 
 
 
 (-0.83) 
 
 (-1.105) 
 
 (-0.791) 
 
 (0.438) 
 
 TIME 
 
 -0.826E-4 
 
 -0.911 E-4 
 
 -0.820E-4 
 
 0.355E-4 
 
 
 (-0.83) 
 
 (-1.047) 
 
 (-0.840) 
 
 (0.347) 
 
 Adjusted 
 
 0.806 
 
 0.811 
 
 0.810 
 
 0.761 
 
 Log-likelihood 
 
 287.092 
 
 287.071 
 
 287.617 
 
 281.198 
 
 Durbin-Watson 
 
 2.118 
 
 2.114 
 
 2.124 
 
 2.078 
 
 Notes: t-statistics are in parentheses, elasticities (at means) are in brackets. Coefficient estimates on promotion variables (the 
 b's) are calculated as (h^j,/RPROCPB+b^^^*RPROSUW^ in model 1; as h*(RPROCPB+RPROSUN)''^ in model 2; as 
 h^p^*RPROCPB'^^ and b^^^*RPROSUhP'' in model 3, and as b^^^ *RPROCPB"- in model 4, Model 4, which does not use 
 Sunsweet's promotional expenditures, failed one of the three Ramsey tests for misspecification. 
 
 25 
 
Figure 3.2. U.S. Per Capita, Monthly Prune Consumption (Retail) - Actual versus Fitted Values, September 
 1992 to July 1996 
 
 (a) Actual and Fitted Values 
 
 0.025 
 
 (b) Fitted Values with and without Promotion, Using Actual Prices 
 
 0.025 
 
 c 0.020 
 
 o 
 I/) 
 
 u 
 
 ^ 0.015 
 
 IH 
 
 a, 
 
 tn 
 fi 
 
 ^ 0.005 
 
 0.000 
 
 — ^ 
 
 CN (N 
 
 On Os 
 
 0) 
 CD 
 
 > 
 
 o 
 
 2 
 I 
 
 ON 
 
 CN 
 
 Fitted 
 
 -Fitted: Without Promotion 
 
 OS 
 
 I 
 
 (0 
 
 CO 
 ON 
 
 (50 
 
 < 
 I 
 
 00 
 
 cn 
 
 ON 
 
 I 
 
 u 
 
 o 
 
 T 1 
 
 CO 
 
 o\ 
 
 c 
 
 « 
 
 CO 
 
 CN 
 
 
 
 
 
 On 
 
 ON 
 
 ON 
 
 ON 
 
 1 
 
 u 
 
 1 
 
 3 
 
 1 
 
 1 
 
 
 u 
 
 1) 
 
 < 
 
 
 0 
 
 1 
 
 D 
 
 1 — 1 
 
 o 
 
 CN 
 
 1 
 
 CN 
 
 ID 
 On 
 I 
 
 in 
 
 o\ 
 
 c 
 
 3 
 
 LD 
 On 
 I 
 
 CI. 
 
 CT) 
 
 On On 
 
 2^ ^ 
 
 I— I CO 
 
 > 
 o 
 
 2 
 
 NO 
 
 (N 
 
 NO 
 CJN 
 
 I 
 
 >^ 
 
 ns 
 
 ^ CN 
 
 Source: Actual values from Infoscan, IRI. Fitteci values from Authors' calculations. 
 
 26 
 
enous, second, whether promotion alone was exog- 
 enous, and finally, whether prices and promotion to- 
 gether may be treated as exogenous. The test compares 
 two different sets of estimates of the coefficients. Un- 
 der the null hypothesis of exogeneity, OLS is appro- 
 priate. Under the alternative hypothesis, price and 
 promotion are endogenous, and a different estimation 
 technique, instrumental variables or two-stage least 
 squares, must be used. The Hausman test involves a 
 comparison of the two sets of estimates. If the esti- 
 mates differ significantly, this is taken as evidence 
 against the null hypothesis of exogeneity. 
 
 To perform the Hausman test, then, we re-estimated 
 the model using an auxihary regression procedure (see 
 Davidson and Mackinnon 1993). We did so for each 
 case (price endogenous, promotion endogenous, and 
 both endogenous) for the models in table 3.4.'" In ev- 
 ery model, we rejected the hypothesis that both price 
 and promotion are exogenous. Specifically, we found 
 that price and CPB's promotion were endogenous in 
 models 1 and 3, price and the aggregate promotion 
 variable (RPROCPB+RPROSUW^ were endogenous 
 in model 2, and CPB's promotion alone was endog- 
 enous in model 4. Given these results, our next step 
 was to re-estimate each of the four models presented 
 in table 3.4 using a two-stage least squares (2SLS) pro- 
 cedure that treats both price and promotion as endog- 
 enous. 
 
 Results from Models Estimated by Two-Stage Least 
 Squares (2SLS) 
 
 The results from estimating models 1-4 using 2SLS 
 are reported in table 3.5. Since the only difference be- 
 tween the four models in table 3.4 and those in table 
 3.5 is the method of estimation (each model includes 
 the same variables), the OLS and 2SLS results are eas- 
 ily compared. While the results in the two tables are 
 broadly similar, close examination of the estimated 
 coefficients and their associated elasticities reveals 
 important differences resulting from estimation 
 method. First, there is much more variation in the es- 
 timated price and income elasticities among the four 
 models estimated using 2SLS methods in table 3.5 than 
 among the same four models estimated by OLS in table 
 3.4. Second, there are significant differences between 
 the estimated coefficients for the same models esti- 
 mated by different methods. Following is a brief com- 
 parison that highlights the results. 
 
 Among the 2SLS models in table 3.5, model 4 was 
 rejected as implausible, since it led to a positive own- 
 price elasticity of demand. Proper specification re- 
 
 quires that Sunsweet promotion be included as an ex- 
 planatory variable. Model 2, when estimated by 2SLS, 
 was also rejected because the hypothesis that the ef- 
 fects of generic and brand advertising were equal, as 
 reflected by the restriction of equal effects between the 
 two types of promotion, was not supported. The re- 
 maining two models in table 3.5 (1 and 3) imply es- 
 sentially the same things. In both models, the coeffi- 
 cient on income is negative (it was positive in the three 
 acceptable OLS models) but statistically insignificant. 
 The calculated income elasticities are very large and 
 negative (-6.7 to -7.5), values that would be difficult to 
 justify if they were statistically significant. While it is 
 plausible that prune demand falls with increased in- 
 come, these elasticities seem implausibly large. The 
 own-price elasticity of demand is substantially smaller 
 (-0.1 to -0.2) than in the OLS models, but the estimated 
 price coefficients are not statistically significant in ei- 
 ther set of models. Importantly, the elasticity of de- 
 mand with respect to CPB promotion is much larger 
 in the 2SLS models (0.13 to 0.15) than in the OLS mod- 
 els (about 0.05), while the elasticity of demand with 
 respect to Sunsweet promotion is smaller and no 
 longer statistically significant in the 2SLS models. 
 
 A Summary of Monthly Demand Estimates 
 
 In summary, the four-step procedure used to esti- 
 mate the monthly demand models for dried prunes 
 resulted in several alternatives being considered. In 
 the first step, a screening procedure was used to se- 
 lect the independent variables to be included in the 
 estimated model. The preferred model included vari- 
 ables for the retail price of prunes, the retail price of 
 other dried fruit, CPB and Sunsweet promotion, and 
 per capita income, all in real terms, as well as season- 
 ality (quarters) and a time trend. The variables in- 
 cluded in this model explained a high proportion of 
 the variation in prune consumption, and most of the 
 estimated coefficients, while not always significant, 
 were consistent with expectations. 
 
 The second step was to subject the estimated mod- 
 els to a set of diagnostic tests. These tests indicated 
 that three of the four models were properly specified 
 with regard to functional form and the variables in- 
 cluded. In the fourth model, omitting the effect of 
 Sunsweet promotion did result in a specification er- 
 ror While model 1 in table 3.3 had some evidence of 
 heteroskedasticity, it was a problem that could be ig- 
 nored without biasing the results. No corrections for 
 heteroskedasticity were required. 
 
 10. For tests on promotion, we tested whether only CPB's promotion was endogenous, whether only Sunsweet's promo- 
 tion was endogenous, and whether both were jointly endogenous. 
 
 27 
 
Table 3.5. 2SLS Estimates Comparing Different Specifications of Promotional Expenditures in the Preferred 
 Monthly Model 
 
 Independent Variables 
 
 (1) 
 
 (2) 
 
 (3) 
 
 (4) 
 
 
 0 1 A? 
 
 
 
 n lA^ 
 
 U. i4D 
 
 
 V 1 ./ 
 
 \ I .U/ O/ 
 
 
 
 KFF 
 
 -0.847E-3 
 
 -0.817E-2 
 
 -0.136E-2 
 
 0.371 E-2 
 
 
 \-V.lll 
 
 1 m ti"! 
 
 V 1 .31 
 
 v~U. l/v) 
 
 \\J.ty/ D) 
 
 
 f-D 1911 
 
 r.i 1 791 
 
 L-l.l/ZJ 
 
 
 
 RFb 
 
 -0.907E-2 
 
 -0.999E-2 
 
 -0.950E-2 
 
 -0.927E-2 
 
 
 r 1 81 ^ 
 
 
 (.-i.yDi; 
 
 (-l./OZ) 
 
 
 F-l 49Q1 
 
 f-l Ry^i 
 
 
 
 KFKULFB 
 
 U.76oh-4 
 
 0.699E-2 
 
 0.725E-2 
 
 0.901E-2 
 
 
 
 CC 979'! 
 
 0 114^ 
 
 
 
 LU. 1^/J 
 
 
 rn 19Q1 
 
 LU. 1/7 J 
 
 rn 1 A1 1 
 
 KFRUbUN 
 
 0.488E-5 
 
 0.699E-2 
 
 0.140E-2 
 
 
 
 en A'^\ 
 
 CC 979 
 
 1,0.// /; 
 
 ^^n i^i^i 
 
 
 
 [0.034] 
 
 [0.220] 
 
 [0.050] 
 
 
 
 -U.DUotl-D 
 
 n 01 Dv A 
 
 -U.D4t>C-D 
 
 
 
 (-1.39) 
 
 (-0.315) 
 
 (-1.132) 
 
 (-1.749) 
 
 
 [-7.482] 
 
 [-1.119] 
 
 [-6.711] 
 
 [-8.289] 
 
 WIN 
 
 0.172E-2 
 
 0.152E-2 
 
 0.174E-2 
 
 0.173E-2 
 
 
 (2.89) 
 
 (2.811) 
 
 (3.037) 
 
 (2.742) 
 
 SPR 
 
 0.303E-2 
 
 0.142E-2 
 
 0.286E-2 
 
 0.336E-2 
 
 
 (3.59) 
 
 (2.803) 
 
 (2.819) 
 
 (5.764) 
 
 SUM 
 
 0.773E-3 
 
 -0.302E-3 
 
 0.651E-3 
 
 0.103E-2 
 
 
 (0.97) 
 
 (-0.678) 
 
 (0.726) 
 
 (1.575) 
 
 TIME 
 
 0.142E-3 
 
 -0.144E-4 
 
 0.124E-3 
 
 0.157E-3 
 
 
 (1.00) 
 
 (-0.145) 
 
 (0.803) 
 
 (1.233) 
 
 Durbin-Watson 
 
 2.101 
 
 2.128 
 
 2.116 
 
 2.112 
 
 Notes: t-statistics are in parentheses, elasticities (at means) are in brackets. Coefficient estimates on promotion variables (the 
 b's) are calculated as {h^^/RPROCPB+h^^^*RPROSUW'^ in model 1; as h*{RPROCPB+RPROSUW^' in model 2; as 
 h^^^*RPROCPB'''- and \^^*RPROSUN"^ in model 3, and as b^^^ *RPROCPB"^ in model 4. The added instrumental variables 
 used in the 2SLS estimation were a squared trend variable, the real price of raisins, and the real price of all other dried fruit. 
 Endogenous, right-hand-side variables in the 2SLS models were chosen based on Hausman tests. These endogenous vari- 
 ables are RPROCPB+RPROSUN and RPP in model 2; RPP and RPROCPB in models 1 and 3, and RPROCPB alone in model 
 4. 
 
 28 
 
The third step in the monthly estimation procedure 
 was to retain the variables included in the preferred 
 model and investigate four alternative specifications 
 of the promotion variables. Three of the four models 
 estimated using OLS procedures resulted in statisti- 
 cally acceptable results that were quite stable from 
 model to model. The fourth model, which excluded 
 Sunsweet promotion, was found to be misspecified. 
 Regardless of the square-root form used, CPB and 
 Sunsweet promotion expenditures always had statis- 
 tically significant positive effects on prune demand. 
 The estimated elasticities for CPB promotion expen- 
 ditures ranged from 0.048 to 0.052, while the elastici- 
 ties for Sunsweet promotion expenditures ranged from 
 0.159 to 0.168. 
 
 Application of the Hausman test resulted in rejec- 
 tion of the null hypothesis that price and promotion 
 were exogenous variables in the OLS estimates in table 
 3.4, and acceptance of the alternative hypothesis that 
 they were, instead, endogenous. The fourth and final 
 step was to re-estimate each of the OLS models pre- 
 sented in table 3.4 using 2SLS procedures. The 2SLS 
 results, while preferred from a statistical standpoint, 
 do raise some questions. The coefficients and elastici- 
 ties are much more variable among models than those 
 estimated by OLS. The price elasticities estimated by 
 2SLS were much smaller than those estimated by OLS, 
 and the 2SLS income elasticities became very nega- 
 tive while the coefficients on time became positive. The 
 elasticity of demand with respect to CPB promotion 
 increased in the 2SLS estimates, while Sunsweet pro- 
 motion elasticities decreased and became statistically 
 insignificant. While we are unsure of the exact 
 "causes" of these changes, the nature of the data cer- 
 tainly played a role. Because of the relatively short 
 period covered by the data, there was little variation 
 in observed prices or income. There is also correla- 
 tion between income and time, and we were unable to 
 identify a statistical substitute for dried prunes. Hence, 
 even though the statistical tests support the use of 
 2SLS, the OLS models are more satisfactory in terms 
 of the consistency of the parameter estimates with prior 
 beliefs about plausible or likely values. And, the fact 
 that the OLS models imply smaller elasticities of de- 
 mand response to CPB expenditure means that, in one 
 sense, the OLS estimates are more conservative. Im- 
 portantly, across all specifications, the effect of CPB 
 promotion was statistically significant, and the coeffi- 
 cients were very similar among the models. 
 
 3.3 Aggregate Annual Demand Models, 1949-1995 
 
 Annual data were available for the period 1949 
 through 1995, as documented in section 2 of this re- 
 port. The essential theory for an annual model of de- 
 mand is identical to that for the monthly model. The 
 primary differences are that, with an annual model, 
 seasonality is absent, and the longer time period of 
 analysis may allow some more useful variation in rela- 
 tive prices and incomes to have occurred, as well as 
 the potential for explicitly accounting for the dynamic 
 effects of changing demographic variables, such as the 
 age distribution of the population. On the other hand, 
 the longer time period needed to estimate an annual 
 model also means that it is less likely that the param- 
 eters of any estimated model will remain stable over 
 time, without structural change. 
 
 Models 
 
 As with the monthly demand model, decisions 
 must be made about the functional form for demand 
 and the variables to be included. The equation for the 
 annual model is: 
 
 Q = B +B RPP + B RPS +B RING + 
 
 '^0 ^PP I ^PS t '^INC t 
 
 B IRPROMO + B AGE65 + 
 BTIME+B Q +e 
 
 '^T '^LAG f-1 t 
 
 (3.5) 
 
 This annual demand model (equation 3.5) includes 
 many of the same variables as the monthly demand 
 model (equation 3.4). The differences from the 
 monthly model relate to the dynamic specification, in 
 terms of the role of the fraction of people over 65 years 
 old, the aggregation of Sunsweet and CPB promotion, 
 and the inclusion of the lagged dependent variable. " 
 The age and trend variables were discussed earlier. 
 The lagged dependent variable often is interpreted as 
 reflecting partial adjustment of desired quantities con- 
 sumed from year to year, or habit persistence. How- 
 ever, it can also be included to detect the possibility of 
 an incorrect specification, rather than true dynamic 
 effects on consumption. 
 
 Data 
 
 The variables used in the annual demand model 
 are defined in table 3.6. Several of these variables were 
 graphed in section 2 of the report, and the trends were 
 
 11. Annual Sunsweet promotion data were not available for the years 1949 to 1968. 
 
 29 
 
Table 3.6. Description of Variables in the Annual Demand Model 
 
 Variable 
 
 Definition 
 
 Units 
 
 Data Source 
 
 TIME, 
 
 RPP. 
 
 RPR 
 
 RING, 
 
 RPROMO, 
 
 Year 
 
 Quantity of dried California prunes 
 shipped in the United States per 
 million people from August 1 of 
 year t to July 31 of year t+1 
 
 Real average price of prunes 
 received by growers, in year-of- 
 harvest 
 
 Real average price of raisins 
 received by growers in year-of- 
 harvest 
 
 Real average U.S. per capita 
 income in the calendar year after 
 the year-of-harvest 
 
 Percentage of U.S. population 65 
 years old or older in the calendar 
 year after the year-of-harvest 
 
 Real expenditures by the California 
 Prune Board and Sunsweet Growers 
 on all types of domestic promotion 
 (includes advertising, promotion, 
 and public relations), from August 
 1 of year t to July 31 of year t-i-1 
 
 years 
 
 pounds of dried 
 prunes per person 
 per year 
 
 real (1995=1.00) 
 dollars per pound 
 of prunes in 
 processed 
 condition 
 
 real dollars per 
 pound of raisins 
 
 real dollars per 
 person per year 
 
 fraction between 
 0 and 1 
 
 millions of real 
 dollars per year 
 
 Prune shipments from 
 California Prune 
 Board. U.S. 
 population from U.S. 
 Statistical Abstract 
 
 Agricultural 
 
 Statistics, 
 
 USDA 
 
 Agricultural 
 
 Statistics, 
 
 USDA 
 
 U.S. Statistical 
 Abstract, 
 
 U.S. Dept. of Commerce 
 
 U.S. Statistical 
 Abstract, U.S. Dept. 
 of Commerce 
 
 California Prune 
 Board and Sunsweet 
 Growers 
 
 Table 3.7. Summary Statistics of the Variables in the Annual Demand Model 
 
 Variable 
 
 N 
 
 Mean 
 
 Standard 
 Deviation 
 
 Minimum 
 
 Maximum 
 
 Q 
 
 47 
 
 0.601 
 
 0.224 
 
 0.372 
 
 1.153 
 
 RPP 
 
 47 
 
 0.639 
 
 0.152 
 
 0.396 
 
 1.026 
 
 RPR 
 
 47 
 
 0.666 
 
 0.250 
 
 0.433 
 
 1.296 
 
 RING 
 
 47 
 
 17.293 
 
 4.800 
 
 9.622 
 
 24.385 
 
 AGE65 
 
 47 
 
 0.105 
 
 0.015 
 
 0.081 
 
 0.130 
 
 RPROMO 
 
 47 
 
 3.821 
 
 2.541 
 
 0.0 
 
 8.776 
 
 TIME 
 
 47 
 
 1972 
 
 13.711 
 
 1949 
 
 1995 
 
 30 
 
discussed there. Summary statistics for these variables 
 are shown in table 3.7. Data can be found in appendix 
 table A3.2. 
 
 Some issues arose from the fact that the annual ob- 
 servation periods differed among the key economic 
 variables. This meant that some lead-lag relationships 
 among the observed variables could arise, even though 
 the underlying data generating process was one in 
 which contemporaneous observations of the indepen- 
 dent variables were relevant for the determination of 
 consumption. This issue is not important for the in- 
 come and the age variables, which simply do not vary 
 significantly from one year to the next. The only real 
 decision was whether to use the price from the fall 
 period at the beginning of the current year, in conjunc- 
 tion with the quantity and expenditure on promotion 
 from the same period through to the end of that twelve- 
 month period. To the extent that this price is not an 
 accurate representation of the relevant annual prices, 
 bias is introduced into the model. 
 
 Estimation Results and Selection of the Preferred Annual 
 Model 
 
 The annual demand model presented in equation 
 3.5 was estimated using the annual data described 
 above. As with the monthly model, the statistical ef- 
 fects of deleting various variables were investigated 
 as indicators of the robustness of certain aspects of the 
 model. The estimation results for models estimated 
 by OLS are shown in table 3.8. Variants of the model 
 shown in column (1) of the table were derived by re- 
 stricting selected parameters in the most general form 
 to zero (in other words, by dropping certain potential 
 explanatory variables). For each equation, the vari- 
 ables appearing in the model are those for which esti- 
 mated coefficients are reported. Note that the results 
 in table 3.8 were obtained using the square-root of to- 
 tal promotion.'^ 
 
 All of the models estimated had relatively high 
 adjusted values — the model with the lowest value 
 still explained over three-fourths of the variation in 
 per capita prune consumption. High values, how- 
 ever, are quite common with time-series data. Per- 
 haps the most striking aspect of these models is that 
 few of the coefficients are statistically significant. In 
 particular, the coefficient on promotion was not sta- 
 tistically significant in any of the annual models. We 
 checked whether this was also true of the model with 
 real promotion (RPROMO) in levels, rather than 
 square-root form, and found that RPROMO was sta- 
 tistically significant only if AGE65 and QSLAG were 
 
 left out of the model. Apparently, AGE65 and the 
 lagged dependent variable explain so much of the 
 variation in prune consumption that we cannot sepa- 
 rate out an effect of promotion. Only the coefficients 
 on the price of prunes, RPP, and last year's domestic 
 shipments, QSLAG, have consistently significant t-sta- 
 tistics, while the AGE65 variable appears to be signifi- 
 cant when QSLAG is deleted. Note also that the (one- 
 year) own-price elasticity of demand for prunes is quite 
 small, around -0.3 in all models. However, the long- 
 run elasticity, calculated by taking the price coefficient 
 and dividing by one minus the coefficient on the 
 lagged dependent variable, is roughly -1 . 
 
 When the lagged dependent variable is not in- 
 cluded, the Durbin-Watson statistic shows evidence 
 of significant autocorrelation. Even when QSLAG is 
 included in the model, there appear to be significant 
 problems with unexplained dynamic effects. Promo- 
 tion still does not have any statistically significant ef- 
 fect on demand, over the longer period covered by this 
 annual model, when such dynamic effects are ac- 
 counted for. 
 
 After examination of the results, the model closest 
 to being a "preferred" model in table 3.8 is the one 
 originally specified in equation 3.5, which appears in 
 column (1). However, this model shows no statisti- 
 cally significant effects of promotion. 
 
 This result might be more worrisome if we were 
 more confident about the aimual models. However, 
 application of the diagnostic tests that were described 
 in detail in section 3.2 confirmed that the annual mod- 
 els could not be used with much conviction. We used 
 the DIAGNOSTIC procedure in SHAZAM to perform 
 several tests on model (1) in table 3.8. We also per- 
 formed the tests on models where RPROMO enters 
 linearly. All three of the variants of Ramsey's specifi- 
 cation error test (RESET) led us to reject the hypoth- 
 esis of no misspecification in all of the models shown 
 in table 3.8. Not passing the model specification test 
 means that other tests cannot be relied upon. As can 
 be seen in the tables, dynamic effects must be ac- 
 counted for, as an autocorrelation correction appears 
 to be called for in the model where the lagged depen- 
 dent variable is not included. 
 
 12. Several other models and alternative estimation techniques were investigated. Corrections for autocorrelation pro- 
 vided results that were sinailar to those in table 3.8, as did models in which promotion entered linearly, rather than in 
 the square-root form. 
 
 31 
 
Table 3.8. Coefficient and Elasticity Estimates from Annual Prune Demand Models 
 
 Independent Variables 
 
 (1) 
 
 (2) 
 
 (3) 
 
 (4) 
 
 Constant 
 
 30.555 
 
 21.621 
 
 116.740 
 
 5.714 
 
 
 (1.754) 
 
 (1.414) 
 
 /A A\ 
 
 (4.284) 
 
 /A '~7/'/'\ 
 
 (0.766) 
 
 RPP 
 
 -0.257 
 
 -0.363 
 
 -0.305 
 
 -0.265 
 
 
 (-4.221) 
 
 (-2.887) 
 
 (-2.810) 
 
 (-4.294) 
 
 
 l-0.2/yj 
 
 [-0.385J 
 
 1-0.324] 
 
 [-0.288] 
 
 RPR 
 
 0.005 
 
 -0.129 
 
 -0.112 
 
 0.011 
 
 
 (0.111) 
 
 (-1.481) 
 
 (-1.501) 
 
 (0.244) 
 
 
 10.005J 
 
 1-0.143J 
 
 1-0.124] 
 
 ff\ r\i 
 
 [0.012] 
 
 RINC 
 
 0.007 
 
 -0.010 
 
 0.029 
 
 -0.003 
 
 
 (0.547) 
 
 (-0.454) 
 
 (1.363) 
 
 (-0.234) 
 
 
 IU.196J 
 
 [-0.292J 
 
 [0.845] 
 
 [-0.075] 
 
 RPROMO^i^ 
 
 0.007 
 
 -0.017 
 
 -0.001 
 
 0.004 
 
 
 (0.423) 
 
 (-0.533) 
 
 (-0.010) 
 
 (0.232) 
 
 
 [0.014] 
 
 [-0.034] 
 
 [-0.002] 
 
 ro 0081 
 
 1 IJVIL 
 
 -U.Ulo 
 
 -U.UlU 
 
 -U.Uol 
 
 -U.ULW 
 
 
 (-i.72y) 
 
 (-1.30O) 
 
 (-4.229) 
 
 / A ''7r\o\ 
 
 (-0.708) 
 
 AGE 65 
 
 8.346 
 
 
 33.476 
 
 
 
 (1.572) 
 
 
 (3.981) 
 
 
 QSLAG 
 
 0.721 
 
 
 
 0.780 
 
 
 (9.386) 
 
 
 
 (11.440) 
 
 
 [0.740] 
 
 
 
 [0.801] 
 
 AdjR2 
 
 0.945 
 
 0.779 
 
 0.838 
 
 0.943 
 
 Durbin-Watson 
 
 na 
 
 0.383 
 
 0.521 
 
 na 
 
 Sample 
 
 1948-1995 
 
 1949-1995 
 
 1949-1995 
 
 1948-1995 
 
 Notes: t-statistics are in parentheses, elasticities (at means) are in brackets. Elasticity in the row RPROMO^'^ is the elasticity of 
 demand with respect to promotional expenditure, not its square root, na = not applicable. 
 
 32 
 
4. SIMULATION MODEL AND BENEFIT-COST ANALYSIS 
 
 In this section, the estimated monthly demand pa- 
 rameters from the previous sections are used to esti- 
 mate the gross and net benefits to the Cahfornia prune 
 industry from its expenditures on promotion. Both 
 the OLS and 2SLS models of monthly retail demand 
 are used to show a range of estimated values for gross 
 and net benefits. The analysis includes the 
 misspecified models that omit brand promotion, to 
 provide a measure of the difference in estimated ben- 
 efits that would result from excluding, or not acquir- 
 ing data on, a relevant variable. 
 
 4.1 Approaches for Evaluating the Benefits from 
 Promotion 
 
 Measuring the welfare impacts of promotion 
 funded by check-offs requires (a) a conceptual struc- 
 tural model of the industry market equilibrium, (b) 
 estimates of supply and demand parameters that can 
 be used to define the values for the parameters in the 
 structural model, (c) estimates of the demand response 
 to promotion expenditures, and (d) information to al- 
 low a transformation of the effects of promotion 
 (through retail demand shifts) and assessments or 
 check-offs (through commodity supply shifts) into 
 measures of benefits and costs. 
 
 Conceptual Model of Supply and Demand 
 
 The econometric work discussed in the previous 
 section allows us to estimate how much the quantities 
 sold of prunes would increase in response to a given 
 increase in promotional expenditures, holding prices 
 (and other variables) constant. This does not, how- 
 ever, tell us how much sales will actually increase when 
 promotion changes, since prices cannot be assumed 
 to remain constant. Indeed, the increase in prices fol- 
 lowing a promotion-induced shift in demand is an 
 important source of the benefits realized by growers 
 and packers of prunes. In order to properly evaluate 
 the industry's demand-shifting activities, therefore, we 
 must account for both demand effects and the response 
 of supply to increased price. 
 
 Demand Shifts from Promotion : The diagram in 
 figure 4.1 illustrates the conceptual supply and de- 
 mand relationships for a typical year f. In the figure, 
 the line labeled S, represents the supply curve for 
 prunes. It shows the quantities available to domestic 
 consumers at various prices; at higher prices, more 
 fresh prunes are available domestically, while at lower 
 prices, larger quantities of prunes are diverted to other 
 uses, such as to the export market, or they may be left 
 unharvested. The line labeled D^(RPROCPB^) repre- 
 sents the demand curve: at higher prices, consumers 
 
 Figure 4.1. Conceptual Supply and Demand Model 
 
 Price 
 Dollars per pound 
 
 Pounds per year 
 
 33 
 
purchase a smaller quantity of prunes than at lower 
 prices, holding other factors constant. In particular, 
 the promotional expenditure by the CPB is held con- 
 stant at its actual value, RPROCPB^, along this demand 
 curve." The equilibrium price, at which quantities 
 supplied and demanded are the same, is the price ob- 
 served at point E: price P, is consistent with the ob- 
 served quantity Q^. 
 
 In this example, the effects of a ten percent increase 
 in CPB promotion, RPROCPB^ , are illustrated by the 
 outward shift in the demand curve. The new curve is 
 labeled D^(RPR0CPB»1.1). The econometric model 
 allows us to estimate the horizontal distance of the 
 demand shift, identified by A in the diagram. In the 
 OLS model given in equation 3.4b, for example (col- 
 umn 1 of table 3.3 or column 2 of table 3.4), the coeffi- 
 cient on {RPROCPB)''^ is 0.535x101 Suppose the ac- 
 tual monthly promotion expenditure is $1 million (in 
 1996 dollars). This means that a $0.1 million (i.e., ten 
 percent) increase in total promotional expenditures 
 would be expected to lead to an increase in per-capita 
 prune consumption of 0.535x10-2(1.10 ' - 1.00 ') = 
 0.26x10'^ pounds per month, if there is no change in 
 price. 
 
 Multiplying by the population {POP^ = 265.5 mil- 
 lion in 1996) yields the total horizontal demand shift 
 from a ten-percent increase in promotional expendi- 
 tures, about a 2.1 percent increase in consumption, at 
 constant prices. However, this is greater than the ac- 
 tual increase in consumption that would result. An 
 increase in price is needed to bring forth the additional 
 quantities to satisfy the increased demand. This is re- 
 flected in the fact that the supply curve slopes up. The 
 new equilibrium is given by the point where the new 
 demand curve crosses the supply curve, £'. Price and 
 quantity both increase to the new equilibrium values 
 
 p; and q;. 
 
 Producer Benefits : The evaluation of promotional 
 expenditures requires both an estimate of the incre- 
 ments in prices and quantities due to the expenditures, 
 and a measure of the costs of supplying the additional 
 quantities to the market. In the diagram, increased 
 prices call forth additional supplies; these supplies 
 come at a cost, which, in the case of a perennial crop, 
 may largely be the earnings foregone from other uses 
 of existing production (e.g., exports), rather than new 
 production. The sellers who were already in the mar- 
 ket at price P^ profit by the increase in price to P^'; their 
 gain is (P,' - P,)Q,, or the area P^'aEP^ in the diagram. 
 
 The gains to the sellers of additional supplies are much 
 smaller, as they must be reduced by the cost (includ- 
 ing revenues from foregone sales in other markets) of 
 the additional quantities. This benefit is given by the 
 area of the triangle aEE' in the diagram. The total gain 
 to prune producers is given by the area of the trap- 
 ezoid P^E'EP-, this represents the gain in producer sur- 
 plus resulting from the ten percent increase in promo- 
 tional expenditures. Changes in producer surplus co- 
 incide with changes in producers' economic profit, 
 from the production of prunes, in such a situation. The 
 only information required, in addition to the 
 econometrically estimated demand parameters for re- 
 sponses to prices and promotion, is information on the 
 supply response to price. 
 
 Costs of Assessments to Finance Promotion : The 
 gain in producer surplus is not adjusted for the cost of 
 the increase in promotional expenditures. To evalu- 
 ate the profitability of these expenditures, the gain 
 must be set against the cost. We use two measures of 
 cost. One measure is the total cost of the marginal 
 increase in expenditure. However, when the promo- 
 tional cost is financed by a per-pound assessment, 
 some of the incidence of the assessment falls on con- 
 sumers, as a result of increased retail prices. Thus, the 
 total cost may differ from the cost to producers. A sec- 
 ond measure compares the benefits to producers with 
 the producers' share of the cost, allowing that some of 
 the costs of the assessment are borne by consumers. 
 
 Figure 4.2 shows the same initial supply and de- 
 mand curves as in the above figure, labeled S^d^) and 
 D,, where X, represents the actual assessment per pound 
 in year f , and the equilibrium is at point E, with price 
 P, and quantity Q^. An increase in the assessment per 
 pound is reflected as a shift up of the supply curve to 
 S,(t,') by the amount of the increase (given by simply 
 adding the additional assessment to the previous price 
 at which producers would be willing to supply any 
 given quantity along the supply curve). This leads to 
 a new equilibrium, at point E', with a higher consumer 
 price P,', a smaller net producer price {b = P,' - x/), and 
 a smaller quantity produced and consumed, Q^'. 
 
 The extent of the consumer price increase depends 
 on the slopes of the supply and demand curves. If 
 supply were fixed and unresponsive to price (so that 
 the supply curve is a vertical line), there would be no 
 increase in the consumer price and all of the additional 
 assessment would be borne by producers. The more 
 
 13. The econometric work discussed in the previous section provides estimates of the shape and position of this line. In 
 particular, the inverse of the price coefficient is the line's slope (the price coefficient tells the change in quantity 
 demanded in response to a unit increase in price), while the horizontal position of the curve is given by the sum of the 
 products of the values of each of the other variables, in year or month t, times their corresponding coefficients. 
 
 34 
 
Figure 4.2. Incidence of Assessments 
 
 price-responsive (the flatter, or more price elastic) is 
 supply, the smaller will be the proportion of the as- 
 sessment borne by producers. 
 
 The additional amount of assessment revenue is 
 equal to 1,'Q,' - T^Q,. For small changes in the assess- 
 ment, this is approximately equal to the change in the 
 assessment times the final quantity: (x,'-X,)Q,'. In fig- 
 ure 4.2, this is equal to area P^'E'ah. This area corre- 
 sponds closely to the full social cost of the change in 
 assessment to finance a change in promotional expen- 
 diture of that amount (it leaves out the area of the tri- 
 angle E'Efl, which is negligible for the small changes 
 in assessments to be considered here). The loss of pro- 
 ducer surplus (or profit) associated with the same in- 
 crease in the assessment is equal to area P^Eab, only a 
 fraction of the total amount being spent. In the work 
 below, we compare producer benefits and the two al- 
 ternative measures of producer costs. 
 
 4.2 An Approximation Using Elasticities 
 
 Before turning to the specification of the supply side 
 and the simulation models for computing benefit-cost 
 ratios, we report results from an approximation pro- 
 cedure, using just the estimated elasticities of demand 
 with respect to price and promotion, and the promo- 
 
 tion intensity (promotion expenditure per dollar of 
 sales). 
 
 As discussed by Alston et al. (1997), much of the 
 literature on optimal primary product promotion rests 
 on two foundational papers on the economics of ad- 
 vertising: Dorfman and Steiner (1954) and Nerlove and 
 Waugh (1961). According to the Dorfman-Steiner theo- 
 rem, given fixed output, a monopolist will maximize 
 profits by setting the advertising budget such that the 
 increase in gross revenue resulting from a one dollar 
 increase in advertising expenditure is equal to the own- 
 price elasticity of demand for the product. That is, 
 
 dv a a , dv a 
 
 — -ri or — = — where a = . 
 
 da V rj da V 
 
 In this equation, a is the advertising expenditure, v 
 is the value of sales (the product of price, p and the 
 quantity sold, q), a is the elasticity of demand with 
 respect to advertising, and 7] is the absolute value of 
 the elasticity of demand with respect to the price. 
 
 The Dorfman-Steiner result may be applicable to a 
 number of primary products where output is fixed 
 (e.g., by a quota) and a marketing organization adver- 
 tises on behalf of producers. However, the more rel- 
 evant reference for a study of promotion by a producer 
 group without the ability to control output is that by 
 
 35 
 
4.1. In all of the instances examined, the estimated 
 benefit from a marginal increase in the promotion ex- 
 penditure exceeds 1.0. 
 
 4.3 Simulation Model 
 
 The simulation model combines results from the 
 monthly demand analysis with assumed supply pa- 
 rameters. We compare the pattern of consumption 
 and prices predicted by the estimated model, given 
 the actual historical promotion activities, w^ith the cor- 
 responding values predicted by the model following 
 a counterfactual one percent increase in promotional 
 expenditure in every month in the sample (Septem- 
 ber 1992 to July 1996). We also simulate changes in 
 the assessments jointly with the corresponding 
 changes in promotion. We use the differences between 
 these actual and counterfactual scenarios to calculate 
 measures of the marginal gross and net benefits. 
 
 Since changes in more than one month are to be 
 simulated, it is necessary to be able to aggregate ben- 
 efits and costs over time. A natural impulse may be 
 simply to add them up. This is appropriate only if 
 past benefits or costs could not have been invested to 
 earn some interest. If the relevant interest rate is not 
 zero, past benefits and costs should be compounded 
 forward to the present. We computed present values 
 of benefits and costs using an annual interest (com- 
 pounding) rate of three percent (a reasonable value 
 for the long-term, risk free, real rate of interest). 
 
 Table 4.1. Approximation of p, the Benefit from a Marginal Increase in Promotion Expenditure 
 
 Models 
 
 Mean 
 
 Minimum 
 
 Maximum 
 
 
 Based 
 
 on OLS Models 
 
 
 1 
 
 3.29 
 
 2.51 
 
 4.62 
 
 2 
 
 2.90 
 
 2.23 
 
 4.06 
 
 3 
 
 2.42 
 
 1.86 
 
 3.15 
 
 4 
 
 6.45 
 
 4.96 
 
 8.39 
 
 
 Based on 2SLS Models 
 
 
 1 
 
 36.44 
 
 20.41 
 
 72.81 
 
 2 
 
 1.42 
 
 1.09 
 
 1.98 
 
 3 
 
 10.55 
 
 8.11 
 
 13.71 
 
 4 
 
 4.82 
 
 3.70 
 
 6.26 
 
 Notes: The regression 
 
 estimates for the models above are given in 
 
 tables 3.4 and 3.5. 
 
 14. It is different from the Nerlove-Waugh condition for optimal advertising financed in a lump-sum fashion (but eqmvalent 
 if the producers' share of the lump sum is equivalent to their share of a check-off). 
 
 Nerlove and Waugh (1961). Like Dorfman and Steiner 
 (1954), Nerlove and Waugh (1961) modeled a case 
 where advertising is funded in a lump-sum way, un- 
 related to output, with the implication that all of the 
 advertising cost is borne by producers. That approach 
 has been adopted in many subsequent studies of pri- 
 mary product promotion. Alston, Carman, and 
 Chalfant (1994) extended the Nerlove-Waugh model 
 to the situation where advertising is funded by a per 
 unit check-off. The condition for optimal advertising 
 that they derived is the same as the Dorfman-Steiner 
 condition for optimal advertising by a monopolist with 
 fixed output.''' 
 
 The same logic leads to the result that p, the ben- 
 efit from a marginal increase in promotion expendi- 
 ture (for an increase in promotion financed by a check- 
 off), can be approximated using 
 
 da da 
 
 ap 
 
 a V _ a 
 Vl\ a |77|i 
 
 30 
 
 where i = a/vis the intensity of promotion (or adver- 
 tising). This result is intuitively clear, in the case where 
 supply is fixed and the producer benefits from adver- 
 tising are exactly equal to the price increase induced 
 by the demand shift, multiplied by the quantity sup- 
 plied. It also applies more generally when advertis- 
 ing is financed by a check-off. 
 
 Using the results from the preferred monthly re- 
 gression models, we computed values for p at every 
 monthly data point. These results are given in table 
 
 36 
 
The Supply Model 
 
 To conduct the benefit-cost analysis, the preferred 
 demand model is combined with an assumed supply 
 function. First, from the demand side, the predicted 
 quantities were calculated by substituting the actual 
 values for each of the explanatory variables into the 
 estimated equation. For instance, for model 3.4b (col- 
 umns 2 in tables 3.4 and 3.5), 
 
 Q +b RPP + b RPS + b REXP + 
 
 / 0 PP f PR ( EXP / 
 
 i) IRPROCPB +RPROSUN +b SPR + 
 
 PRO\ I I SPR t 
 
 b WIN +b SUM +b TIME 
 
 WIN I SUM t T t 
 
 (4.1) 
 
 where the hats ('^) on parameters denote their esti- 
 mated values, and the hat on denotes the predicted 
 value given the estimated coefficients and the values 
 of the exogenous variables. We replicated this step 
 for each of the four models, estimated by both OLS 
 and 2SLS. 
 
 Next, the supply function was defined to be of the 
 constant elasticity form and to pass through the points 
 defined by the predicted quantities from the demand 
 model. That is, the supply function is of the form 
 
 Q =AR' where A =Q Ir' , (4.2) 
 
 and R^ is the producer return per pound in year I, de- 
 fined as R^ = (l-Xj)Pj, where T, is the actual promotional 
 expenditure per pound consumed in year t, expressed 
 as a fraction of the market price in year f (i.e., the rate 
 of assessment required to finance the actual promo- 
 tional expenditure). is a parameter that varies from 
 month to month to ensure that, given the actual val- 
 ues of prices and the other exogenous variables, each 
 month the supply equation passes through the point 
 defined by the predicted quantity from the demand 
 model and the actual price. This means that we can 
 combine the calibrated supply model and the esti- 
 mated demand model to simulate the past actual prices 
 and predicted quantities. 
 
 Supply functions were calibrated using alternative 
 supply elasticities (£) of 0, 0.5, 1.0, 2.0, and 5.0. This 
 range of elasticities reflects a range of lengths of run; 
 it also reflects our uncertainty about the exact impli- 
 cations of international trade in prunes for the elastic- 
 ity of the residual supply to the domestic market. 
 
 Changes in producer surplus were calculated by 
 integrating the function over the range of a price 
 change. In practice, this translates to using the fol- 
 lowing formula for the change in producer surplus: 
 
 R'Q'-RQ 
 
 APS '—^ (4.3) 
 
 ' 1 + e 
 
 Simulations 
 
 Using these definitions of supply and demand 
 equations, we first replicated the past: by equating 
 the equations for supply and demand and solving for 
 market equilibrium, we obtained values of actual 
 prices and predicted quantities (from the demand 
 model), given the actual values for the exogenous vari- 
 ables. In addition, we simulated counterfactual sce- 
 narios, by using hypothetical values for the exogenous 
 variables. 
 
 Counterfactual simulations were conducted by: 
 
 • using hypothetical values for the CPB's promo- 
 tional expenditure in every year (1 .01 times the actual 
 values) with actual assessment rates (in practice we 
 define "actual" assessment rates by expressing total 
 promotional expenditure as a fraction of the total value 
 of production) 
 
 • using hypothetical values for the assessment rate 
 in every year (1.01 times the actual values) with actual 
 promotional expenditure 
 
 • changing both the promotional expenditures and 
 the assessments (setting both at 1 .01 times the actual 
 values). 
 
 For each simulation, we calculated two measures 
 of producer marginal costs of promotion: (a) the cost 
 of the marginal promotional expenditure, and (b) the 
 producer surplus loss associated with an assessment 
 sufficient to generate the same amount of additional 
 expenditure. Only selected summary results are re- 
 ported below. 
 
 An important issue is to know how much confi- 
 dence can be placed in the particular values of the ben- 
 efit-cost measures. How confident can we be that the 
 net benefits are greater than, say, $1 million, given a 
 "best" estimate of, say, $2 million? The precision of 
 our estimates of the benefits depends on the precision 
 of our estimates of the underlying parameters, but in 
 ways that are not easy to see clearly. To evaluate the 
 precision of our measures of benefits and costs, we con- 
 ducted Monte Carlo simulations, following the ap- 
 proach laid out in Alston et al. (1997). These simula- 
 tions yield confidence intervals on our welfare mea- 
 sures, permitting us to make statements such as that a 
 95% confidence interval for the benefit-cost ratio is 
 formed by the interval from a:l to b:l, where a is a 
 lower confidence limit and b is an upper confidence 
 limit. 
 
 In practice, to do this requires an iterative process 
 where first a particular set of values of the parameters 
 is drawn at random from the estimated joint statisti- 
 cal distribution. This set of parameters is substituted 
 into the demand equation and used to generate pre- 
 dicted quantities which are, in turn, used to param- 
 eterize the supply equation for each supply elasticity. 
 Then these supply equations are used with the demand 
 
 37 
 
Table 4.2. Marginal Benefit-Cost Ratios for Prune Promotion: A Comparison of Estimates from Four 
 Regressions using both OLS and 2SLS 
 
 Supply Elasticity 
 
 Series 0.0 0.5 1.0 2.0 5.0 
 
 Benefit-Cost Ratios from OLS Models 
 
 Model 1 
 
 Producer Benefits /Producer Costs 
 
 3.00 
 
 2.99 
 
 2.99 
 
 2.99 
 
 2.99 
 
 Producer Benefits /Total Expenses 
 
 3.00 
 
 1.32 
 
 0.85 
 
 0.50 
 
 0.22 
 
 Model 2 
 
 
 
 
 
 
 Producer Benefits /Producer Costs 
 
 2.65 
 
 2.65 
 
 2.65 
 
 2.65 
 
 2.65 
 
 Producer Benefits /Total Expenses 
 
 2.65 
 
 1.20 
 
 0.78 
 
 0.46 
 
 0.20 
 
 Model 3 
 
 
 
 
 
 
 Producer Benefits /Producer Costs 
 
 3.82 
 
 3.85 
 
 3.86 
 
 3.86 
 
 3.87 
 
 Producer Benefits/Total Expenses 
 
 3.82 
 
 1.48 
 
 0.92 
 
 0.52 
 
 0.23 
 
 Model 4 
 
 
 
 
 
 
 Producer Benefits /Producer Costs 
 
 10.19 
 
 10.29 
 
 10.30 
 
 10.31 
 
 10.32 
 
 Producer Benefits /Total Expenses 
 
 10.19 
 
 3.02 
 
 1.78 
 
 0.97 
 
 0.41 
 
 
 Benefit-Cost Ratios from 2SLS Models 
 
 Model 1 
 
 
 
 
 
 
 Producer Benefits /Producer Costs 
 
 29.32 
 
 29.61 
 
 29.63 
 
 29.64 
 
 29.65 
 
 Producer Benefits /Total Expenses 
 
 29.32 
 
 5.49 
 
 3.03 
 
 1.60 
 
 0.66 
 
 Model 2 
 
 
 
 
 
 
 Producer Benefits /Producer Costs 
 
 0.86 
 
 0.86 
 
 0.86 
 
 0.86 
 
 0.86 
 
 Producer Benefits/Total Expenses 
 
 0.86 
 
 0.61 
 
 0.48 
 
 0.33 
 
 0.17 
 
 Model 3 
 
 
 
 
 
 
 Producer Benefits/Producer Costs 
 
 16.80 
 
 17.02 
 
 17.05 
 
 17.07 
 
 17.08 
 
 Producer Benefits/Total Expenses 
 
 16.80 
 
 4.56 
 
 2.64 
 
 1.43 
 
 0.61 
 
 Model 4 
 
 
 
 
 
 
 Producer Benefits /Producer Costs 
 
 5.82 
 
 5.86 
 
 5.88 
 
 5.89 
 
 5.91 
 
 Producer Benefits /Total Expenses 
 
 5.82 
 
 3.43 
 
 2.44 
 
 1.54 
 
 0.74 
 
 Notes: Present Values are in millions of constant (August 1996) dollars using 3 percent (annual) compounding. 
 
 38 
 
equation to conduct the counterfactual scenarios, and 
 then to evaluate the scenarios. The detailed results 
 from the Monte Carlo study are reported in appendix 
 tables A4.1-A4.3. 
 
 Benefit-Cost Ratios 
 
 Table 4.2 reports marginal benefits and costs of 
 prune promotion implied by each of the four demand 
 models estimated by both OLS and 2SLS, using five 
 alternative values for the supply elasticity. A real dis- 
 count rate of 3 percent per annum was used to com- 
 pound the costs and benefits. In each case, we report 
 both producer benefits relative to producer costs asso- 
 ciated with a change in assessment to finance a change 
 in promotion, along with the producer benefits rela- 
 tive to the total costs of the change in promotional ex- 
 penditure (not just the producer share). For example, 
 the first row in table 4.2 refers to model 1 estimated by 
 OLS and the first column represents a supply elastic- 
 ity of zero. Thus, using a supply elasticity of zero and 
 a demand model based on OLS estimation of model 1, 
 the benefit-cost ratio for a one percent increase in pro- 
 motional expenditure is 3.00, regardless of whether we 
 distinguish between producer cost and total cost (they 
 are the same when the supply elasticity is zero). Look- 
 ing across the columns in this row, we can see the ef- 
 fects of the increase in the supply elasticity. We can 
 see that the supply elasticity does not affect the ben- 
 efit-cost ratio when we consider only the producer 
 
 share of the cost of the checkoff — the benefit-cost ra- 
 tio is essentially 3.00 regardless of the value of the sup- 
 ply elasticity. However, as the supply elasticity rises, 
 producers receive progressively smaller benefits from 
 a given demand increase. Hence, the benefit-cost ra- 
 tio computed using total promotional expenditure 
 declines with increases in the supply elasticity, going 
 from 3.00 when supply is fixed, to 0.22 when the sup- 
 ply elasticity is 5.0. This pattern is repeated across all 
 the models (and, looking across the rows and com- 
 paring within a column, we can see the effects of the 
 different demand models). Thus, a primary issue is 
 whether incremental promotion is financed by incre- 
 mental changes in assessments or in a lump-sum fash- 
 ion (see Goddard et al. (1994), Alston, Carman, and 
 Chalfant (1994), and Alston et al. (1997)). In what fol- 
 lows, we emphasize the results obtained by compar- 
 ing producer benefits to producer costs, rather than to 
 total expenses, where total expenses are partly borne 
 by producers and partly by consumers, to the extent 
 that costs of the assessment are passed on through 
 higher consumer prices. 
 
 Table 4.3 reports the point estimates for the ratio of 
 producer benefits to producer costs from table 4.2, cal- 
 culated using a supply elasticity of 1 .0 (but represen- 
 tative of all supply elasticities), with the correspond- 
 ing results from the Monte Carlo simulation. We re- 
 port the mean and the upper and lower limits from a 
 95 percent confidence interval. Recall that, in the OLS 
 
 Table 4.3. Marginal Benefit-Cost Ratio for a Supply Elasticity of One 
 
 Simulation Results 
 
 OLS Model Point Mean 95% Lower 95% Upper 
 
 Estimate Bound Bound 
 
 1* 3.0 10.2 0.3 41.6 
 
 2* 2.7 9.0 0.6 33.2 
 
 3» 3.9 18.4 0.4 49.5 
 
 4 10.3 25.2 1.1 91.1 
 
 Simulation Results 
 
 2SLS Model 
 
 Point 
 
 Mean 
 
 95% Lower 
 
 95% Upper 
 
 
 Estimate 
 
 
 Bound 
 
 Bound 
 
 1* 
 
 29.6 
 
 14.4 
 
 0.4 
 
 71.1 
 
 2 
 
 0.9 
 
 2.0 
 
 0.3 
 
 7.4 
 
 3* 
 
 17.1 
 
 20.7 
 
 0.2 
 
 66.8 
 
 4 
 
 5.9 
 
 5.9 
 
 1.1 
 
 118.2 
 
 Notes: Models marked * were not rejected on grounds of implausible estimates or statistical problems. 
 
 39 
 
models, model 4, which did not include Sunsweet pro- 
 motion, was rejected, but models 1, 2, and 3 were not. 
 Comparing the point estimates for these three mod- 
 els, the estimated benefit-cost ratios fall within a rela- 
 tively narrow range from 2.65 to 3.85. An analyst who 
 had inadvertently omitted Sunsweet promotion from 
 the analysis (model 4) would have estimated a ben- 
 efit-cost ratio of 10.3 — a clear example of omitted- 
 variable bias leading to a biased benefit-cost ratio. 
 
 While the OLS benefit-cost estimates for models 1, 
 2, and 3 range from 2.7 to 3.9, the point estimates for 
 the 2SLS models 1 and 3 are much greater (recall that 
 the 2SLS models 2 and 4 were rejected in section 3.2). 
 The point estimate for model 1 using 2SLS is 29.6, and 
 that for model 3 is 17.1. These larger benefit-cost ra- 
 tios result from both a larger promotion elasticity for 
 CPB promotion and a smaller own-price elasticity of 
 demand in the 2SLS models than in the OLS models.'^ 
 
 15. The effect of the own-price elasticity is apparent when comparing OLS and 2SLS for model 2. Although the effect of 
 CPB promotion is slightly greater in the 2SLS case, the benefit-cost ratio is considerably smaller than in the OLS 
 model. The reason is that the larger promotion elasticity under 2SLS is offset by the very large (in absolute value) 
 own-price elasticity of demand. 
 
 40 
 
5. ANALYSIS OF TEST-MARKET STUDIES 
 
 The California Prune Board has periodically con- 
 ducted test-market studies to evaluate the effective- 
 ness of its promotional expenditures. As part of our 
 overall evaluation of the Board's promotional activi- 
 ties, we performed our own analysis of data collected 
 in two recent test-market studies. One study, con- 
 ducted in 1990, involved an analysis of television ad- 
 vertising in three test cities, with three comparable cit- 
 ies monitored as controls. The other study, conducted 
 in conjunction with Safeway stores in 1995, was an 
 evaluation of the effect of in-store information on 
 prune sales. These studies provide an opportunity to 
 further our understanding of the effects of alternative 
 types of promotion and provide a basis to confirm or 
 contradict results from the analysis of the monthly and 
 annual demand models. Following is our analysis of 
 these two test-market studies. 
 
 5.1 Six-City Study of Television Advertising 
 
 The objective of this study was to estimate the ef- 
 fect of television advertisements on prune sales. The 
 initial study was conducted by Nielson Marketing 
 Research, based on the method of matched-market 
 evaluation. Three cities were selected as test markets, 
 where TV ads for prunes were run for 12 weeks in 1990, 
 and three cities were selected as control markets in 
 which no TV ads for prunes were run. The test and 
 the control cities were selected based on similarity of 
 volume of prune sales prior to the test, promotion his- 
 tory, and availability of major brands (Table 5.1). 
 
 Table 5.1. Test and Control Markets in the Television 
 Ad Test Campaign 
 
 Test Cities 
 
 Matching Control Cities 
 
 TV ads run 
 
 No TV ads 
 
 Denver 
 
 Omaha 
 
 Hartford 
 
 Philadelphia 
 
 Kansas City 
 
 Chicago 
 
 Source: Nielson Marketing Research. 
 
 The TV ads featured generic advertising of dried 
 prunes. In 1990, an advertising agency was commis- 
 sioned by the Prune Board to create a new advertising 
 campaign to promote prunes. The result was a series 
 of television commercials with the overall title "Prune 
 Presenter." In these television commercials, a magi- 
 cian performed sleight-of-hand tricks to portray 
 prunes as containing more vitamins and minerals than 
 other fruits. Expenditures on the TV ads in the three 
 test cities over the twelve weeks were at a rate equiva- 
 lent to expenditures of $4.2 million annually in the 
 United States. 
 
 The TV ads were run for 12 weeks in 1990, from 
 early September to the end of November. The differ- 
 ences in prune sales between the test and control mar- 
 kets were based on sales from September 1990 to Feb- 
 ruary 1991; the 12 weeks when the TV ads were run 
 and an additional 12 weeks after the completion of the 
 TV ads (Table 5.2). 
 
 Table 5.2. Time Frame of the Television Advertising Test 
 
 Same month; previous year 
 < > 
 
 (9/89) 
 
 (2/90) 
 
 Earlier months; same year 
 
 (3/90) 
 
 Testing period 
 
 (8/90) (9/90) 
 
 (2/91) 
 
 TV ads run Post TV Ad 
 
 (9/90) 
 (11/90) 
 
 (12/90) 
 (2/91) 
 
 41 
 
Table 5.3. Overall Effect of Television Advertisements on Prune Sales 
 
 Market 
 
 Percentage Change 
 
 Adjustment 
 
 Ad market 
 
 +3.4 
 
 Unadjusted for pre-test differences 
 
 No-ad market 
 
 -8.3 
 
 Unadjusted for pre-test differences 
 
 Net difference 
 
 +9.1 
 
 Adjusted for pre-test differences 
 
 Total U.S. 
 
 -1.9 
 
 Unadjusted for pre-test differences 
 
 Source: Nielson Marketing Researcii. 
 
 
 Table 5.4. Effect of Television Advertisements on Prune Sales 
 
 in Individual Markets 
 
 Test Cities 
 
 Matching Control Cities 
 
 Net difference in prune sales 
 
 TV Ads run 
 
 No TV ads 
 
 - percentage - 
 
 Denver 
 
 Omaha 
 
 16.0 
 
 Hartford 
 
 Philadelphia 
 
 21.3 
 
 Kansas City 
 
 Chicago 
 
 3.4 
 
 Average 
 
 Average 
 
 9.1 
 
 Source: A.C. Nielson 1991. 
 
 Nielson Marketing Research conducted an initial 
 analysis of the test market results. Their method in- 
 volved first adjusting sales for the periods during and 
 after the TV ads for any pre-advertising differences 
 between markets, based on sales trends in the months 
 prior to the TV ads, March to August 1990. Second, 
 the percentage change in sales was computed for each 
 city, as the ratio of adjusted sales from September 1990 
 to February 1991 to sales in the same 24 weeks of the 
 previous year (September 1989 to February 1990). 
 Third, net differences in the percentage change in sales 
 were computed by comparing each test city and its 
 control city. Nielson found that prune sales in the test 
 markets were up a net 9.1 percent, relative to the test 
 markets, when compared with the previous year (table 
 5.3), a result Nielson claimed was statistically signifi- 
 cant at the 1 percent level. 
 
 The campaign's effectiveness apparently varied 
 across the test cities (table 5.4). In Hartford and Den- 
 ver sales were up 21.3 percent and 16.0 percent, re- 
 spectively, relative to the control, but the sales differ- 
 ence in Kansas City was only 3.4% relative to its con- 
 trol, Chicago, and was not statistically significant. 
 
 Figure 5.1 summarizes the effects of the ad cam- 
 paign on sales of dried prunes. Sales increased rela- 
 
 tive to control in the test markets in September, the 
 first month of the campaign, but the difference then 
 diminished in October, and adjusted sales in the test 
 markets were actually less than in the control markets 
 in November. However, sales in the test markets then 
 increased significantly relative to control in the post- 
 test months of December-February. 
 
 The results reported in tables 5.3 and 5.4 and fig- 
 ure 5.1 suggest a positive impact for the TV ad cam- 
 paign, but our ability to draw firm inferences is ham- 
 pered by confounding influences that occurred dur- 
 ing the period of the market test. In particular, the test 
 period coincided with a significant increase in the in- 
 cidence of deals on prunes offered by the retail trade, 
 especially in the control markets."' The percentage of 
 prunes sold with a deal is shown in figure 5.2. The 
 incidence of deals in the control markets was especially 
 high relative to the test markets in October and No- 
 vember, perhaps explaining why the differential ef- 
 fects of the TV ad campaign diminished during those 
 months. 
 
 The analysis-of-variance method employed by 
 Nielson Marketing does not permit these various con- 
 founding effects to be distinguished, but, in principle, 
 it is possible to do so using the econometric approach 
 
 16. These deals or trade promotions included displays, cents-off coupons, and bargains such as 'l3uy one, get a second 
 for 50 percent off." 
 
 42 
 
Figure 5.1. Percentage Change in Prune Sales in Test and Control Markets 
 
 ^ 45 
 
 
 
 
 
 
 
 
 iz 
 
 >, 
 
 
 
 
 
 .2 35 
 > 
 
 ♦ Test (TV Ads) 
 
 ■ Control (No TV Ads) 
 
 ■ 
 
 1 
 
 ♦ 
 
 
 E 
 
 c 
 
 o 
 B 
 
 S 15 
 
 CO 
 > 
 
 bC 5 
 
 c 
 
 re 
 
 ♦ i 
 
 m 
 
 t 
 
 ♦ 
 
 ■ 
 
 < 
 
 > 
 
 ■ 
 
 T 
 
 ■ 
 
 
 rcentage c 
 
 June July August September October November December January FebAary 
 
 i 
 
 
 E. 
 
 -15 
 
 
 
 
 
 
 
 
 Source: Data compiled by Nielson Marketing and provided by California Prune Board. 
 
 Figure 5.2. Share of Prunes Sold with a Deal 
 
 40.00 
 
 35.00 
 
 ■S 30.00 
 rs 
 
 ■S 25.00 
 
 T3 20.00 
 
 ^ 15.00 
 
 05 
 
 10.00 
 5.00 
 0.00 
 
 -5.00 
 -10.00 
 
 ♦ Test (TV Ads) 
 
 
 
 
 
 
 
 
 ■ Control (No TV Ads) 
 
 
 
 
 
 ♦ 
 
 
 
 
 
 
 ■ 
 
 ■ 
 
 
 
 
 
 1 
 
 1 
 
 
 
 
 
 
 
 
 
 
 
 
 ♦ 
 
 T 
 
 ■ 
 
 
 
 i 
 
 * 
 
 ♦ 
 
 
 
 
 
 
 
 m 
 ♦ 
 
 
 1 
 
 
 
 
 
 1 
 
 1 
 
 
 
 
 ♦ 
 
 
 
 
 
 
 June July August 
 
 September 
 
 October 
 
 November 
 
 December 
 
 January 
 
 FebAary 
 
 Source: Data compiled by Nielson IVlarketing and provided by California Prune Board. 
 
 43 
 
adopted in this study. We thus undertook an inde- 
 pendent examination of the test market data. 
 
 Re-evaluation of the TV Advertising Test 
 
 The data for this analysis consist of monthly prune 
 sales for nine months for each of three test-market cit- 
 ies and three control cities, 54 observations in all. The 
 time period consists of three months prior to the TV 
 ads, the three months of the ad campaign, and three 
 months following the end of the ad campaign. 
 
 The statistical model used for this analysis is a lin- 
 ear model of consumer demand, similar to those used 
 earlier in this report. In this model, the volume of 
 prune sales at retail is expressed as a function of the 
 price of prunes and the alternative types of promo- 
 tion that were undertaken during this time in the test 
 and control markets. An algebraic specification of the 
 model is as follows: 
 
 Q - b„ + fc, PRICE , + b, DISPLAY , + b,PRINT , 
 
 ^n.t 0 1 n,l 2 n,t 3 n.t 
 
 + b COUPON , + b,DEAL% , + b TV-PRIOR 
 
 4 n.t 5 II,/ 6 
 
 -H b,TV-DURING + bJV-AFTER + e „ (5.1) 
 
 7 8 ii,r 
 
 where n = 1 6 denotes cities and t - 1 9 denotes 
 monthly time periods. 
 
 The variables used in the model are defined in table 
 5.5, and summary statistics for the variables are re- 
 ported in table 5.6. We have data for the number of 
 displays, print ads, the number of coupon ads, and 
 the percentage of sales tied to a deal for both the test 
 
 and control markets for the entire nine-month period 
 of the analysis, thereby enabling us to separate these 
 effects from any sales impacts due to the TV ads. The 
 presence of the TV ad campaign is measured by 1 0,1) 
 indicator or "dummy" variables. Thus, for example, 
 TV-PRIOR assumes a value of one for a test-market 
 city during the three-month period prior to the ads 
 running, and is zero otherwise. 
 
 Results of the analysis are reported in table 5.7. The 
 first column reports results for the full model as set 
 forth in equation 5.1. Several aspects of this model 
 are worth noting. First, the estimated price elasticity 
 of demand is -0.57, a value quite consistent with esti- 
 mates provided elsewhere in this report. Second, the 
 effects of the TV ad campaign on demand are positive 
 and significant, during both the test period {TV-DUR- 
 ING) and the post-test period (TV-AFTER). Pre-test 
 differences in prune sales between the test and con- 
 trol cities {TV-PRIOR) are small and are not statisti- 
 cally significant. 
 
 DEAL%, the variable used to measure the percent- 
 age of prunes sold on a deal, also has a positive and 
 significant effect on demand. This result is not sur- 
 prising, because the deals reduce the effective price of 
 prunes to buyers. The number of ads run with cou- 
 pons also had a positive impact on sales, but the effect 
 was not quite significant at the five percent level. The 
 remaining two variables, PRINT and DISPLAY, have 
 negative coefficients. These effects run counter to ex- 
 pectations, but they are not statistically significant. 
 
 Table 5.5. Description of Variables in the Television Ad Demand Model 
 
 Variable 
 
 Definition 
 
 
 Pounds of dried prunes purchased per million dollars of retail purchases in city n during 
 
 
 month t 
 
 PRICE , 
 
 Retail price of prunes in city n during month t ($/lb) 
 
 DISPLAY , 
 
 Number of displays in city n during month t 
 
 PRINT , 
 
 Number of A, B, or C ads in city n during month t 
 
 COUPON , 
 
 Number of ads with coupons in city n during month t 
 
 DEALX , 
 
 Percentage of sales tied with a deal, such as two-for-one 
 
 TV-PRIOR 
 
 Dummy variable equal to one for the test city n during the 3 months before TV ads were 
 
 
 aired 
 
 TV-DURING 
 
 Dummy variable equal to one for the test city n during the 3 months TV ads were aired 
 
 TV-AFTER 
 
 Dummy variable equal to one for the test city n during the 3 months after TV ads were 
 
 
 aired 
 
 44 
 
Table 5.6. Summary Statistics of the Variables in the Television Ad Demand Model 
 
 Variable 
 
 M 
 i\ 
 
 Mean 
 
 DlallClarQ 
 
 Deviation 
 
 iVllIllIIlLlIIl 
 
 \/f -3 Vl TVll 1 m 
 
 iVldAlll IU.111 
 
 Q„.> 
 
 □4 
 
 Z14.0 
 
 D4.0 
 
 LJJ.Kj 
 
 JO 1 .U 
 
 FKlLt , 
 
 n.l 
 
 54 
 
 1 Q1 
 
 i.yi 
 
 n 1 Q 
 u. ly 
 
 1 A 
 
 i.D 
 
 A.JO 
 
 U lb FLAY , 
 
 n,t 
 
 54 
 
 lo.U 
 
 11 n 
 ll.U 
 
 u.u 
 
 4: 1 
 
 PRINT 
 
 >i,f 
 
 54 
 
 21.1 
 
 22.7 
 
 0.0 
 
 82.0 
 
 COUPON , 
 
 54 
 
 3.1 
 
 8.7 
 
 0.0 
 
 38 
 
 
 54 
 
 10.2 
 
 10.1 
 
 0.0 
 
 40.0 
 
 TV-PRIOR 
 
 54 
 
 0.17 
 
 0.38 
 
 0.0 
 
 1.0 
 
 TV-DURING 
 
 54 
 
 0.17 
 
 0.38 
 
 0.0 
 
 1.0 
 
 TV-AFTER 
 
 54 
 
 0.17 
 
 0.38 
 
 0.0 
 
 1.0 
 
 suggesting that these mechanisms had little indepen- 
 dent effect on prune sales. 
 
 To test for the sensitivity of these results to alterna- 
 tive specifications, we re-estimated the model exclud- 
 ing individually or in combination the insignificant 
 variables, DISPLAY, PRINT, and COUPON. The re- 
 sults from estimating these models are provided in 
 columns (2) through (4) in table 5.7. When PRINT is 
 excluded, the effect of coupon ads is significant at the 
 five percent level. Importantly, the effects of the vari- 
 ables of most interest, PRICE and the TV dummy vari- 
 ables, are little affected by the inclusion or exclusion 
 of the other variables, thus reinforcing our confidence 
 about those estimates. 
 
 The estimated effect of the TV ads on prune sales 
 during and after the TV ads was positive and statisti- 
 cally significant at the one percent level. The TV ads 
 increased prune sales more after the ads concluded 
 than during the period when the ads were run. Both 
 deals and coupons, which reduce the effective price 
 paid by consumers, also increased prune sales. Thus, 
 our analysis of the matched-cities test-market data sug- 
 gests strongly that the television ad campaign was suc- 
 cessful in increasing prune sales both during the test 
 and after it. That is, even after accounting for effects 
 of other types of promotion on prune sales, the TV 
 ads were effective in increasing sales of prunes.'" 
 
 Burke Marketing Research conducted two tele- 
 phone surveys to measure awareness and attitudes to- 
 
 ward prunes. A pre- wave survey was conducted in 
 early September, prior to the TV ads, and a post-wave 
 survey was conducted in mid-December, after the ad- 
 vertising campaign was concluded. Consumers in the 
 test markets exhibited greater awareness of prune ad- 
 vertising after the campaign was completed (Keeble 
 1992), explaining perhaps the relatively greater impact 
 of the ads on sales after the campaign had ended. 
 
 5.2 In-Store Promotion 
 
 The objective of this market study was to estimate 
 the effect of interactive kiosks or display terminals on 
 prune sales. Store shoppers could push a button la- 
 beled "prunes" on an interactive kiosk and follow di- 
 rections to obtain recipes and to see prune advertise- 
 ments. The interactive kiosks were referred to as 
 Safeway New PICS, because they were tested in 
 Safeway stores located in or near Phoenix. 
 
 The research program consisted of 10 test stores 
 matched with 10 control stores. The stores were 
 matched based on similarity of total store volume, 
 shopper demographics, and shares of urban and sub- 
 urban shoppers. Retail sales were measured in each 
 store before and during the introduction of the PICS. 
 
 The California Prune Board was one of 34 partici- 
 pants in the Safeway New PICS test. The promotional 
 activities were conducted during three periods in 1995, 
 
 17. It is important from a statistical perspective that we consider the effects on prune demand of DISPLAY, PRINT, 
 COUPON, and DEAL%, but we caution against attempts to provide much interpretation of the numerical results. 
 Although these variables all relate to various forms of prune promotion, they were not the focus of the test market 
 study. Although we do have quantitative information on these variables from the data set generated by Nielson 
 Marketing Research, we do not have any details on the types of displays, print ads, coupons, or deals that were in 
 effect. Attempts to obtain this information from Nielson were unsuccessful. 
 
 18. Unfortunately the data available from the test market do not enable us to conduct a cost-benefit evaluation of the TV 
 ad campaign. 
 
 45 
 
Table 5.7. Econometric Results: Estimated Effects of Television Ads on Prune Sales 
 
 Independent Variables 
 
 
 U; 
 
 (3) 
 
 (4) 
 
 Lonstant 
 
 297.36 
 
 298.10 
 
 321.81 
 
 322.86 
 
 
 {D.U/) 
 
 
 CC 
 
 (5.97) 
 
 PKlLt , 
 
 -64.01 
 
 -65.71 
 
 -78.77 
 
 -80.85 
 
 
 
 
 
 (-Z.64) 
 
 
 [-0.57] 
 
 [-0.58] 
 
 [-0.70] 
 
 [-0 721 
 
 DKPr AY 
 
 n.t 
 
 
 -U.Dt) 
 
 
 
 
 (-1.04) 
 
 (-1.23) 
 
 
 
 Pl^TAJT 
 
 n "^i 
 -yj.ol 
 
 
 -U.JO 
 
 
 
 (-1.38) 
 
 
 (-1.52) 
 
 
 cm TPflKl 
 
 n,f 
 
 n 71 
 
 U./ i 
 
 U.oZ 
 
 U.oz 
 
 n 7/1 
 
 
 (1.94) 
 
 (2.19) 
 
 (1.64) 
 
 (1 87) 
 
 TIF AT 
 
 n,\ 
 
 J.V/ 
 
 9 7n 
 z./u 
 
 9 79 
 Z./Z 
 
 
 
 (5.83) 
 
 (5.94) 
 
 (5.58) 
 
 (5 78) 
 
 
 10 7 
 lU./ 
 
 1 "5 ^8 
 1 J. JO 
 
 11 1 Q 
 
 1 "3 Q1 
 
 ij.yi 
 
 
 (0.76) 
 
 (0.96) 
 
 (0.82) 
 
 (1.03) 
 
 TV-DuRING 
 
 44.8 
 
 47.89 
 
 40.06 
 
 41.98 
 
 
 (2.90) 
 
 (3.13) 
 
 (2.96) 
 
 (3 11) 
 
 TV-AFTER 
 
 83.2 
 
 83.08 
 
 83.34 
 
 82.22 
 
 
 ^1 ^ 
 
 (3. JO) 
 
 (3.DD) 
 
 (D.DD) 
 
 Buse K- 
 
 0.73 
 
 0.73 
 
 0.72 
 
 0.72 
 
 Buse Raw- 
 
 
 
 
 
 Moment 
 
 0.98 
 
 0.98 
 
 0.98 
 
 0.98 
 
 Notes: 54 observations, t-values are in parentheses and estimated values of the price elasticity of demand calculated at 
 sample means are in brackets. The dependent variable is ^, pounds of dried prunes purchased per million dollars of retail 
 purchases in city n in month t. 
 
 46 
 
Table 5.8. Test Design of New PICS at Safewaxj Stores 
 
 Presence of 
 
 Period 
 
 Weeks 
 
 Dates 
 
 Type of Promotion 
 
 Safeway New PICS 
 
 Base 
 
 1-8 
 
 June 4 - July 29 
 
 
 No 
 
 1 
 
 1-4 
 
 Aug. 6 - Sept. 2 
 
 Advertising 
 
 Yes 
 
 2 
 
 5-8 
 
 Sept. 3 - Sept. 30 
 
 Advertising and Recipes 
 
 Yes 
 
 3 
 
 9-12 
 
 Oct. 1 - Oct. 28 
 
 Advertising and Recipes 
 
 Yes 
 
 as shown in table 5.8. The ads in period 1 were differ- 
 ent from the ads in periods 2 and 3. 
 
 During the three test periods, the Safeway New 
 PICS were placed in the test stores, but not in the con- 
 trol stores. The effect of the Safeway New PICS on 
 sales was measured in terms of the net percentage 
 changes in sales between the test and control stores. 
 Over the 12- week test period, retail sales of prunes 
 increased in both the test and the control stores when 
 compared with the base period. The percentage dif- 
 ference in sales of pitted prunes in PIC stores, relative 
 to control stores, was 19.9 percent in the first period of 
 the program, grew to 29.2 percent in period 2 and then 
 dropped to 21 .6 percent in the third period. The aver- 
 age percentage difference over all three periods was 
 23.6 percent. The larger net percentage increases in 
 sales in periods 2 and 3 may reflect either more effec- 
 tive ads or the release of recipes, or a combination of 
 the two factors. 
 
 The number of button pushes for prunes on the 
 New PICS declined from 2,840 in period 1, to 1,080 in 
 period 2, and 800 in period 3. Store shoppers explored 
 the New PICS more when they were first introduced. 
 The high use of the PICS machines in period 1 may 
 have contributed to higher sales in periods 2 and 3. 
 
 5.3 Conclusions from the Test-Market Analysis 
 
 The primary importance of the test-market studies 
 is that they confirm broadly the key results contained 
 in our primary analyses. Both the TV ad campaign 
 and the New PICS sales promotion increased prune 
 sales significantly relative to control markets. Both test- 
 market studies suggest that the effects of prune pro- 
 motion are rather durable. The TV ad test market 
 study also helped to increase our confidence that the 
 price elasticity of demand for prunes is inelastic, with 
 estimated values in the range of -0.57 to -0.72. 
 
 47 
 
6. CONCLUSION 
 
 This study has analyzed the effectiveness of prod- 
 uct promotion in the Cahfornia prune industry. The 
 economic theory of consumer demand was used to 
 specify empirical models to explain prune consump- 
 tion as a function of prune prices, expenditures on 
 prune promotion, and other relevant variables. Three 
 complementary analyses were conducted. The main 
 analysis used data for monthly intervals from Septem- 
 ber 1992 to July 1996. A secondary data set consisted 
 of annual observations from 1949 to 1995. A third 
 analysis evaluated the results of a test-market study 
 of prune advertising in six U.S. cities. 
 
 The two main sources of expenditures on promo- 
 tion of California prunes are the California Prune 
 Board (CPB) and Sunsweet Growers. Results from 
 analysis of the monthly, retail data support strongly 
 the proposition that prune advertising and promotion 
 has been an effective mechanism for increasing the 
 demand for prunes. Across alternative model specifi- 
 cations examined and reported in part 3, prune pro- 
 motion had a consistently statistically significant, posi- 
 tive effect on per capita domestic prune consumption. 
 For the various models estimated using ordinary least 
 squares (OLS), the elasticity of prune demand with 
 respect to CPB promotion generally ranged from 0.048 
 to 0.052, meaning that a ten percent increase in expen- 
 ditures on generic promotion would have induced 
 about a 0.5 percent increase in consumption, holding 
 price and other explanatory variables constant. Be- 
 cause of concern that some of the explanatory vari- 
 ables might be endogenous, the preferred model was 
 re-estimated using two-stage least squares (2SLS). Pro- 
 motion by the CPB remained a positive and statisti- 
 cally significant determinant of prune sales in the 
 model estimated by 2SLS, with the estimated elastic- 
 ity with respect to promotion being slightly larger than 
 in the OLS model. The models based on the monthly 
 data performed well against diagnostic tests, causing 
 us to have reasonable confidence in the specification 
 and, in turn, the statistical results pertaining to the ef- 
 fects of promotion. 
 
 The models based on the annual data did not per- 
 form as well. Promotion, measured in this case by 
 annual, aggregate real expenditures by the CPB and 
 Sunsweet on all types of domestic promotion, was 
 generally not found to be a statistically significant de- 
 terminant of demand. However, the diagnostic tests 
 generally rejected the hypothesis that the annual mod- 
 els were specified correctly, reflected in dynamic ef- 
 fects on demand that the models were unable to cap- 
 ture adequately. 
 
 We used both an approximation method and a 
 simulation approach to translate the effects of promo- 
 
 tion on demand into estimates of marginal benefits to 
 prune growers. Because of our greater faith in the data 
 underpinning the monthly analysis of demand, the su- 
 perior statistical performance of those demand mod- 
 els, and their congruence with the results from the test- 
 market analysis reported in part 5, we based our ben- 
 efit-cost analysis on results from the monthly models. 
 The simulation analysis required a complete model of 
 the industry, including supply response. Since a sup- 
 ply analysis was not a component of the present study, 
 simulations were conducted for a range of alternative 
 synthetic supply functions. These simulations, but- 
 tressed by some complementary algebraic derivations, 
 enabled us to estimate the marginal returns to the in- 
 dustry from expenditures on advertising and promo- 
 tion. 
 
 As part 4 discusses in detail, the marginal benefit- 
 cost ratio for advertising and promotion can hinge on 
 the value of the price elasticity of supply, depending 
 on how the expenditures are funded. The marginal 
 returns refer to the revenues generated from an incre- 
 mental expenditure on advertising and promotion. 
 The ratio of producer benefits to producer incidence 
 of the check-off, however, does not depend on the sup- 
 ply elasticity. We emphasize this measure. 
 
 Optimal allocation of expenditures to advertising 
 and promotion calls for expanding expenditures until 
 the marginal dollar spent just yields a dollar back in 
 benefits. The simulation analyses suggest that the in- 
 dustry stopped short of this optimizing condition dur- 
 ing the 1992-1996 period covered by the monthly data. 
 Considering just the models that were not rejected, the 
 marginal benefit of an additional dollar of expendi- 
 ture, given the amounts actually expended, ranged 
 upward from $2.65, suggesting that additional expen- 
 ditures on advertising and promotion would have gen- 
 erated positive net revenue to producers. 
 
 Only when producers are (implausibly) assumed 
 to bear the entire cost of the promotion is it possible to 
 derive average benefit-cost ratios less than 1:1, and this 
 is only possible for supply elasticities of 1 .0 or greater. 
 
 We conclude that promotion of California prunes 
 conducted by the CPB has increased the demand for 
 prunes and returns to producers of prunes. Over the 
 four-year period analyzed in the monthly model, the 
 results suggest that investments by prune growers in 
 promotion through the CPB yielded them marginal 
 returns of at least $2.65 for every dollar spent. More- 
 over, marginal benefit-cost ratios in the range of 2.7:1, 
 and higher, indicate that the industry could have prof- 
 itably invested even more in promotion during this 
 period. 
 
 48 
 
APPENDIX TABLES 
 
 Table A2.1. Selected Fruit Juices: U.S. Per Capita Consumption (in gallons) * 
 
 Crop 
 year 
 
 
 Citrus 
 
 
 Total 
 citrus 
 
 
 Non-Citrus 
 
 
 Total 
 noncitrus 
 
 Total 
 
 Orange 
 
 Grapefruit 
 
 Lemon 
 
 Lime 
 
 Apple 
 
 Grape 
 
 Pineapple 
 
 Prune 
 
 1971 
 
 3.81 
 
 0.68 
 
 0.09 
 
 0.01 
 
 4.59 
 
 0.53 
 
 0.21 
 
 0.26 
 
 0.12 
 
 1.13 
 
 5.71 
 
 1972 
 
 4.18 
 
 0.67 
 
 0.10 
 
 0.01 
 
 4.96 
 
 0.58 
 
 0.30 
 
 0.26 
 
 0.11 
 
 1.25 
 
 6.21 
 
 1973 
 
 4.19 
 
 0.71 
 
 0.15 
 
 0.01 
 
 5.07 
 
 0.45 
 
 0.19 
 
 0.25 
 
 0.07 
 
 0.96 
 
 6.03 
 
 1974 
 
 4.32 
 
 0.68 
 
 0.09 
 
 0.01 
 
 5.10 
 
 0.39 
 
 0.24 
 
 0.20 
 
 0.10 
 
 0.93 
 
 6.03 
 
 1975 
 
 4.66 
 
 0.69 
 
 0.24 
 
 0.01 
 
 5.60 
 
 0.49 
 
 0.25 
 
 0.18 
 
 0.08 
 
 1.00 
 
 6.61 
 
 1976 
 
 5.18 
 
 0.56 
 
 0.09 
 
 0.01 
 
 5.84 
 
 0.57 
 
 0.23 
 
 0.21 
 
 0.09 
 
 1.10 
 
 6.93 
 
 1977 
 
 5.01 
 
 0.75 
 
 0.17 
 
 0.01 
 
 5.94 
 
 0.52 
 
 0.22 
 
 0.20 
 
 0.11 
 
 1.06 
 
 6.99 
 
 1978 
 
 4.31 
 
 0.79 
 
 0.18 
 
 0.00 
 
 5.29 
 
 0.66 
 
 0.17 
 
 0.24 
 
 0.09 
 
 1.15 
 
 6.44 
 
 1979 
 
 4.46 
 
 0.76 
 
 0.10 
 
 0.00 
 
 5.32 
 
 0.80 
 
 0.30 
 
 0.24 
 
 0.10 
 
 1.44 
 
 6.77 
 
 1980 
 
 4.95 
 
 0.58 
 
 0.13 
 
 0.01 
 
 5.66 
 
 0.89 
 
 0.23 
 
 0.28 
 
 0.09 
 
 1.49 
 
 7.15 
 
 1981 
 
 4.72 
 
 0.72 
 
 0.25 
 
 0.01 
 
 5.69 
 
 1.08 
 
 0.25 
 
 0.30 
 
 0.09 
 
 1.73 
 
 7.42 
 
 1982 
 
 4.30 
 
 0.69 
 
 0.18 
 
 0.01 
 
 5.18 
 
 0.96 
 
 0.24 
 
 0.28 
 
 0.10 
 
 1.58 
 
 6.75 
 
 1983 
 
 5.78 
 
 0.61 
 
 0.17 
 
 0.01 
 
 6.56 
 
 1.21 
 
 0.24 
 
 0.29 
 
 0.08 
 
 1.82 
 
 8.38 
 
 1984 
 
 4.82 
 
 0.33 
 
 0.12 
 
 0.01 
 
 5.28 
 
 1.32 
 
 0.33 
 
 0.28 
 
 0.06 
 
 1.99 
 
 7.27 
 
 1985 
 
 4.81 
 
 0.61 
 
 0.15 
 
 0.01 
 
 5.57 
 
 1.53 
 
 0.28 
 
 0.27 
 
 0.07 
 
 2.16 
 
 7.72 
 
 1986 
 
 5.16 
 
 0.48 
 
 0.11 
 
 0.01 
 
 5.77 
 
 1 (TO 
 
 1.53 
 
 0.23 
 
 0.34 
 
 0.07 
 
 TIT 
 1.1/ 
 
 /.94 
 
 1987 
 
 5.08 
 
 0.68 
 
 0.21 
 
 0.01 
 
 5.98 
 
 1.52 
 
 0.22 
 
 0.39 
 
 0.07 
 
 2.19 
 
 8.17 
 
 1988 
 
 5.33 
 
 0.37 
 
 0.10 
 
 0.01 
 
 5.80 
 
 1.62 
 
 0.30 
 
 0.42 
 
 0.06 
 
 2.40 
 
 8.21 
 
 1989 
 
 4.63 
 
 0.60 
 
 0.11 
 
 0.01 
 
 5.34 
 
 1.60 
 
 0.26 
 
 0.42 
 
 0.07 
 
 2.35 
 
 7.69 
 
 1990 
 
 3.85 
 
 0.62 
 
 0.14 
 
 0.02 
 
 4.63 
 
 1.45 
 
 0.30 
 
 0.44 
 
 0.04 
 
 2.23 
 
 6.86 
 
 1991 
 
 4.79 
 
 0.41 
 
 0.13 
 
 0.02 
 
 5.36 
 
 1.73 
 
 0.28 
 
 0.49 
 
 0.04 
 
 2.53 
 
 7.89 
 
 1992 
 
 4.33 
 
 0.40 
 
 0.12 
 
 0.01 
 
 4.87 
 
 1.52 
 
 0.35 
 
 0.50 
 
 0.03 
 
 2.40 
 
 7.27 
 
 1993 
 
 5.14 
 
 0.59 
 
 0.17 
 
 0.01 
 
 5.91 
 
 1.57 
 
 0.38 
 
 0.47 
 
 0.04 
 
 2.45 
 
 8.37 
 
 1994P 
 
 5.27 
 
 0.54 
 
 0.18 
 
 0.01 
 
 6.00 
 
 1.79 
 
 0.35 
 
 0.41 
 
 0.04 
 
 2.59 
 
 8.60 
 
 Notes: 1 . Single-strength equivalent. 
 P. Preliminary. 
 
 Source: USDA/Economic Research Service. 
 
 49 
 
Table A2.2. Selected Commercial Fruits and Vegetables (farm weight): U.S. Per Capita Consumption (in pounds) 
 
 Fruit Total fruit"" 
 
 Year Fresh' Processing^ Wine Including Excluding 
 
 Grapes grapes grapes 
 
 1970 
 
 101.2 
 
 128.8 
 
 17.3 
 
 247.2 
 
 230.0 
 
 1971 
 
 100.3 
 
 133.5 
 
 24.4 
 
 258.2 
 
 233.8 
 
 1972 
 
 94.8 
 
 129.3 
 
 17.3 
 
 241.4 
 
 224.1 
 
 1973 
 
 96.5 
 
 131.7 
 
 27.5 
 
 255.6 
 
 228.2 
 
 1974 
 
 95.6 
 
 133.2 
 
 25.5 
 
 254.3 
 
 228.8 
 
 1975 
 
 101.8 
 
 144.5 
 
 23.9 
 
 270.1 
 
 246.2 
 
 1976 
 
 101.5 
 
 149.1 
 
 24.6 
 
 275.2 
 
 250.6 
 
 1977 
 
 99.7 
 
 163.7 
 
 25.7 
 
 289.1 
 
 263.4 
 
 1978 
 
 103.4 
 
 148.0 
 
 29.2 
 
 280.6 
 
 251.4 
 
 1979 
 
 100.1 
 
 145.0 
 
 28.9 
 
 274.1 
 
 245.2 
 
 1980 
 
 104.8 
 
 153.1 
 
 31.5 
 
 289.5 
 
 257.9 
 
 1981 
 
 103.6 
 
 152.6 
 
 27.6 
 
 283.8 
 
 256.2 
 
 1982 
 
 107.4 
 
 147.6 
 
 33.9 
 
 288.8 
 
 255.0 
 
 1983 
 
 110.0 
 
 161.0 
 
 27.3 
 
 298.2 
 
 271.0 
 
 1984 
 
 112.6 
 
 147.4 
 
 30.0 
 
 289.9 
 
 259.9 
 
 1985 
 
 110 6 
 
 1 57 9 
 
 31 3 
 
 994 9 
 
 
 1986 
 
 117.3 
 
 153.5 
 
 29.4 
 
 300.3 
 
 270.9 
 
 1987 
 
 121.6 
 
 155.5 
 
 26.2 
 
 303.2 
 
 277.1 
 
 1988 
 
 120.9 
 
 150.2 
 
 27.6 
 
 298.8 
 
 271.2 
 
 1989 
 
 123.1 
 
 141.2 
 
 25.8 
 
 290.0 
 
 264.2 
 
 1990 
 
 116.5 
 
 144.1 
 
 23.6 
 
 284.3 
 
 260.6 
 
 1991 
 
 113.2 
 
 151.7 
 
 23.0 
 
 287.9 
 
 264.8 
 
 1992 
 
 123.6 
 
 138.8 
 
 27.0 
 
 289.4 
 
 262.4 
 
 1993 
 
 124.9 
 
 153.4 
 
 24.9 
 
 303.3 
 
 278.4 
 
 1994 
 
 126.7 
 
 152.8 
 
 22.5 
 
 302.0 
 
 279.5 
 
 (continued) 
 
 50 
 
Table A2.2 (continued). Selected Commercial Fruits and Vegetables (farm weight): U.S. Per Capita 
 Consumption (in pounds) 
 
 Year 
 
 
 
 Vegetables 
 
 
 
 Total 
 vegetables' 
 
 Total fruit & 
 vegetables' 
 Including Excluding 
 
 
 Fresh* 
 
 Canning' 
 
 Freezing^ 
 
 Dehyd.^ 
 
 Pulses' 
 
 
 
 
 
 
 
 
 grapes 
 
 grapes 
 
 1970 
 
 152.9 
 
 99.4 
 
 45.1 
 
 30.6 
 
 7.6 
 
 335.5 
 
 582.8 
 
 565.5 
 
 1971 
 
 146.7 
 
 106.4 
 
 46.8 
 
 31.0 
 
 7.5 
 
 338.5 
 
 596.6 
 
 572.2 
 
 1972 
 
 150.0 
 
 103.0 
 
 47.0 
 
 30.0 
 
 6.7 
 
 336.7 
 
 578.1 
 
 560.9 
 
 1973 
 
 146.6 
 
 96.7 
 
 51.9 
 
 30.6 
 
 7.9 
 
 333.8 
 
 589.4 
 
 562.0 
 
 1974 
 
 144.6 
 
 98.1 
 
 52.6 
 
 31.7 
 
 6.2 
 
 333.2 
 
 587.6 
 
 562.0 
 
 1975 
 
 147.1 
 
 96.6 
 
 54.0 
 
 32.2 
 
 7.2 
 
 337.1 
 
 607.2 
 
 583.4 
 
 1976 
 
 146.4 
 
 102.2 
 
 58.8 
 
 32.9 
 
 7.0 
 
 347.3 
 
 622.6 
 
 598.0 
 
 1977 
 
 147.0 
 
 100.6 
 
 60.5 
 
 28.9 
 
 6.9 
 
 343.9 
 
 633.0 
 
 607.3 
 
 1978 
 
 141.8 
 
 95.8 
 
 59.9 
 
 30.0 
 
 5.9 
 
 333.3 
 
 613.8 
 
 584.7 
 
 1979 
 
 146.8 
 
 99.5 
 
 56.5 
 
 29.8 
 
 6.8 
 
 339.4 
 
 613.5 
 
 584.5 
 
 1980 
 
 149.2 
 
 101.7 
 
 52.6 
 
 27.1 
 
 5.8 
 
 336.5 
 
 626.0 
 
 594.4 
 
 1981 
 
 142.8 
 
 96.3 
 
 59.1 
 
 28.3 
 
 6.0 
 
 332.5 
 
 616.3 
 
 588.7 
 
 1982 
 
 148.6 
 
 94.7 
 
 54.7 
 
 29.4 
 
 6.9 
 
 334.3 
 
 623.1 
 
 589.2 
 
 1983 
 
 148.5 
 
 96.2 
 
 56.1 
 
 29.5 
 
 7.0 
 
 337.1 
 
 635.4 
 
 608.1 
 
 1984 
 
 154.0 
 
 101.8 
 
 63.6 
 
 29.8 
 
 5.5 
 
 354.7 
 
 644.6 
 
 614.6 
 
 1985 
 
 156.2 
 
 98.9 
 
 65.0 
 
 30.4 
 
 7.6 
 
 358.1 
 
 653.0 
 
 621.7 
 
 1986 
 
 156.3 
 
 99.5 
 
 64.9 
 
 31.0 
 
 7.3 
 
 359.0 
 
 659.3 
 
 629.9 
 
 1987 
 
 162.3 
 
 98.9 
 
 67.2 
 
 29.9 
 
 5.7 
 
 363.9 
 
 667.1 
 
 641.0 
 
 1988 
 
 167.5 
 
 94.6 
 
 64.4 
 
 29.3 
 
 7.5 
 
 363.3 
 
 662.1 
 
 634.5 
 
 1989 
 
 172.3 
 
 102.2 
 
 67.6 
 
 29.9 
 
 6.3 
 
 378.2 
 
 668.2 
 
 642.4 
 
 1990 
 
 166.3 
 
 110.6 
 
 70.6 
 
 31.8 
 
 7.1 
 
 386.4 
 
 670.6 
 
 647.0 
 
 1991 
 
 163.2 
 
 113.1 
 
 73.1 
 
 32.6 
 
 7.9 
 
 389.9 
 
 677.7 
 
 654.7 
 
 1992 
 
 171.3 
 
 110.8 
 
 72.0 
 
 32.1 
 
 8.1 
 
 394.3 
 
 683.7 
 
 656.7 
 
 1993 
 
 172.0 
 
 111.7 
 
 77.5 
 
 33.0 
 
 7.8 
 
 402.0 
 
 705.3 
 
 680.3 
 
 1994 
 
 170.8 
 
 108.0 
 
 79.4 
 
 32.2 
 
 8.0 
 
 398.3 
 
 700.3 
 
 677.8 
 
 Notes: 1. Includes oranges, tangerines, tangelos, lemons, limes, grapefruit, apples, apricots, avocados, bananas, 
 cantaloupes, cherries, cranberries, grapes, honeydew, kiwifruit, mangoes, nectarines, peaches, pears, 
 pineapples, papayas, plums, prunes, strawberries, and watermelon. 
 
 2. Excludes wine grapes. 
 
 3. Computed from unrounded data. 
 
 4. Includes asparagus, broccoli, carrots, cauliflower, celery, sweet com, lettuce, onions, tomatoes, artichokes, 
 garlic, eggplant, cucumbers, bell peppers, cabbage, green beans, mushrooms, potatoes, and sweetpotatoes. 
 
 5. Includes asparagus, snap beans, carrots, sweet corn, pickles, green peas, tomatoes, potatoes, mushrooms, 
 and miscellaneous vegetables. 
 
 6. Includes asparagus, snap beans, broccoli, carrots, cauliflower, sweet com, green peas, potatoes, and 
 miscellaneous vegetables. 
 
 7. Includes potatoes. 
 
 8. Includes dry peas, lentils, and dry edible beans. 
 Source: USDA/ Economic Research Service. 
 
 51 
 
Table A2.3. Fresh Fruits (retail-weight equivalent): U.S. Per Capita Consumption (in pounds)^ 
 
 Year^ 
 
 
 
 Citrus 
 
 
 
 Total 
 Citrus^ 
 
 Non-Citrus 
 
 Oranges & 
 Temples 
 
 Tangerines 
 & Tangelos 
 
 Lemons 
 
 Limes 
 
 Grapefruit 
 
 Apples 
 
 Apricots 
 
 1970 
 
 15.68 
 
 2.13 
 
 1.98 
 
 0.17 
 
 7.97 
 
 27.92 
 
 16.34 
 
 0.11 
 
 1971 
 
 15.26 
 
 2.22 
 
 2.16 
 
 0.16 
 
 8.29 
 
 28.10 
 
 15.77 
 
 0.12 
 
 1972 
 
 14.04 
 
 1.96 
 
 1.79 
 
 0.20 
 
 8.31 
 
 26.30 
 
 14.91 
 
 0.08 
 
 1973 
 
 14.00 
 
 1.97 
 
 1.86 
 
 0.20 
 
 8.31 
 
 26.35 
 
 15.48 
 
 0.08 
 
 1974 
 
 13.99 
 
 2.13 
 
 1.93 
 
 0.19 
 
 7.96 
 
 26.20 
 
 15.75 
 
 0.06 
 
 1975 
 
 15.40 
 
 2.45 
 
 1.87 
 
 0.21 
 
 8.11 
 
 28.04 
 
 18.71 
 
 0.07 
 
 1976 
 
 14.30 
 
 2.25 
 
 1.82 
 
 0.23 
 
 8.98 
 
 27.58 
 
 16.40 
 
 0.09 
 
 1977 
 
 13.04 
 
 2.50 
 
 2.03 
 
 0.22 
 
 7.50 
 
 25.29 
 
 15.86 
 
 0.08 
 
 1978 
 
 13.04 
 
 1.98 
 
 2.05 
 
 0.21 
 
 8.09 
 
 25.38 
 
 17.23 
 
 0.07 
 
 1979 
 
 11.15 
 
 1.92 
 
 1.84 
 
 0.26 
 
 7.07 
 
 22.23 
 
 16.45 
 
 0.07 
 
 1980 
 
 13.85 
 
 2.10 
 
 1.84 
 
 0.34 
 
 7.08 
 
 25.21 
 
 18.43 
 
 0.09 
 
 1981 
 
 11.99 
 
 1.93 
 
 1.93 
 
 0.40 
 
 6.45 
 
 22.70 
 
 16.18 
 
 0.09 
 
 1982 
 
 11.34 
 
 1.97 
 
 1.98 
 
 0.36 
 
 6.99 
 
 22.64 
 
 16.84 
 
 0.07 
 
 1983 
 
 14.59 
 
 2.14 
 
 2.23 
 
 0.49 
 
 7.60 
 
 27.04 
 
 17.54 
 
 0.07 
 
 1984 
 
 11.51 
 
 1.97 
 
 2.07 
 
 0.43 
 
 5.80 
 
 21.78 
 
 17.62 
 
 0.11 
 
 1985 
 
 11.25 
 
 1.44 
 
 2.21 
 
 0.53 
 
 5.34 
 
 20.76 
 
 16.57 
 
 0.15 
 
 1986 
 
 13.03 
 
 1.52 
 
 2.37 
 
 0.55 
 
 5.95 
 
 23.42 
 
 17.13 
 
 0.09 
 
 1987 
 
 12.43 
 
 1.69 
 
 2.38 
 
 0.48 
 
 6.16 
 
 23.14 
 
 20.00 
 
 0.07 
 
 1988 
 
 13.48 
 
 1.68 
 
 2.37 
 
 0.53 
 
 6.49 
 
 24.56 
 
 19.04 
 
 0.14 
 
 1989 
 
 11.80 
 
 1.63 
 
 2.29 
 
 0.66 
 
 6.41 
 
 22.79 
 
 20.37 
 
 0.09 
 
 1990 
 
 12.00 
 
 1.24 
 
 2.50 
 
 0.63 
 
 4.29 
 
 20.66 
 
 18.82 
 
 0.14 
 
 1991 
 
 8.20 
 
 1.31 
 
 2.50 
 
 0.72 
 
 5.69 
 
 18.42 
 
 17.45 
 
 0.12 
 
 1992 
 
 12.52 
 
 1.84 
 
 2.44 
 
 0.98 
 
 5.77 
 
 23.54 
 
 18.48 
 
 0.14 
 
 1993 
 
 13.82 
 
 1.78 
 
 2.55 
 
 0.91 
 
 6.05 
 
 25.11 
 
 18.40 
 
 0.12 
 
 1994'' 
 
 12.67 
 
 2.01 
 
 2.57 
 
 0.93 
 
 5.91 
 
 24.09 
 
 18.77 
 
 0.16 
 
 (continued) 
 
 52 
 
Table A2.3 (continued). Fresh Fruits (retail-weight equivalent): U.S. Per Capita Consumption (in pounds)^ 
 
 Non-Citrus, cont. 
 
 Year^ 
 
 Avocados 
 
 Bananas 
 
 Cherries 
 
 Cran- 
 berries 
 
 Grapes 
 
 Kiwi- 
 fruit 
 
 Mangoes 
 
 Peaches & 
 nectarines 
 
 1970 
 
 0.80 
 
 17.38 
 
 0.47 
 
 0.17 
 
 2.63 
 
 na 
 
 0.07 
 
 5.53 
 
 1971 
 
 0.42 
 
 18.06 
 
 0.63 
 
 0.19 
 
 2.31 
 
 na 
 
 0.08 
 
 5.38 
 
 1972 
 
 0.77 
 
 17.92 
 
 0.36 
 
 0.15 
 
 2.29 
 
 na 
 
 0.08 
 
 3.69 
 
 1973 
 
 0.40 
 
 18.16 
 
 0.69 
 
 0.18 
 
 2.62 
 
 na 
 
 0.11 
 
 4.05 
 
 1974 
 
 0.66 
 
 18.49 
 
 0.54 
 
 0.14 
 
 2.85 
 
 na 
 
 0.12 
 
 4.12 
 
 1975 
 
 1.16 
 
 17.64 
 
 0.65 
 
 0.14 
 
 3.29 
 
 na 
 
 0.16 
 
 4.73 
 
 1976 
 
 0.68 
 
 19.25 
 
 0.77 
 
 0.18 
 
 3.22 
 
 na 
 
 0.18 
 
 4.88 
 
 1977 
 
 1.11 
 
 19.21 
 
 0.59 
 
 0.18 
 
 3.22 
 
 na 
 
 0.14 
 
 4.84 
 
 1978 
 
 1.05 
 
 20.19 
 
 0.50 
 
 0.17 
 
 2.81 
 
 na 
 
 0.20 
 
 5.79 
 
 1979 
 
 1.18 
 
 20.98 
 
 0.63 
 
 0.13 
 
 3.14 
 
 na 
 
 0.20 
 
 6.33 
 
 1980 
 
 0.79 
 
 20.77 
 
 0.64 
 
 0.14 
 
 3.61 
 
 na 
 
 0.24 
 
 6.73 
 
 1981 
 
 1.96 
 
 21.48 
 
 0.50 
 
 0.20 
 
 3.69 
 
 na 
 
 0.19 
 
 6.53 
 
 1982 
 
 1.46 
 
 22.54 
 
 0.49 
 
 0.20 
 
 5.20 
 
 0.07 
 
 0.28 
 
 5.08 
 
 1983 
 
 1.74 
 
 21.25 
 
 0.68 
 
 0.13 
 
 5.09 
 
 0.09 
 
 0.41 
 
 5.16 
 
 1984 
 
 2.06 
 
 22.18 
 
 0.66 
 
 0.12 
 
 5.54 
 
 0.14 
 
 0.41 
 
 6.36 
 
 1985 
 
 1.72 
 
 23.48 
 
 0.40 
 
 0.13 
 
 6.23 
 
 0.13 
 
 0.41 
 
 5.22 
 
 1986 
 
 1.42 
 
 25.82 
 
 0.46 
 
 0.14 
 
 6.46 
 
 0.13 
 
 0.46 
 
 5.54 
 
 1987 
 
 2.21 
 
 25.02 
 
 0.67 
 
 0.12 
 
 6.41 
 
 0.23 
 
 0.53 
 
 5.75 
 
 1 OQQ 
 
 1.49 
 
 24.29 
 
 0.50 
 
 0.11 
 
 7.01 
 
 0.23 
 
 0.36 
 
 6.41 
 
 1989 
 
 1.45 
 
 24.71 
 
 0.59 
 
 0.19 
 
 7.22 
 
 0.30 
 
 0.48 
 
 5.56 
 
 1990 
 
 1.01 
 
 24.36 
 
 0.37 
 
 0.23 
 
 7.21 
 
 0.45 
 
 0.51 
 
 5.27 
 
 1991 
 
 1.32 
 
 25.13 
 
 0.38 
 
 0.25 
 
 6.61 
 
 0.40 
 
 0.81 
 
 6.11 
 
 1992 
 
 1.35 
 
 27.26 
 
 0.50 
 
 0.23 
 
 6.54 
 
 0.30 
 
 0.64 
 
 5.72 
 
 1993 
 
 2.04 
 
 26.80 
 
 0.41 
 
 0.18 
 
 6.41 
 
 0.49 
 
 0.85 
 
 5.65 
 
 1994P 
 
 1.24 
 
 28.06 
 
 0.50 
 
 0.30 
 
 6.67 
 
 0.46 
 
 0.93 
 
 5.19 
 
 (continued) 
 
 53 
 
Table A2.3 (continued). Fresh Fruits (retail-weight equivalent): U.S. Per Capita Consumption (in pounds)^ 
 
 
 
 
 
 Non-Citrus, 
 
 cont. 
 
 
 Total 
 
 Total 
 
 Year^ 
 
 Pears 
 
 Pine- 
 
 Papayas 
 
 Plums & 
 
 Straw- 
 
 Melons 
 
 Non-Citrus^ 
 
 Fresh 
 
 
 
 apples 
 
 
 prunes 
 
 berries 
 
 
 Fruit' 
 
 
 1970 
 
 1.80 
 
 0.67 
 
 0.11 
 
 1.40 
 
 1.60 
 
 19.50 
 
 68.57 
 
 96.49 
 
 1971 
 
 2.41 
 
 0.61 
 
 0.09 
 
 1.22 
 
 1.68 
 
 18.90 
 
 67.86 
 
 95.96 
 
 1972 
 
 2.17 
 
 0.74 
 
 0.11 
 
 1.03 
 
 1.53 
 
 18.50 
 
 64.33 
 
 90.63 
 
 1973 
 
 2.44 
 
 0.87 
 
 0.13 
 
 1.09 
 
 1.45 
 
 18.10 
 
 65.87 
 
 92.22 
 
 1974 
 
 2.36 
 
 0.86 
 
 0.15 
 
 1.43 
 
 1.68 
 
 16.00 
 
 65.22 
 
 91.42 
 
 1975 
 
 2.60 
 
 0.98 
 
 0.16 
 
 1.26 
 
 1.65 
 
 16.10 
 
 69.30 
 
 97.34 
 
 1976 
 
 2.68 
 
 1.09 
 
 0.19 
 
 1.19 
 
 1.52 
 
 17.20 
 
 69.53 
 
 97.11 
 
 1977 
 
 2.26 
 
 1.29 
 
 0.24 
 
 1.47 
 
 1.76 
 
 17.70 
 
 69.94 
 
 95.23 
 
 1978 
 
 2.18 
 
 1.37 
 
 0.24 
 
 1.46 
 
 1.95 
 
 18.20 
 
 73.42 
 
 98.79 
 
 1979 
 
 2.18 
 
 1.39 
 
 0.16 
 
 1.54 
 
 1.75 
 
 17.40 
 
 73.55 
 
 95.78 
 
 1980 
 
 2.48 
 
 1.42 
 
 0.20 
 
 1.46 
 
 1.81 
 
 16.30 
 
 75.12 
 
 100.33 
 
 1981 
 
 2.68 
 
 1.48 
 
 0.21 
 
 1.62 
 
 2.00 
 
 17.50 
 
 76.30 
 
 99.00 
 
 1982 
 
 2.70 
 
 1.58 
 
 0.16 
 
 1.01 
 
 2.18 
 
 20.00 
 
 79.85 
 
 102.50 
 
 1983 
 
 2.84 
 
 1.60 
 
 0.17 
 
 1.34 
 
 2.14 
 
 17.80 
 
 78.05 
 
 105.09 
 
 1984 
 
 2.41 
 
 1.43 
 
 0.25 
 
 1.75 
 
 2.73 
 
 21.80 
 
 85.58 
 
 107.36 
 
 1985 
 
 2.65 
 
 1.40 
 
 0.17 
 
 1.36 
 
 2.75 
 
 21.90 
 
 84.65 
 
 105.41 
 
 1986 
 
 2.83 
 
 1.64 
 
 0.17 
 
 1.23 
 
 2.66 
 
 22.40 
 
 88.58 
 
 112.00 
 
 1987 
 
 3.34 
 
 1.55 
 
 0.18 
 
 1.82 
 
 2.87 
 
 22.10 
 
 92.85 
 
 115.99 
 
 1988 
 
 3.06 
 
 1.67 
 
 0.15 
 
 1.63 
 
 3.07 
 
 21.60 
 
 90.73 
 
 115.28 
 
 1989 
 
 3.04 
 
 1.86 
 
 0.13 
 
 1.34 
 
 2.99 
 
 24.10 
 
 94.44 
 
 117.23 
 
 1990 
 
 3.06 
 
 1.95 
 
 0.17 
 
 1.47 
 
 2.98 
 
 22.40 
 
 90.39 
 
 111.05 
 
 1991 
 
 3.00 
 
 1.82 
 
 0.16 
 
 1.35 
 
 3.29 
 
 21.20 
 
 89.40 
 
 107.82 
 
 1992 
 
 2.98 
 
 1.90 
 
 0.23 
 
 1.69 
 
 3.32 
 
 23.00 
 
 94.29 
 
 117.83 
 
 1993 
 
 3.21 
 
 1.95 
 
 0.27 
 
 1.22 
 
 3.35 
 
 22.80 
 
 94.15 
 
 119.26 
 
 1994 P 
 
 3.30 
 
 1.94 
 
 0.29 
 
 1.53 
 
 3.68 
 
 23.70 
 
 96.70 
 
 120.79 
 
 Notes: 1 . Uses U.S. total population, July 1 for everything except apples, grapes, and pears, which use January 1 of 
 the year following that indicated. 
 
 2. Citrus fruits are on a crop-year basis beginning in year preceding that indicated. Noncitrus fruits are on a 
 calendar-year basis except apples, grapes, and pears which are on a crop year-basis beginning in year 
 indicated. 
 
 3. Computed from unrounded data, 
 na. Not available. 
 
 P. Preliminary. 
 
 Source; USDA/ Economic Research Service. 
 
 54 
 
Table A2.4. Dried Fruits: U.S. Per Capita Consumption (in poundsy 
 
 Year^ 
 
 Apples 
 
 Apricots 
 
 Dates^* 
 
 Figs 
 
 Peaches 
 
 Pears 
 
 Prunes'* 
 
 Raisins 
 
 Total' 
 
 1970 
 
 0.11 
 
 0.06 
 
 0.26 
 
 0.22 
 
 0.02 
 
 0.01 
 
 0.69 
 
 1.35 
 
 2.72 
 
 1971 
 
 0.06 
 
 0.04 
 
 0.26 
 
 0.20 
 
 0.02 
 
 0.01 
 
 0.58 
 
 1.43 
 
 2.60 
 
 1972 
 
 0.08 
 
 0.04 
 
 0.25 
 
 0.13 
 
 0.02 
 
 0.01 
 
 0.49 
 
 1.04 
 
 2.06 
 
 1973 
 
 0.14 
 
 0.05 
 
 0.33 
 
 0.18 
 
 0.01 
 
 0.01 
 
 0.55 
 
 1.38 
 
 2.65 
 
 1974 
 
 0.11 
 
 0.03 
 
 0.26 
 
 0.16 
 
 0.01 
 
 0.01 
 
 0.51 
 
 1.29 
 
 2.38 
 
 1975 
 
 0.13 
 
 0.05 
 
 0.34 
 
 0.16 
 
 0.02 
 
 0.01 
 
 0.60 
 
 1.29 
 
 2.60 
 
 1976 
 
 0.13 
 
 0.06 
 
 0.33 
 
 0.17 
 
 0.02 
 
 0.01 
 
 0.53 
 
 1.28 
 
 2.53 
 
 1977 
 
 0.12 
 
 0.06 
 
 0.36 
 
 0.16 
 
 0.02 
 
 0.01 
 
 0.49 
 
 1.25 
 
 2.47 
 
 1978 
 
 0.12 
 
 0.04 
 
 0.34 
 
 0.17 
 
 0.01 
 
 0.01 
 
 0.43 
 
 1.10 
 
 2.22 
 
 1979 
 
 0.14 
 
 0.06 
 
 0.26 
 
 0.17 
 
 0.01 
 
 0.01 
 
 0.38 
 
 1.31 
 
 2.34 
 
 1980 
 
 0.10 
 
 0.03 
 
 0.14 
 
 0.13 
 
 0.01 
 
 0.01 
 
 0.43 
 
 1.46 
 
 2.31 
 
 1981 
 
 0.10 
 
 0.05 
 
 0.18 
 
 0.14 
 
 0.02 
 
 0.01 
 
 0.46 
 
 1.54 
 
 2.50 
 
 1982 
 
 0.11 
 
 0.08 
 
 0.26 
 
 0.14 
 
 0.02 
 
 0.01 
 
 0.42 
 
 1.52 
 
 2.56 
 
 1983 
 
 0.15 
 
 0.09 
 
 0.25 
 
 0.14 
 
 0.04 
 
 0.01 
 
 0.46 
 
 1.58 
 
 2.72 
 
 1984 
 
 0.16 
 
 0.09 
 
 0.32 
 
 0.13 
 
 0.04 
 
 0.01 
 
 0.39 
 
 1.90 
 
 3.04 
 
 1985 
 
 0.14 
 
 0.03 
 
 0.24 
 
 0.13 
 
 0.02 
 
 0.01 
 
 0.47 
 
 1.92 
 
 2.96 
 
 1986 
 
 0.10 
 
 0.08 
 
 0.15 
 
 0.14 
 
 0.01 
 
 0.01 
 
 0.44 
 
 1.83 
 
 2.76 
 
 1987 
 
 0.15 
 
 0.05 
 
 0.17 
 
 0.18 
 
 0.02 
 
 0.01 
 
 0.62 
 
 1.88 
 
 3.08 
 
 1988 
 
 0.15 
 
 0.08 
 
 0.23 
 
 0.15 
 
 0.02 
 
 0.01 
 
 0.58 
 
 2.07 
 
 3.29 
 
 1989 
 
 0.14 
 
 0.10 
 
 0.23 
 
 0.16 
 
 0.01 
 
 0.01 
 
 0.63 
 
 1.92 
 
 3.20 
 
 1990 
 
 0.10 
 
 0.07 
 
 0.23 
 
 0.20 
 
 0.01 
 
 0.01 
 
 0.97 
 
 1.80 
 
 3.39 
 
 1991 
 
 0.10 
 
 0.08 
 
 0.22 
 
 0.15 
 
 0.02 
 
 0.01 
 
 0.73 
 
 1.78 
 
 3.09 
 
 1992 
 
 0.15 
 
 0.10 
 
 0.16 
 
 0.16 
 
 0.02 
 
 0.01 
 
 0.58 
 
 1.62 
 
 2.80 
 
 1993 
 
 0.18 
 
 0.09 
 
 0.21 
 
 0.21 
 
 0.01 
 
 0.01 
 
 0.68 
 
 1.86 
 
 3.25 
 
 1994 P 
 
 0.19 
 
 0.14 
 
 0.15 
 
 0.19 
 
 0.01 
 
 0.01 
 
 0.71 
 
 1.72 
 
 3.12 
 
 Notes: 1. Processed weight. 
 
 2. Beginning July 1 for apples, apricots, peaches, and pears; September 1 for dates, and August 
 1 for figs, prunes, and raisins. 
 
 3. Pits-in basis. 
 
 4. Excludes quantities used for juice. 
 
 5. Computed from unrounded numbers. 
 P. Preliminary. 
 
 Source: USDA/Economic Research Service. 
 
 55 
 
Table A2.5. Domestic Shipments of California Prunes 
 
 Year 
 
 Dried 
 Pitted 
 Prunes 
 
 Dried 
 Whole 
 Prunes 
 
 Juice 
 and 
 Concentrate 
 
 Canned 
 
 Baoy 
 Food 
 
 or 
 Puree 
 
 U.S. 
 Govt. 
 
 Other 
 Human 
 
 Total 
 Human 
 
 Stock 
 Food 
 
 1949 
 
 657 
 
 85,615 
 
 36,582 
 
 2,912 
 
 4,288 
 
 17,074 
 
 227 
 
 147,355 
 
 207 
 
 1950 
 
 700 
 
 85,222 
 
 30,728 
 
 4,305 
 
 5,079 
 
 5,987 
 
 295 
 
 132,316 
 
 159 
 
 1951 
 
 670 
 
 86,204 
 
 31,728 
 
 2,441 
 
 3,408 
 
 1,360 
 
 61 
 
 125,872 
 
 0 
 
 1952 
 
 749 
 
 84,077 
 
 36,063 
 
 2,745 
 
 3,950 
 
 2,032 
 
 287 
 
 129,903 
 
 5 
 
 1953 
 
 902 
 
 76,162 
 
 37,769 
 
 2,306 
 
 3,687 
 
 3,565 
 
 304 
 
 124,695 
 
 1,031 
 
 1954 
 
 908 
 
 81,557 
 
 38,801 
 
 3,045 
 
 3,803 
 
 496 
 
 624 
 
 129,234 
 
 53 
 
 1955 
 
 883 
 
 70,593 
 
 40,620 
 
 4,738 
 
 3,435 
 
 na 
 
 832 
 
 121,101 
 
 378 
 
 1956 
 
 746 
 
 73,448 
 
 48,920 
 
 5,747 
 
 3,703 
 
 na 
 
 1,888 
 
 134,452 
 
 1,305 
 
 1957 
 
 833 
 
 79,596 
 
 43,560 
 
 4,426 
 
 2,589 
 
 na 
 
 1,703 
 
 132,707 
 
 1 
 
 1958 
 
 507 
 
 57,474 
 
 34,095 
 
 4,327 
 
 1,967 
 
 na 
 
 1,481 
 
 99,851 
 
 143 
 
 1959 
 
 620 
 
 60,605 
 
 41,476 
 
 4,323 
 
 2,589 
 
 na 
 
 2,005 
 
 111,618 
 
 48 
 
 1960 
 
 795 
 
 56,757 
 
 43,781 
 
 4,415 
 
 2,200 
 
 na 
 
 1,268 
 
 109,216 
 
 47 
 
 1961 
 
 1,831 
 
 56,100 
 
 42,377 
 
 3,724 
 
 2,254 
 
 na 
 
 1,599 
 
 107,885 
 
 547 
 
 1962 
 
 1,998 
 
 58,271 
 
 44,328 
 
 3,592 
 
 2,210 
 
 na 
 
 1,144 
 
 111,543 
 
 265 
 
 1963 
 
 3,206 
 
 54,760 
 
 48,454 
 
 4,295 
 
 2,200 
 
 na 
 
 2,500 
 
 115,415 
 
 576 
 
 1964 
 
 5,521 
 
 57,248 
 
 51,513 
 
 9,760 
 
 2,051 
 
 na 
 
 2,896 
 
 128,989 
 
 758 
 
 1965 
 
 7,606 
 
 53,850 
 
 54,414 
 
 8,981 
 
 2,007 
 
 na 
 
 1,911 
 
 128,769 
 
 1,224 
 
 1966 
 
 8,424 
 
 47,167 
 
 37,629 
 
 3,273 
 
 2,093 
 
 1,299 
 
 888 
 
 100,773 
 
 682 
 
 1967 
 
 12,703 
 
 44,566 
 
 44,925 
 
 3,761 
 
 2,041 
 
 213 
 
 846 
 
 109,055 
 
 1,068 
 
 1968 
 
 13,528 
 
 42,347 
 
 46,482 
 
 3,440 
 
 1,598 
 
 13,931 
 
 965 
 
 122,291 
 
 473 
 
 1969 
 
 14,978 
 
 38,030 
 
 41,415 
 
 3,345 
 
 1,782 
 
 4,076 
 
 132 
 
 103,758 
 
 443 
 
 1970 
 
 15,906 
 
 37,428 
 
 42,924 
 
 2,951 
 
 1,285 
 
 17,569 
 
 133 
 
 118,196 
 
 661 
 
 1971 
 
 19,780 
 
 40,337 
 
 42,121 
 
 2,990 
 
 1,316 
 
 11,258 
 
 53 
 
 117,855 
 
 649 
 
 1972 
 
 11,319 
 
 32,729 
 
 37,049 
 
 2,798 
 
 857 
 
 994 
 
 na 
 
 85,746 
 
 10 
 
 1973 
 
 13,671 
 
 40,137 
 
 47,798 
 
 3,764 
 
 1,304 
 
 2,011 
 
 na 
 
 108,685 
 
 430 
 
 1974 
 
 11,131 
 
 36,578 
 
 38,250 
 
 2,832 
 
 1,096 
 
 0 
 
 na 
 
 89,887 
 
 374 
 
 (continued) 
 
 56 
 
Table A2.5 (continued). Domestic Shipments of California Prunes 
 
 Year 
 
 Dried 
 Pitted 
 Prunes 
 
 Dri pH 
 Whole 
 Prunes 
 
 Tnif^A 
 
 and 
 Concentrate 
 
 Canned 
 
 Baby 
 Food 
 
 or 
 Puree 
 
 U.S. 
 Govt. 
 
 Other 
 Human 
 
 Total 
 Human 
 
 Stock 
 Food 
 
 1975 
 
 13,994 
 
 40,487 
 
 48,157 
 
 2,706 
 
 1,161 
 
 185 
 
 na 
 
 106,690 
 
 1,025 
 
 1976 
 
 14,280 
 
 36,300 
 
 43,345 
 
 2,616 
 
 1,665 
 
 0 
 
 na 
 
 98,206 
 
 706 
 
 1977 
 
 13,626 
 
 38,309 
 
 44,289 
 
 3,385 
 
 1,191 
 
 0 
 
 5 
 
 100,805 
 
 20 
 
 1978 
 
 11,981 
 
 30,994 
 
 38,515 
 
 2,461 
 
 1,142 
 
 0 
 
 0 
 
 85,093 
 
 0 
 
 1979 
 
 11,857 
 
 30,159 
 
 42,171 
 
 2,142 
 
 1,031 
 
 0 
 
 0 
 
 87,360 
 
 0 
 
 1980 
 
 14,913 
 
 31,408 
 
 45,241 
 
 2,068 
 
 1,011 
 
 2,932 
 
 0 
 
 97,573 
 
 166 
 
 1981 
 
 15,942 
 
 31,083 
 
 39,858 
 
 1,909 
 
 789 
 
 3,711 
 
 0 
 
 93,292 
 
 183 
 
 1982 
 
 17,626 
 
 29,035 
 
 38,800 
 
 1,609 
 
 762 
 
 6,506 
 
 0 
 
 94,338 
 
 0 
 
 1983 
 
 20,080 
 
 27,270 
 
 39,715 
 
 1,785 
 
 891 
 
 3,667 
 
 0 
 
 93,408 
 
 0 
 
 1984 
 
 19,601 
 
 24,776 
 
 36,330 
 
 1,802 
 
 744 
 
 1,662 
 
 0 
 
 84,915 
 
 0 
 
 1985 
 
 23,102 
 
 25,324 
 
 35,805 
 
 1,839 
 
 707 
 
 3,536 
 
 0 
 
 90,313 
 
 0 
 
 1986 
 
 28,763 
 
 25,237 
 
 33,786 
 
 1,812 
 
 540 
 
 593 
 
 0 
 
 90,731 
 
 0 
 
 1987 
 
 33,288 
 
 23,443 
 
 40,912 
 
 1,872 
 
 683 
 
 2,828 
 
 0 
 
 103,026 
 
 0 
 
 1988 
 
 37,676 
 
 22,591 
 
 40,419 
 
 1,563 
 
 865 
 
 5,637 
 
 0 
 
 108,751 
 
 0 
 
 1989 
 
 42,234 
 
 23,580 
 
 43,103 
 
 1,668 
 
 2,074 
 
 5,058 
 
 0 
 
 117,717 
 
 0 
 
 1990 
 
 44,166 
 
 20,856 
 
 40,688 
 
 1,395 
 
 1,092 
 
 504 
 
 0 
 
 108,701 
 
 0 
 
 1991 
 
 43,059 
 
 16,707 
 
 38,128 
 
 1,280 
 
 1,168 
 
 3,964 
 
 0 
 
 104,306 
 
 0 
 
 1992 
 
 47,932 
 
 13,137 
 
 36,465 
 
 1,304 
 
 1,377 
 
 403 
 
 na 
 
 100,618 
 
 na 
 
 1993 
 
 42,171 
 
 12,269 
 
 33,207 
 
 1,406 
 
 1,502 
 
 235 
 
 na 
 
 90,790 
 
 na 
 
 1994 
 
 48,281 
 
 8,443 
 
 31,419 
 
 1,216 
 
 1,733 
 
 342 
 
 na 
 
 91,434 
 
 na 
 
 1995 
 
 46,383 
 
 9,457 
 
 35,292 
 
 1,281 
 
 1,755 
 
 288 
 
 na 
 
 94,456 
 
 na 
 
 Source: California Prune Board, Annual Reports. 
 
 57 
 
Table A2.6. Expenditures by the California Prune Board 
 
 Year 
 
 Domestic 
 Merchand. 
 and Pub. Rel. 
 
 Domestic 
 Advertising 
 
 Export 
 Promotion 
 
 Industry 
 Research 
 
 Total 
 
 1949 
 
 15 
 
 0 
 
 0 
 
 0 
 
 15 
 
 1950 
 
 13 
 
 0 
 
 0 
 
 0 
 
 13 
 
 1951 
 
 24 
 
 15 
 
 3 
 
 0 
 
 42 
 
 1952 
 
 66 
 
 215 
 
 9 
 
 0 
 
 290 
 
 1953 
 
 68 
 
 268 
 
 18 
 
 0 
 
 355 
 
 1954 
 
 63 
 
 369 
 
 3 
 
 0 
 
 435 
 
 1955 
 
 72 
 
 331 
 
 1 
 
 0 
 
 405 
 
 1956 
 
 58 
 
 336 
 
 66 
 
 0 
 
 460 
 
 1957 
 
 120 
 
 365 
 
 63 
 
 6 
 
 554 
 
 1958 
 
 102 
 
 249 
 
 39 
 
 8 
 
 397 
 
 1959 
 
 112 
 
 286 
 
 35 
 
 7 
 
 439 
 
 1960 
 
 110 
 
 297 
 
 19 
 
 18 
 
 443 
 
 1961 
 
 115 
 
 319 
 
 52 
 
 15 
 
 501 
 
 1962 
 
 110 
 
 319 
 
 75 
 
 19 
 
 523 
 
 1963 
 
 125 
 
 299 
 
 115 
 
 10 
 
 548 
 
 1964 
 
 121 
 
 365 
 
 114 
 
 35 
 
 635 
 
 1965 
 
 159 
 
 981 
 
 142 
 
 31 
 
 1,313 
 
 1966 
 
 156 
 
 809 
 
 114 
 
 40 
 
 1,119 
 
 1967 
 
 158 
 
 848 
 
 14 
 
 44 
 
 1,065 
 
 1968 
 
 195 
 
 688 
 
 61 
 
 88 
 
 1,032 
 
 1969 
 
 156 
 
 903 
 
 39 
 
 59 
 
 1,156 
 
 1970 
 
 164 
 
 903 
 
 7 
 
 80 
 
 1,154 
 
 1971 
 
 190 
 
 1,049 
 
 7 
 
 80 
 
 1,326 
 
 1972 
 
 170 
 
 777 
 
 7 
 
 67 
 
 1,021 
 
 1973 
 
 219 
 
 2,103 
 
 93 
 
 91 
 
 2,506 
 
 1974 
 
 154 
 
 1,706 
 
 138 
 
 92 
 
 2,090 
 
 (continued) 
 
 58 
 
Table A2.6 (continued). Expenditures by the California Prune Board 
 
 Year 
 
 Domestic 
 Merchand. 
 and Pub. Rel. 
 
 Domestic 
 Advertising 
 
 Export 
 Promotion 
 
 Industry 
 
 Research 
 
 Total 
 
 1975 
 
 40 
 
 0 
 
 43 
 
 50 
 
 133 
 
 1976 
 
 0 
 
 0 
 
 0 
 
 69 
 
 69 
 
 1977 
 
 0 
 
 0 
 
 0 
 
 62 
 
 62 
 
 1978 
 
 53 
 
 0 
 
 0 
 
 67 
 
 120 
 
 1979 
 
 336 
 
 171 
 
 0 
 
 78 
 
 585 
 
 1980 
 
 302 
 
 3,299 
 
 0 
 
 90 
 
 3,691 
 
 1981 
 
 366 
 
 3,939 
 
 0 
 
 90 
 
 4,395 
 
 1982 
 
 462 
 
 3,586 
 
 0 
 
 109 
 
 4,157 
 
 1983 
 
 486 
 
 2,829 
 
 0 
 
 105 
 
 3,420 
 
 1984 
 
 576 
 
 3,771 
 
 0 
 
 88 
 
 4,435 
 
 1985 
 
 842 
 
 3,006 
 
 112 
 
 89 
 
 4,049 
 
 1986 
 
 987 
 
 2,037 
 
 595 
 
 85 
 
 3,704 
 
 1987 
 
 2,108 
 
 3,289 
 
 835 
 
 138 
 
 6,370 
 
 1988 
 
 1,539 
 
 4,697 
 
 1,055 
 
 260 
 
 7,552 
 
 1989 
 
 2,103 
 
 5,042 
 
 992 
 
 236 
 
 8,372 
 
 1990 
 
 2,326 
 
 3,191 
 
 1,050 
 
 253 
 
 6,820 
 
 1991 
 
 2,835 
 
 3,057 
 
 883 
 
 269 
 
 7,044 
 
 1992 
 
 3,020 
 
 2,905 
 
 933 
 
 312 
 
 7,170 
 
 1993 
 
 1,535 
 
 1,300 
 
 925 
 
 300 
 
 4,060 
 
 1994 
 
 2,675 
 
 3,769 
 
 940 
 
 348 
 
 7,732 
 
 1995 
 
 3,075 
 
 3,850 
 
 1,070 
 
 379 
 
 8,374 
 
 Source: California Prune Board. 
 
 59 
 
Table A3.1. U.S. Dried Prune Data Used in the Monthly Models (1992-96) 
 
 
 Week 
 
 Total Q 
 
 POP 
 
 EXP 
 
 PP 
 
 PR 
 
 PF 
 
 
 
 
 obs 
 
 Ending 
 
 (lbs) 
 
 (millions) 
 
 ($) 
 
 ($) 
 
 ($) 
 
 ($) 
 
 ($) 
 
 ($) 
 
 (8-96 =1) 
 
 1 
 
 6-Sep-92 
 
 3,417,100 
 
 255.42 
 
 15,931 
 
 1.94 
 
 2.11 
 
 1.84 
 
 4 71 
 
 
 
 2 
 
 4-Oct-92 
 
 3,744,500 
 
 255.59 
 
 15,976 
 
 1.91 
 
 2.12 
 
 1.84 
 
 4 63 
 
 1 rtd^ 811 
 
 U.070 
 
 3 
 
 l-Nov-92 
 
 4,292,200 
 
 255.77 
 
 16,190 
 
 1.90 
 
 2.15 
 
 1.85 
 
 4 37 
 
 
 n Qoi 
 
 4 
 
 29-NOV-92 
 
 4,331,700 
 
 255.94 
 
 16,213 
 
 1.93 
 
 2.19 
 
 1.76 
 
 3 95 
 
 1 c;qq qcjo 
 
 
 5 
 
 27-Dec-92 
 
 4,327,500 
 
 256.11 
 
 16,202 
 
 1.92 
 
 2.28 
 
 1.74 
 
 4 01 
 
 1 ■^y? 4.^7 
 
 
 6 
 
 24-Jan-93 
 
 4,771,900 
 
 256.28 
 
 16,434 
 
 1.91 
 
 2.18 
 
 1.81 
 
 4 97 
 
 1 'iQfi S^Q 
 
 l,J7U,0^" 
 
 U."U/ 
 
 7 
 
 21-Feb-93 
 
 4,980,800 
 
 256.46 
 
 16,492 
 
 1.91 
 
 2.15 
 
 1.81 
 
 4.44 
 
 1 891 77S 
 
 n Qin 
 
 8 
 
 21-Mar-93 
 
 4,933,600 
 
 256.63 
 
 16,550 
 
 1.93 
 
 2.15 
 
 1.82 
 
 4 50 
 
 Z.,L/Orr,.JU^ 
 
 n Q1 
 
 9 
 
 18-Apr-93 
 
 5,148,700 
 
 256.80 
 
 16,742 
 
 1.88 
 
 2.11 
 
 1.78 
 
 4.27 
 
 1 Q18 (^QO 
 
 n Q1 R 
 
 10 
 
 16-May-93 
 
 4,240,500 
 
 256.98 
 
 16,765 
 
 1.94 
 
 2.18 
 
 1.86 
 
 4 59 
 
 1 ^1^ 149 
 
 n Qi 7 
 
 n 
 
 13-Jun-93 
 
 3,890,300 
 
 258.14 
 
 16,788 
 
 1.95 
 
 2.17 
 
 1.86 
 
 4 65 
 
 884 94^^ 
 
 0 QI S 
 
 U." lO 
 
 12 
 
 ll-Jul-93 
 
 3,412,400 
 
 258.31 
 
 16,765 
 
 1.95 
 
 2.17 
 
 1.87 
 
 4.72 
 
 1 007 RC\f\ 
 
 0 QI 5^ 
 
 13 
 
 8-Aug-93 
 
 3,241,700 
 
 258.49 
 
 16,812 
 
 1.98 
 
 2.18 
 
 1.88 
 
 4.73 
 
 833 311 
 
 0 Q71 
 
 14 
 
 5-Sep-93 
 
 3,215,100 
 
 258.66 
 
 16,847 
 
 2.02 
 
 2.18 
 
 1.89 
 
 4.76 
 
 
 n Q99 
 
 15 
 
 3-Oct-93 
 
 3,456,300 
 
 258.83 
 
 16,916 
 
 2.03 
 
 2.18 
 
 1.89 
 
 4.69 
 
 81 5 936 
 
 
 16 
 
 31-Oct-93 
 
 3,809,800 
 
 259.00 
 
 16,983 
 
 2.04 
 
 2.20 
 
 1.89 
 
 4.33 
 
 1 335 997 
 
 0 Q9^^ 
 
 17 
 
 28-NOV-93 
 
 4,230,500 
 
 259.18 
 
 16,999 
 
 2.06 
 
 2.22 
 
 1.81 
 
 4.03 
 
 1 986 493 
 
 n Q97 
 yj.yz./ 
 
 18 
 
 26-Dec-93 
 
 4,188,100 
 
 259.35 
 
 16,999 
 
 2.05 
 
 2.33 
 
 1.80 
 
 4.00 
 
 1 783 1 83 
 
 0 Q97 
 
 19 
 
 23-Jan-94 
 
 4,097,700 
 
 259.52 
 
 17,155 
 
 2.06 
 
 2.24 
 
 1.87 
 
 4.43 
 
 
 n Q9Q 
 
 20 
 
 20-Feb-94 
 
 4,470,600 
 
 259.70 
 
 17,214 
 
 2.06 
 
 2.21 
 
 1.87 
 
 4 60 
 
 1917 Q'^^i 
 
 
 21 
 
 20-Mar-94 
 
 4,947,700 
 
 259.87 
 
 17,273 
 
 2.04 
 
 2.19 
 
 1.85 
 
 4.62 
 
 1 812 888 
 
 
 22 
 
 17-Apr-94 
 
 4,584,700 
 
 260.04 
 
 17,386 
 
 2.04 
 
 2.20 
 
 1.85 
 
 4.50 
 
 1 559 480 
 
 
 23 
 
 15-May-94 
 
 3,890,700 
 
 260.21 
 
 17,398 
 
 2.12 
 
 2.25 
 
 1.90 
 
 4.55 
 
 1 1 79 631 
 
 yj.yjo 
 
 24 
 
 12-Jun-94 
 
 3,701,600 
 
 260.66 
 
 17,457 
 
 2.13 
 
 2.23 
 
 1.90 
 
 4.68 
 
 1 DDI 
 
 n QJ.1 
 
 25 
 
 lO-JuI-94 
 
 3,394,500 
 
 260.83 
 
 17,567 
 
 2.11 
 
 2.24 
 
 1.92 
 
 4.77 
 
 1,503,771 
 
 0.943 
 
 26 
 
 7-Aug-94 
 
 3,283,000 
 
 261.01 
 
 17,638 
 
 2.10 
 
 2.25 
 
 1.94 
 
 4.64 
 
 1,190,957 
 
 0.947 
 
 27 
 
 4-Sep-94 
 
 3,100,600 
 
 261.18 
 
 17,685 
 
 2.11 
 
 2.22 
 
 1.93 
 
 4.70 
 
 599,865 
 
 0.950 
 
 28 
 
 2-Oct-94 
 
 3,377,500 
 
 261.35 
 
 17,697 
 
 2.11 
 
 2.26 
 
 1.96 
 
 4.62 
 
 1,030,314 
 
 0.950 
 
 29 
 
 30-Oct-94 
 
 3,671,500 
 
 261.52 
 
 17,860 
 
 2.09 
 
 2.25 
 
 1.94 
 
 4.06 
 
 1,976,239 
 
 0.950 
 
 30 
 
 27-NOV-94 
 
 4,062,700 
 
 261.70 
 
 17,892 
 
 2.12 
 
 2.28 
 
 1.88 
 
 3.90 
 
 1,471,390 
 
 0.952 
 
 (continued) 
 
 60 
 
Table A3.1 (continued). U.S. Dried Prune Data Used in the Monthly Models (1992-96) 
 
 
 Week 
 
 Toial Q 
 
 POP 
 
 EXP 
 
 PP 
 
 PR 
 
 Pf 
 
 PS 
 
 VROMO 
 
 CPl 
 
 obs 
 
 Ending 
 
 (lbs) 
 
 (millions) 
 
 ($) 
 
 ($) 
 
 ($) 
 
 ($) 
 
 ($) 
 
 ($) 
 
 (8-96 =1) 
 
 31 
 
 25-Dec-94 
 
 4,042,500 
 
 261.87 
 
 17,892 
 
 2.08 
 
 2.37 
 
 1.85 
 
 3.86 
 
 2,281,549 
 
 0.952 
 
 32 
 
 22-Tan-95 
 
 3 882 100 
 
 262.04 
 
 18,115 
 
 2.10 
 
 2.31 
 
 1.92 
 
 4.17 
 
 1,600,298 
 
 0.955 
 
 33 
 
 19.Feb-95 
 
 4,722,700 
 
 262.22 
 
 18,187 
 
 2.13 
 
 2.30 
 
 1.93 
 
 4.28 
 
 1,480,893 
 
 0.959 
 
 34 
 
 19-Mar-95 
 
 4,508,200 
 
 262.39 
 
 18,248 
 
 2.11 
 
 2.32 
 
 1.95 
 
 4.30 
 
 1,967,286 
 
 0.962 
 
 35 
 
 16-Apr-95 
 
 4,470,800 
 
 262.56 
 
 18,459 
 
 2.10 
 
 2.31 
 
 1.94 
 
 4.13 
 
 1,756,616 
 
 0.966 
 
 36 
 
 14-Mav-95 
 
 3 743 300 
 
 262.73 
 
 18,496 
 
 2.16 
 
 2.34 
 
 1.96 
 
 4.28 
 
 1,349,716 
 
 0.968 
 
 37 
 
 ll-Tun-95 
 
 3 665 900 
 
 263.03 
 
 18,532 
 
 2.16 
 
 2.35 
 
 1.98 
 
 4.31 
 
 1,366,585 
 
 0.969 
 
 38 
 
 9-Jul-95 
 
 3 341 000 
 
 263.20 
 
 18,561 
 
 2.13 
 
 2.33 
 
 1.98 
 
 4.33 
 
 1,266,180 
 
 0.969 
 
 39 
 
 6-Aue-95 
 
 3,204,300 
 
 263.38 
 
 18,610 
 
 2.15 
 
 2.33 
 
 1.99 
 
 4.25 
 
 1,022,729 
 
 0.972 
 
 40 
 
 3-Sep-95 
 
 3,044,300 
 
 263.55 
 
 18,647 
 
 2.14 
 
 2.33 
 
 2.01 
 
 4.38 
 
 785,201 
 
 0.974 
 
 41 
 
 l-Oct-95 
 
 3 285 900 
 
 263.72 
 
 18,707 
 
 2.15 
 
 2.35 
 
 2.02 
 
 4.11 
 
 1,182,914 
 
 0.977 
 
 42 
 
 29-Oct-95 
 
 3,783,700 
 
 263.89 
 
 18,742 
 
 2.13 
 
 2.34 
 
 2.00 
 
 4.07 
 
 1,719,983 
 
 0.977 
 
 43 
 
 26-Nov-95 
 
 4,081,500 
 
 264.07 
 
 18,732 
 
 2.16 
 
 2.32 
 
 1.89 
 
 3.89 
 
 1,724,766 
 
 0.976 
 
 44 
 
 24-Dec-95 
 
 4,114,400 
 
 264.24 
 
 18,719 
 
 2.10 
 
 2.40 
 
 1.87 
 
 3.93 
 
 2,216,297 
 
 0.976 
 
 45 
 
 21-Jan-96 
 
 4,052,400 
 
 264.41 
 
 19,079 
 
 2.12 
 
 2.37 
 
 1.95 
 
 4.13 
 
 2,044,989 
 
 0.982 
 
 46 
 
 18-Feb-96 
 
 4,538,000 
 
 264.59 
 
 19,140 
 
 2.16 
 
 2.37 
 
 1.96 
 
 4.24 
 
 2,015,955 
 
 0.985 
 
 47 
 
 17-Mar-96 
 
 4,486,700 
 
 264.76 
 
 19,239 
 
 2.13 
 
 2.33 
 
 1.94 
 
 4.23 
 
 2,290,121 
 
 0.990 
 
 48 
 
 14-Apr-96 
 
 4,389,600 
 
 264.93 
 
 19,313 
 
 2.12 
 
 2.36 
 
 1.95 
 
 4.00 
 
 2,051,264 
 
 0.994 
 
 49 
 
 12-May-96 
 
 3,757,700 
 
 265.10 
 
 19,397 
 
 2.16 
 
 2.42 
 
 1.99 
 
 4.33 
 
 1,193,532 
 
 0.996 
 
 50 
 
 9-Jun-96 
 
 3,392,900 
 
 265.28 
 
 19,409 
 
 2.18 
 
 2.43 
 
 2.02 
 
 4.36 
 
 1,341,806 
 
 0.996 
 
 51 
 
 7-Jul-96 
 
 3,260,300 
 
 265.45 
 
 19,446 
 
 2.14 
 
 2.44 
 
 2.04 
 
 4.30 
 
 858,772 
 
 0.998 
 
 Notes: All prices and expenditure variables are in nominal dollars. For sources, see text. PROMO=PROCPB+PROSUN. 
 
 61 
 
Table A3.2. U.S. Dried Prune Data Used in the Annual Models (1949-95) 
 
 Year 
 
 QS 
 
 AGE65 
 
 PP 
 
 PR 
 
 INC 
 
 POP 
 
 CPI 
 
 
 (1000 lbs) 
 
 (% of POP age 
 65 and older) 
 
 ($/lb) 
 
 ($/lb.) 
 
 ($/ person) 
 
 (millions) 
 
 (1995=1.0) 
 
 1949 
 
 172,000 
 
 0.081 
 
 0.08 
 
 0.07 
 
 1,501 
 
 149.19 
 
 0.156 
 
 1950 
 
 172,000 
 
 0.083 
 
 0.12 
 
 0.13 
 
 1,657 
 
 152.27 
 
 0.158 
 
 1951 
 
 174,000 
 
 0.084 
 
 0.09 
 
 0.08 
 
 1,736 
 
 154.88 
 
 0.171 
 
 1952 
 
 170,000 
 
 0.085 
 
 0.12 
 
 0.08 
 
 1,806 
 
 157.55 
 
 0.174 
 
 1953 
 
 154,000 
 
 0.086 
 
 0.11 
 
 0.08 
 
 1,787 
 
 160.18 
 
 0.175 
 
 1954 
 
 164,000 
 
 0.087 
 
 0.11 
 
 0.09 
 
 1,881 
 
 163.03 
 
 0.177 
 
 1955 
 
 142,000 
 
 0.088 
 
 0.14 
 
 0.09 
 
 1,980 
 
 165.93 
 
 0.176 
 
 1956 
 
 148,000 
 
 0.089 
 
 0.10 
 
 0.10 
 
 2,050 
 
 168.90 
 
 0.178 
 
 1957 
 
 160,000 
 
 0.090 
 
 0.10 
 
 0.13 
 
 2,074 
 
 171.98 
 
 0.184 
 
 1958 
 
 116,000 
 
 0.091 
 
 0.20 
 
 0.16 
 
 2,166 
 
 174.88 
 
 0.190 
 
 1959 
 
 122,000 
 
 0.092 
 
 0.18 
 
 0.10 
 
 2,277 
 
 177.83 
 
 0.191 
 
 1960 
 
 116,000 
 
 0.093 
 
 0.20 
 
 0.11 
 
 2,335 
 
 180.67 
 
 0.194 
 
 1961 
 
 116,000 
 
 0.094 
 
 0.17 
 
 0.10 
 
 2,444 
 
 183.69 
 
 0.196 
 
 1962 
 
 120,000 
 
 0.094 
 
 0.13 
 
 0.12 
 
 2,301 
 
 186.54 
 
 0.180 
 
 1963 
 
 116,000 
 
 0.094 
 
 0.15 
 
 0.10 
 
 2,676 
 
 189.24 
 
 0.201 
 
 1964 
 
 126,000 
 
 0.095 
 
 0.12 
 
 0.12 
 
 2,860 
 
 191.89 
 
 0.203 
 
 1965 
 
 122,000 
 
 0.096 
 
 0.12 
 
 0.10 
 
 3,076 
 
 194.30 
 
 0.207 
 
 1966 
 
 112,000 
 
 0.096 
 
 0.16 
 
 0.10 
 
 3,269 
 
 196.56 
 
 0.213 
 
 1967 
 
 114,000 
 
 0.096 
 
 0.14 
 
 0.15 
 
 3,554 
 
 198.71 
 
 0.219 
 
 1968 
 
 112,000 
 
 0.097 
 
 0.15 
 
 0.13 
 
 3,839 
 
 200.71 
 
 0.228 
 
 1969 
 
 106,000 
 
 0.098 
 
 0.14 
 
 0.13 
 
 4,077 
 
 202.68 
 
 0.241 
 
 1970 
 
 106,000 
 
 0.101 
 
 0.10 
 
 0.14 
 
 4,328 
 
 205.05 
 
 0.255 
 
 1971 
 
 120,000 
 
 0.100 
 
 0.14 
 
 0.16 
 
 4,703 
 
 207.66 
 
 0.266 
 
 1972 
 
 88,000 
 
 0.101 
 
 0.27 
 
 0.28 
 
 5,217 
 
 209.90 
 
 0.274 
 
 1973 
 
 108,000 
 
 0.102 
 
 0.23 
 
 0.38 
 
 5,672 
 
 211.91 
 
 0.291 
 
 1974 
 
 96,000 
 
 0.104 
 
 0.22 
 
 0.30 
 
 6,091 
 
 213.85 
 
 0.323 
 
 1975 
 
 108,000 
 
 0.105 
 
 0.20 
 
 0.33 
 
 6,673 
 
 215.97 
 
 0.353 
 
 1976 
 
 102,000 
 
 0.107 
 
 0.21 
 
 0.35 
 
 7,315 
 
 218.04 
 
 0.373 
 
 1977 
 
 104,000 
 
 0.108 
 
 0.25 
 
 0.42 
 
 8,176 
 
 220.24 
 
 0.398 
 
 1978 
 
 86,000 
 
 0.110 
 
 0.35 
 
 0.53 
 
 9,105 
 
 222.59 
 
 0.428 
 
 1979 
 
 84,000 
 
 0.112 
 
 0.41 
 
 0.58 
 
 10,037 
 
 225.06 
 
 0.476 
 
 1980 
 
 92,000 
 
 0.114 
 
 0.34 
 
 0.60 
 
 11,132 
 
 227.73 
 
 0.541 
 
 1981 
 
 94,000 
 
 0.115 
 
 0.33 
 
 0.66 
 
 11,707 
 
 229.97 
 
 0 596 
 
 1982 
 
 94,000 
 
 0.117 
 
 0.34 
 
 0.57 
 
 12,340 
 
 232.19 
 
 n 633 
 
 1983 
 
 94,000 
 
 0.118 
 
 0.33 
 
 0.29 
 
 13 560 
 
 234.31 
 
 0.654 
 
 1984 
 
 88,000 
 
 0.119 
 
 0.35 
 
 0.32 
 
 14,421 
 
 236.35 
 
 0 689 
 
 1985 
 
 96,000 
 
 0.121 
 
 0.34 
 
 0.31 
 
 15,155 
 
 238.47 
 
 0 706 
 
 1986 
 
 108,000 
 
 0.122 
 
 0.41 
 
 0.38 
 
 15,966 
 
 240.65 
 
 0 719 
 
 1987 
 
 114,000 
 
 0.123 
 
 0.37 
 
 0.41 
 
 17,028 
 
 242.80 
 
 0 745 
 
 1988 
 
 120,000 
 
 0.124 
 
 0.39 
 
 0.45 
 
 18,147 
 
 245.02 
 
 0 776 
 
 \J./ / U 
 
 1989 
 
 132,000 
 
 0.124 
 
 0.39 
 
 0.49 
 
 19,170 
 
 247.34 
 
 0.814 
 
 1990 
 
 130,000 
 
 0.126 
 
 0.44 
 
 0.45 
 
 19,663 
 
 249.91 
 
 0.858 
 
 1991 
 
 120,000 
 
 0.126 
 
 0.47 
 
 0.48 
 
 20,609 
 
 252.65 
 
 0.894 
 
 1992 
 
 122,000 
 
 0.127 
 
 0.51 
 
 0.42 
 
 21,224 
 
 255.42 
 
 0.921 
 
 1993 
 
 108,000 
 
 0.127 
 
 0.56 
 
 0.47 
 
 22,059 
 
 258.14 
 
 0.948 
 
 1994 
 
 114,000 
 
 0.128 
 
 0.55 
 
 0.46 
 
 23,193 
 
 260.66 
 
 0.972 
 
 1995 
 
 112,000 
 
 0.130 
 
 0.52 
 
 0.44 
 
 24,385 
 
 263.03 
 
 1.000 
 
 Notes: All dollar figures are in nominal terms. See text for sources. 
 
 62 
 
Table A4.1. Marginal Benefit-Cost Ratios for Prune Promotion: A Comparison of Means from Simulations 
 Based on Four Regressions Using Both OLS and 2SLS 
 
 Supply Elasticity 
 
 Series 0.0 0.5 1.0 2.0 5.0 
 
 Benefit-Cost Ratios from OLS Models 
 
 Model 1 
 
 Producer Benefits /Producer Costs 
 
 10.19 
 
 10.20 
 
 10.20 
 
 10.19 
 
 10.19 
 
 ProHiirpr Rpnpfit*; /Total Fxnenses 
 
 10.19 
 
 1.24 
 
 0.79 
 
 0.47 
 
 0.21 
 
 iviouei z 
 
 
 
 
 
 
 I roQUcer Denenis/ 1 roaucer v„osis 
 
 8.97 
 
 8.97 
 
 8.97 
 
 8.96 
 
 8.96 
 
 Producer Benefits/Total Expenses 
 
 8.97 
 
 1.12 
 
 0.72 
 
 0.43 
 
 0.20 
 
 iviociei o 
 
 
 
 
 
 
 1 roQUcer Deneriis/ 1 roaucer v^osis 
 
 18.14 
 
 18.35 
 
 18.35 
 
 18.35 
 
 18.36 
 
 Producer Benefits/Total Expenses 
 
 18.14 
 
 1.26 
 
 0.79 
 
 0.46 
 
 0.21 
 
 jviouei 4 
 
 
 
 
 
 
 1 roQucer Dtriicrits/ i rutiuctri ^ubib 
 
 24.86 
 
 25.18 
 
 25.18 
 
 25.19 
 
 25.19 
 
 Producer Benefits /Total Expenses 
 
 24.86 
 
 2.24 
 
 1.41 
 
 0.83 
 
 0.37 
 
 
 Benefit-Cost Ratios from 2SLS Models 
 
 Model 1 
 
 
 
 
 
 
 Producer Benefits /Producer Costs 
 
 14.28 
 
 14.42 
 
 14.42 
 
 14.42 
 
 14.43 
 
 Producer Benefits /Total Expenses 
 
 14.28 
 
 2.44 
 
 1.60 
 
 0.98 
 
 0.46 
 
 Model 2 
 
 
 
 
 
 
 Producer Benefits /Producer Costs 
 
 2.01 
 
 2.01 
 
 2.01 
 
 2.01 
 
 2.01 
 
 Producer Benefits/Total Expenses 
 
 2.01 
 
 0.70 
 
 0.51 
 
 0.34 
 
 0.17 
 
 Model 3 
 
 
 
 
 
 
 Producer Benefits /Producer Costs 
 
 20.39 
 
 20.73 
 
 20.73 
 
 20.74 
 
 20.74 
 
 Producer Benefits /Total Expenses 
 
 20.39 
 
 2.08 
 
 1.36 
 
 0.83 
 
 0.39 
 
 Model 4 
 
 
 
 
 
 
 Producer Benefits /Producer Costs 
 
 53.23 
 
 54.24 
 
 54.21 
 
 54.20 
 
 54.20 
 
 Producer Benefits /Total Expenses 
 
 53.23 
 
 3.30 
 
 2.12 
 
 1.27 
 
 0.59 
 
 Notes: For the OLS models, estimates for model 1 are based on 7,438 replications; those for 2 are based on 7,819 replications; 
 those for 3 are based on 7,084 replications, and those for 4 are based on 6,290 replications. For the 2SLS models, estimates for 
 model 1 are based on 4,120 replications; those for 2 are based on 9,587 replications; those for 3 are based on 4,639 replications, 
 and those for 4 are based on 4,535 replications. For both model types 10,000 simulations were run for each model, and we 
 discarded those cases where a random draw chose a negative price, quantity or promotion value. Present Values are in 
 mUhons of constant (August 1996) dollars using 3 percent (annual) compounding. 
 
 63 
 
Table A4.2. Marginal Benefit-Cost Ratios for Prune Promotion: A Comparison of the Lower 95% Boundaries 
 of Four OLS and 2SLS Regressions 
 
 Supply Elasticity 
 
 Series 0.0 0.5 1.0 2.0 5.0 
 
 Benefit-Cost Ratios from OLS Models 
 
 Model 1 
 
 Producer Benefits /Producer Costs 
 
 0.32 
 
 0.32 
 
 0.32 
 
 0.32 
 
 0.32 
 
 
 
 U.Zl 
 
 U. ID 
 
 U. lU 
 
 U.Uo 
 
 Model 2 
 
 
 
 
 
 
 Producer Benefits /Producer Costs 
 
 0.63 
 
 0.63 
 
 0.63 
 
 0.63 
 
 0.63 
 
 Producer Benefits/Total Expenses 
 
 0.63 
 
 0.46 
 
 0.36 
 
 0.25 
 
 0.13 
 
 Model 3 
 
 
 
 
 
 
 Producer Benefits /Producer Costs 
 
 0.34 
 
 0.34 
 
 0.35 
 
 0.35 
 
 0.35 
 
 Producer Benefits /Total Expenses 
 
 0.34 
 
 0.22 
 
 0.16 
 
 0.10 
 
 0.05 
 
 Model 4 
 
 
 
 
 
 
 Producer Benefits/Producer Costs 
 
 1.08 
 
 1.08 
 
 1.08 
 
 1.08 
 
 1.09 
 
 Producer Benefits/Total Expenses 
 
 1.08 
 
 0.77 
 
 0.60 
 
 0.41 
 
 0.21 
 
 
 Benefit-Cost Ratios from 2SLS Models 
 
 Model 1 
 
 
 
 
 
 
 Producer Benefits/Producer Costs 
 
 0.43 
 
 0.43 
 
 0.43 
 
 0.43 
 
 0.43 
 
 Producer Benefits /Total Expenses 
 
 0.43 
 
 0.33 
 
 0.27 
 
 0.20 
 
 0.11 
 
 Model 2 
 
 
 
 
 
 
 Producer Benefits /Producer Costs 
 
 0.32 
 
 0.32 
 
 0.32 
 
 0.32 
 
 0.32 
 
 Producer Benefits /Total Expenses 
 
 0.32 
 
 0.26 
 
 0.22 
 
 0.17 
 
 0.10 
 
 Model 3 
 
 
 
 
 
 
 Producer Benefits /Producer Costs 
 
 0.19 
 
 0.19 
 
 0.19 
 
 0.19 
 
 0.19 
 
 Producer Benefits /Total Expenses 
 
 0.19 
 
 0.15 
 
 0.12 
 
 0.09 
 
 0.05 
 
 Model 4 
 
 
 
 
 
 
 Producer Benefits /Producer Costs 
 
 1.09 
 
 1.09 
 
 1.10 
 
 1.10 
 
 1.10 
 
 Producer Benefits /Total Expenses 
 
 1.09 
 
 0.86 
 
 0.70 
 
 0.51 
 
 0.28 
 
 Notes: For the OLS models, estimates for model 1 are based on 7,438 replications; those for 2 are based on 7,819 replications; 
 those for 3 are based on 7,084 replications, and those for 4 are based on 6,290 repHcations. For the 2SLS models, estimates for 
 model 1 are based on 4,120 repHcations; those for 2 are based on 9,587 replications; those for 3 are based on 4,639 replications, 
 and those for 4 are based on 4,535 replications. For both model types 10,000 simulations were run for each model, and we 
 discarded those cases where a random draw chose a negative price, quantity or promotion value. Present Values are in 
 millions of constant (August 1996) dollars using 3 percent (annual) compounding. 
 
 64 
 
Table A4.3. Marginal Benefit-Cost Ratios for Prune Promotion: A Comparison of the Upper 95% Boundaries 
 of Four OLS and 2SLS Regressions 
 
 Supply Elasticity 
 
 Series 0.0 0.5 1.0 2.0 5.0 
 
 Benefit-Cost Ratios from OLS Models 
 
 Model 1 
 
 Producer Beriefits/ Producer Costs 
 
 41.63 
 
 41.63 
 
 41.62 
 
 41.62 
 
 41.61 
 
 Producer Benefits /Total Expenses 
 
 41.63 
 
 3.11 
 
 1.72 
 
 0.95 
 
 0.41 
 
 Model 2 
 
 
 
 
 
 
 Producer Benefits /Producer Costs 
 
 33.23 
 
 33.21 
 
 33.20 
 
 33.20 
 
 33.19 
 
 Producer Benefits /Total Expenses 
 
 33.23 
 
 2.21 
 
 1.19 
 
 0.63 
 
 0.27 
 
 Model 3 
 
 
 
 
 
 
 Producer Benefits /Producer Costs 
 
 48.87 
 
 49.45 
 
 49.45 
 
 49.46 
 
 49.46 
 
 Producer Benefits/Total Expenses 
 
 48.87 
 
 3.08 
 
 1.70 
 
 0.91 
 
 0.39 
 
 Model 4 
 
 
 
 
 
 
 Producer Benefits /Producer Costs 
 
 89.87 
 
 91.08 
 
 91.07 
 
 91.07 
 
 91.06 
 
 Producer Benefits/Total Expenses 
 
 89.87 
 
 4.64 
 
 2.49 
 
 1.31 
 
 0.55 
 
 
 Benefit-Cost Ratios from 2SLS Models 
 
 Model 1 
 
 
 
 
 
 
 Producer Benefits /Producer Costs 
 
 70.20 
 
 71.12 
 
 71.12 
 
 71.12 
 
 71.12 
 
 Producer Benefits /Total Expenses 
 
 70.20 
 
 6.38 
 
 3.48 
 
 1.87 
 
 0.79 
 
 Model 2 
 
 
 
 
 
 
 Producer Benefits /Producer Costs 
 
 7.44 
 
 7.44 
 
 7.43 
 
 7.43 
 
 7.43 
 
 Producer Benefits /Total Expenses 
 
 7.44 
 
 1.78 
 
 1.04 
 
 0.58 
 
 0.25 
 
 Model 3 
 
 
 
 
 
 
 Producer Benefits /Producer Costs 
 
 65.73 
 
 66.82 
 
 66.84 
 
 66.85 
 
 66.86 
 
 Producer Benefits /Total Expenses 
 
 65.73 
 
 6.05 
 
 3.32 
 
 1.79 
 
 0.76 
 
 Model 4 
 
 
 
 
 
 
 Producer Benefits /Producer Costs 
 
 116.04 
 
 118.18 
 
 118.17 
 
 118.16 
 
 118.16 
 
 Producer Benefits/Total Expenses 
 
 116.04 
 
 7.41 
 
 4.01 
 
 2.13 
 
 0.90 
 
 Notes: For the OLS models, estimates for model 1 are based on 7,438 replications; those for 2 are based on 7,819 replications; 
 those for 3 are based on 7,084 replications, and those for 4 are based on 6,290 replications. For the 2SLS models, estimates for 
 model 1 are based on 4,120 repUcations; those for 2 are based on 9,587 replications; those for 3 are based on 4,639 replications, 
 and those for 4 are based on 4,535 replications. For both model types 10,000 simulations were run for each model, and we 
 discarded those cases where a random draw chose a negative price, quantity or promotion value. Present Values are in 
 millions of constant (August 1996) dollars using 3 percent (annual) compounding. 
 
 65 
 
REFERENCES 
 
 Alston, J.M. and J.A. Chalfant. "Can We Take the Con out of Meat Demand Studies?" Western Journal of Agricul- 
 tural Economics 16, 1 (1991): 36-48. 
 
 Alston, J.M., H.F. Carman, and J.A. Chalfant. "Evaluating Primary Product Promotion: The Returns to Generic 
 Advertising by a Producer Cooperative in a Small, Open Economy," in E.W. Goddard and D.S. Taylor (eds.) 
 Promotion in the Marketing Mix: Wimt Works and Why? Proceedings from the NEC-63 Spring '94 Conference on 
 Toronto, Ontario, April 28-29, 1994. 
 
 Alston, J.M., H.F. Carman, J. Christian, J.H. Dorfman, J.-R. Murua and R.J. Sexton. Optimal Reserve and Export 
 Policies for the California Almond Industry: Theory, Econometrics and Sijimlations. Giannini Foundation Mono- 
 graph no. 42, February 1995. 
 
 Alston, J.M., J.A. Chalfant, J.E. Christian, E. Meng and N.E. Piggott. The California Table Grape Commission's Pro- 
 motion Program: An Evaluation. Giannini Foundation Monograph no. 43, 1997. 
 
 Blaylock, J.R. and D.M. Smallwood. U.S. Demand for Food: Household Expenditures, Demographics, and Projections. 
 USDA/ERS, Technical Bulletin 1713, 1986. 
 
 Bureau of the Census. Statistical Abstract of the United States. Various editions. 
 
 California Agricultural Statistics Service. 2995 California Prune Acreage Survey, Sacramento, CA, 1996. 
 
 California Crop and Livestock Reporting Service. California Fruit and Nut Crops, 1909-1955. Special Publication 
 261, July 1956. 
 
 California Prune Board and the Prune Marketing Committee, California Prune News: Annual Report, reports from 
 1986 to 1992. 
 
 California Prune Board. California Prune News: Annual Report, January, 1996. 
 
 California Prune Board. Generic Monthly Program Evaluations, budgets covering the fiscal years 92/92 to 95/ 
 96 as provided by CPB. 
 
 California Prune Board. Personal Conversation with Bart Minor, Marketing Director of CPB, October 1997. 
 
 California Prune Board. PruneTec, September, 1996. 
 
 California Prune Board. Tlie California Prune Buyer's Guide, 1996. 
 
 Chalfant, J.A. and J.M. Alston. "Accounting for Changes in Tastes" Journal of Political Economy 96(1988):391-410. 
 
 Dale, L., J. Mamer, G. Goldman. Programs to Increase Grower Income in the California Prune Industry, University of 
 California, Berkeley, 1988. 
 
 Davidson, R., and J.G. MacKinnon. Estimation and Inference in Econometrics, Oxford University Press, New York, 
 1993: 238-239. 
 
 Dorfman, R. and P.O. Steiner. "Optimal Advertising and Optimal Quality." American Economic Review 
 44,l(1954):826-836. 
 
 GiacoUni, E. California Prune Industry Report, International Prune Association Conference, May 24, 1993. 
 
 Goddard, E.W. and A. Tielu. "Assessing the Effectiveness of Fluid Milk Advertising in Ontario." Canadian Jour- 
 nal of Agricultural Economics 36(1 988) :261 -278. 
 
 Goddard, E.W. and M.L. McCutcheon. "Optimal Producer Investment in Generic Advertising: The Case of Fluid 
 Milk in Ontario and Quebec." Canadian journal of Agricultural Economics 41(1993):329-347. 
 
 66 
 
Goddard, E.W., G.R. Griffith and J.J. Quilkey. "Optimal Advertising Investment Decisions." Guelph: Depart- 
 ment of Agricultural Economics and Business, University of Guelph, AEB 92/05, November 1992 (revised 
 April 1994). 
 
 Green, R.D., H.F. Carman and K. McManus. "Some Empirical Methods of Estimating Advertising Effects in 
 Demand Systems: An AppUcation to Dried Fruits." Western Journal of Agricultural Economics 16(July 1991):63- 
 71. 
 
 Information Resources, Inc. (IRI), Infoscan Grocery Markets: Data on Fruit Juices and Dried Fruits from July 1992 to 
 August 1996, as provided by Sun-Diamond Growers of California, Pleasanton, Calif. 
 
 International Monetary Fund, International Financial Statistics, Washington, D.C., monthly reports from July 1992 
 to January 1997. 
 
 Keeble, Don. Burke Market Research's written testimony from 1992 California Prune Board public hearing, as 
 provided by California Prune Board, Pleasanton, Calif. 
 
 Lindauer, K. California Prune Industry Report, International Prune Association Conference, May 24, 1993. 
 
 Maddala, G.S. Introduction to Econometrics, Second Edition. New York: Macmillan Publishing Company, 1992. 
 
 Nerlove, M. and F.V. Waugh. "Advertising without Supply Control: Some Implications of a Study of the Adver- 
 tising of Oranges." Journal of Farm Economics 43(1961 ):813-837. 
 
 Nielsen Marketing Research statistics as provided by the California Prune Board, various years. 
 
 Peterson, R. Californian Marketing & Advertising Report, 1994. 
 
 Piggott, N.E., J.A. Chalfant, J.M. Alston, and G.R. Griffith. "Demand Response to Advertising in the Austrahan 
 Meat Industry." American Journal of Agricultural Economics 78(May 1996): 268-279. 
 
 Prescott, R. Charts on U.S. Food Consumption. USDA/ERS, February 1983. 
 
 Prune Administrative Committee, Marketing Policy Report, various reports covering 1964-1986. 
 
 Prune Administrative Committee, The California Prune Industry Marketing Program, various reports covering 
 1947-1965. 
 
 Ramsey, J.B. "Tests for Specification Error in Classical Linear Least Squares Regression Analysis." Journal of the 
 Royal Statistical Society, Series B 31,1(1969):350-371. 
 
 Sunsweet Growers. Monthly Advertising and Merchandising Expenses, budgets covering fiscal years 92/92 to 
 95/96. 
 
 Thorogood, "The Impact of Price on UK Market Size." Report to California Prune Board, June 1994. 
 
 United States Department of Agriculture (USD A), Agricultural Statistics, Washington, D.C., various years. 
 
 USDA Agricultural Marketing Service. How Dried Prunes are Marketed, Bulletin No. 68, 1979. 
 
 USDA Economic Research Service. Food Consumption, Prices, and Expenditures, 1970-94. Statistical Bulletin No. 
 928, 1996. 
 
 U.S. Department of Commerce, Statistical Abstract of the United States 1995, Washington, D.C., issued September 
 1995. 
 
 White, H. "A Heteroskedasticity-Consistent Covariance Matrix Estimator and a Direct test for Heteroskedasticity." 
 Econometrica 48(1980):817-838. 
 
 White, K.J., S.A. Haun, N.G. Horsman, and S.D. Wong. SHAZAM Econometrics Computer Program: User's Refer- 
 ence Manual, Version 6.2. New York: McGraw-Hill Book Company, 1990. 
 
 67 
 
68 
 
GIANNINI FOUNDATION RESEARCH REPORT SERIES 
 
 The Giannini Foundation Research Report Series (ISSN 0899-3068) is designed to communicate re- 
 search to specific professional audiences interested in applications. The first Research Report was issued 
 in 1961 as No. 246, continuing the numbering of the Giannini Foundation Mimeographed Report Series 
 which the Research Report replaced. 
 
 The Giannini Foundation's goal is to encourage research in the various areas of interest to agricultural 
 economists and to support the dissemination of research findings to other researchers and to the public. 
 Foundation membership includes agricultural economists at the Department of Agricultural and Re- 
 source Economics, Davis and at the Department of Agricultural and Resource Economics, Berkeley. As- 
 sociate members include forestry economists in the College of Natural Resources, Berkeley and econo- 
 mists in Soil and Environmental Sciences at Riverside. 
 
 Copies of this Research Report may be ordered from Agriculture and Natural Resource Publications, 
 6701 San Pablo Avenue, Oakland, CA 94608. Exchanges of similar materials are welcome. Inquiries 
 should be sent to Librarian, Department of Agricultural and Resource Economics, University of Califor- 
 nia, Davis, CA 95616-8512. 
 
 The University of California does not discriminate in any of its policies, procedures or practices. The University is an 
 affirmative action/equal opportunity employer