Japanese Perception of Safety and willingness-to-Pay for Imported Rice Kentaro Yoshida1 and Hikaru H. Peterson2 1. Introduction Japan is the largest net food-importing country in the world. The percentage of caloric intake supplied from domestic agricultural produce is 40 percent and has been constantly decreasing since 1960s. Especially, self-sufficiency rate of grain is 28 percent and is at a very low level. Japan imports, however, only 7 percent of total consumption of rice from foreign countries. Border measures and domestic price supports account for such a low rate. The Uruguay Round agreements of the GATT ensured the minimum access to the Japanese rice market. Since then, the quantity of rice imported from various countries has been steadily increasing. The ongoing round of the WTO agricultural negotiation is scheduled to conclude by the end of 2004. According to the negotiating proposal by the Ministry of Agriculture, Forestry and Fisheries of Japan, food safety is considered to be one of the possible reasons to protest against tariff reduction although it is often accused by exporting countries as a non-trade barrier. In addition, Japan mandated rigid country-of-origin labeling (COOL) for all fresh foods in July 2000. For both exporting countries and Japan, it will be important information to understand Japanese consumers’ perception of COOL and food safety of imported rice. In this study, we analyze the consumer response toward the COOL and food safety of imported and domestic rice by choice modeling which is one of the stated preference techniques. In the stated preference questionnaire, respondents are asked to choose from a set of goods with varying attributes and levels. A nested logit model and a random parameters logit model are used to test data from the questionnaire. Choice modeling in marketing of agricultural products, as exemplified by Sato et al. [7], can simulate an actual market behavior of consumers. A number of literatures have been devoted to understand the relative impacts of factors that affect consumer choice, e.g., assessment of consumer preferences for color of veal (West et al. [9]), food safety attributes in fresh apples (Baker [1]), and use of genetically modified organisms (Burton et al. [2]). 1 2 Kentaro Yoshida is assistant professor in the Institute of Policy and Planning Sciences, University of Tsukuba, Japan. E-mail address: yoshidak@sk.tsukuba.ac.jp. Hikaru H. Peterson is assistant professor in the Department of Agricultural Economics, Kansas State University, USA. In general, it is often suggested that Japanese consumers show the strong resistance to imported rice because of its flavor and safety concerns, e.g., post-harvest pesticide, and other chemical residuals. For these reasons, consumers generally prefer domestic produce over foreign produce. For fresh vegetables and other grains, however, Japanese consumers are likely to accept imported produce because they usually buy imported vegetables and grains at grocery stores mainly during out of season. Our experimental study aims at revealing the difference in willingness-to-pay (WTP) between imported and domestic rice when respondents are exposed to imported rice brands along with domestic brands 2. Survey Design and Data Collection The survey is designed to elicit values for several representative imported and domestic brands of rice. Alternatives in the survey are selected from the following brands. The three imported brands are the Koshihikari variety from the U.S. and Australia, and the Akitakomachi variety from China. The four domestic brands were selected based on telephone interviews to managers of several rice shops: the Koshihikari variety from Niigata prefecture, the Akitakomachi variety from Akita prefecture, and the Hitomebore variety from Miyagi prefecture. In addition, a fourth domestic brand, Ibaraki Koshihikari, was selected as a representative local brand familiar to the survey participants in Tokyo. The valuation section of the survey consists of eight separate choice scenarios. Respondents were asked to choose from three alternatives based on the description of rice brands at different prices (option A, B, and C) or to choose option D. Respondents have the option of indicating that they would choose none of them (option D). The 16 choice sets were created using SPSS version 10.0J. Attributes and levels are given in Table 1. Table 1 Attribute levels used in the choice sets (yen per 5 kilograms) U.S. Koshihikari Australian Koshihikari Chinese Akitakomachi Akita Akitakomachi Miyagi Hitomebore Ibaraki Koshihikari Niigata Koshihikari Level 1 Level 2 Level 3 Level 4 Level 5 280 280 280 1880 1880 1780 2280 480 480 480 2080 2080 1980 2480 680 680 680 2280 2280 2180 2680 880 880 880 2480 2480 2380 2880 1080 1080 1080 2680 2680 2580 3080 The survey was pre-tested twice, and then mailed in March 2002 to 400 households in the Tokyo prefecture, all of whom were randomly chosen from a commercial phone directory database, Kurofune 2002. A reminder was sent two weeks after the initial mailing. 256 households (63.8 percent) returned the questionnaires. Since some of them did not respond to the valuation section completely, there were 1,713 usable observations. Selected demographics of the respondents were as follows: 72.6% were female, average age was 58.9 years old ranging from 25 to 88, and average annual income of households was 6,620 thousand yen. 3. Empirical Models Data obtained by choice modeling studies is usually analyzed by multinomial logit (MNL) models (Louviere et al. [5]). Sato et al. [7] demonstrated the performance of MNL by examining the reproducibility of brand choice of domestic rice in a particular region. However, our preliminary surveys indicated that the utility level of foreign brands was much different from that of domestic ones. In addition, Hausman’s specification test of MNL rejected the independence of irrelevant alternatives (IIA) assumption. We then estimated other models in order to relax the IIA assumption (Greene [3]). These two models were a random parameters logit (RPL) model and a nested logit (NL) model. The RPL model allows the parameters to vary over people and choice situations (Train [8], West et al. [9], Kodama [4]). The indirect utility function is defined as follows. U ij X ij ij X ij (b i ) ij where the utility that respondent i chooses j among J possible alternatives is Uij, and εij is an unobserved random term that is i.i.d. extreme value. X is a vector of choice attributes, and is a vector of marginal utilities parameters that can be expressed as the sum of population mean b and individual deviation i. In the two-level nesting structure, the utility of each consumer for each alternative can be decomposed as Ulj = Ul + Uj|l. Ul is the utility of country-level choice depends on household characteristics, and l = 1 (domestic) and 2 (foreign). Uj|l are the utility of brand choice specified to depend on brand-specific characteristics. Given these assumptions, the probability of choosing the jlth alternative is then decomposed as Plj = Pj|lPl, which can be written as: Plj exp X exp X l l I l exp X j|l exp X l l I l j |l where Il is an inclusive value of lth branch and τl is a parameter of the inclusive value. We defined the following variables for two types of models. The choice set of rice brands consists of eight alternatives: U.S. and Australian Koshihikari (US, AU), Chinese Akitakomachi (CH), Niigata Koshihikari (NG), Akita Akitakomachi (AK), Miyagi Hitomebore (MY), Ibaraki Koshihikari (IB), and choosing none of the alternatives (NONE). The utility of brand choice depends on the price (Price) and the brand, where each brand-specific quality is represented by brand-specific binary variables (US, AU, CH, NG, AK, MY, IB). In addition, images regarding safety and taste affect utility in choosing one of the foreign brands. Specifically, binary variables interacted with brand-specific variables (US-SAFE, AU-SAFE, CH-SAFE) were defined for each foreign brand. Their value is one if respondents had highly negative images of the specific imported brand of rice compared with domestic rice, or they were uncertain of its safety. In a same way, binary variables interacted with brand-specific variables were defined for taste (US-TASTE, AU-TASTE, CH-TASTE). If consumers have selected a brand of their choice over time, their utility may additionally be affected by a habit formation factor. Thus, for domestic rice, the utility level is specified to depend on whether it is the brand currently consumed by the respondent. A binary variable interacted with each domestic brand (HAB-AK, HAB-MY, HAB-IB, HAB-NG) is specified to equal one if the respondent stocked rice from the same region as the specific domestic brand at home. Other variables regarding their socio-economic characteristics are also specified as follows: household income in 10,000 yen (INC), the number of children under 18 (NCHILD), and gender variable (GENDER) which equals one for female and zero for male. In addition, their attitude toward foreign produce (ATTFOR) is specified as Likert-type scale, 1 if their answer to a question regarding their purchases of imported vegetables is at least minding and 5 as only purchasing domestic produce. The indirect utility function of the RPL model is defined as follows: URPL = βPP + βUSUS +βAUAU + βCHCH + βNGNG + βAKAK + βMYMY + βIBIB + βUSSUS-SAFE + βUSTUS-TASTE + βAUSAU-SAFE + βAUTAU-TASTE + βCHSCH-SAFE + βCHTCH-TASTE + βHNGHAB-NG + βHAKHAB-AK + βHMYHAB-MY + βHIBHAB-IB + εRPL The indirect utility functions of the NL model are defined as follows: UUS|foreign = βPP + βUSUS + βUS,1US-SAFE + βUS,2US-TASTE + εUS UAU|foreign = βPP + βAUAU + βAU,1AU-SAFE + βAU,2AU-TASTE + εAU UCH|foreign = βPP + βCHCH + βCH,1CH-SAFE + βCH,2CH-TASTE + εCH UNG|domestic = βPP + βNGNG + βHNGHAB-NG + εNG UAK|domestic = βPP + βAKAK + βHAKHAB-AK + εAK UMY|domestic = βPP + βMYMY + βHMYHAB-MY + εMY UIB|domestic = βPP + βIBIB + βHIBHAB-IB + εIB Ul = αl0 + αl1 ATTFOR + αl2 GENDER +αl3NCHILD + αl4 INC + εl, where εUS, εAU, and εCH are correlated, and εNG, εAK, εMY , and, εIB are correlated, there is no correlation between the two set of error terms, and εl are independent. 4. Results The simulated maximum likelihood result for RPL model and the FIML maximum likelihood result for NL model were computed using LIMDEP, Version 7.0. The estimation results are given in Table 2. The estimated coefficients of the RPL model did not improve the result of the estimated MNL model1) in terms of the goodness-of-fit. In addition, standard deviations of coefficients of foreign brands (US (s.d.), AU (s.d.), CH (s.d.)) were not statistically significant. On the other hand, the NL model could derive more reliable estimates than the MNL model according to adjusted pseudo R2 and log likelihood. The coefficient on the safety perception of U.S. rice and the habit formation factor of Miyagi Hitomebore were not statistically significant at 5% level. Based on the estimated parameters, WTP for each brand is calculated as the negative of the coefficient of the specific brand divided by the coefficient on price. Similarly, expected WTP can be obtained for consumers who have highly negative or uncertain images of foreign varieties. Table 3 reports the WTP estimates based on the estimated results of the NL model, along with the average retail price in 2000 by Food Agency of Japan. WTPs for imported brands can be interpreted as WTPs of those who have neutral or positive images. For all imported brands, the average WTP estimates were lower than the reported retail price, and for all domestic brands, the respondents were willing to pay more than the average retail price. The WTP of Niigata variety was the highest, which was consistent with the market price. Average retail prices of domestic brands were at about 10 percent discount relative to their WTPs. It suggests that consumers are currently benefiting the most from their purchases. Moreover, there was a positive habit formation for domestic brands except Miyagi Hitomebore. Consumers were willing to pay more for rice of the same brand as what they purchased last. Among foreign brands, WTP for Australian rice was the highest, followed by the U.S. and China. WTP for U.S. rice was 65 percent of its average retail price, making it most overpriced relative to its average WTP among the three brands, while Australian rice was most closely priced. If the respondent had a highly negative or uncertain image about safety or taste, WTP signed even negative. It is consistent with Japanese consumers’ behavior at the Heisei rice disorder. When it happened in 1993, it was observed that a large number of Japanese consumers threw away imported rice which was sold as a tie-in of domestic rice. Table 2 Estimated coefficients of the random parameters logit and nested logit model Random parameters logit Nested logit Variable Coefficient T-statistic Coefficient T-statistic Attributes in the utility functions Price -0.9291E-3 US 1.551 US (s.d.) 0.4567E-3 AU 1.912 AU (s.d.) 0.7923E-2 CH 1.540 CH (s.d.) 0.2569E-1 AK 2.628 MY 2.506 IB 2.375 NG 2.847 US-SAFE -0.9269 US-TASTE -1.795 AU-SAFE -2.014 AU-TASTE -1.013 CH-SAFE -1.236 CH-TASTE -1.492 HAB-AK 0.8590 HAB-MY 0.7434 HAB-IB 1.292 HAB-NG 1.487 Attributes of the branch choice equations Constantforeign - ATTFORforeign - GENDERforeign - NCHILDforeign - INCforeign - Constantdomestic - ATTFORdomestic - GENDERdomestic - NCHILDdomestic - INCdomestic - Inclusive value parameters FOREIGN - DOMESTIC - No.of observations 1713 Log likelihood -1879.4 Adj. pseudo R2 0.470 -7.236 5.830 0.004 7.998 0.069 3.759 0.179 8.297 8.013 7.989 7.571 -2.524 -5.361 -6.968 -3.285 -2.592 -3.677 4.159 1.537 2.651 8.334 Note: Distribution simulations for RPL are based on 50 draws. - - - - - - - - - - - - -0.1117E-2 1.445 - 1.792 - 1.341 - 3.137 2.997 2.813 3.444 -0.9229 -2.694 -2.279 -1.382 -1.738 -1.866 0.86641 0.8530 1.615 1.511 2.813 -0.6291 -1.041 -0.3333 -0.2166E-3 -0.5757 0.2455 1.127 0.2552 -0.8750E-3 -7.347 2.530 - 3.675 - 2.190 - 8.450 8.130 8.308 7.796 -1.835 -5.221 -5.109 -3.157 -2.871 -3.437 4.152 1.828 2.947 7.953 6.093 -5.844 -4.145 -2.032 -0.829 -0.955 1.545 2.995 0.988 -2.126 0.7284 6.654 0.4146 2.262 1713 -1809.5 0.524 Table 3 Willingness-to-pay estimates for each rice brand (yen per 5 kilograms) WTP Negative or uncertain image Safety Taste Habit Retail price U.S. Koshihikari 1293.8 467.5 -1118.3 1988.5 Australian Koshihikari 1604.4 -436.3 367.4 1927.5 Chinese Akitakomachi 1201.1 -354.9 -469.5 1650.5 Akita Akitakomachi 2809.1 3582.8 2343.5 Miyagi Hitomebore 2683.3 3447.1 2396.5 Ibaraki Koshihikari 2518.7 3965.2 2309.5 Niigata Koshihikari 3084.0 4436.6 2789.0 Note: Average retail price in 2000 reported for 5 kilogram packages by Food Agency, Japan. 5. Summary and Implications Rice purchasing decisions of residents of Tokyo were analyzed in the RPL and NL model. Although the RPL model did not improve the result of the MNL model, the NL model was proved to be more explainable for the rice brand choice including imported rice. Choice modeling to elicit consumers’ WTP for imported brands of rice revealed that the current retail prices for them were higher than the average consumers’ WTP, while domestic brands were priced below the average. This suggests that sales volume of imported rice varieties could increase by lowering the price. If consumers had highly negative images toward imported brands in terms of safety or taste, or if they were not familiar with imported varieties and their safety and taste were unknown to them, their WTP was harshly discounted. Since general Japanese consumers tend to believe that domestic produce is superior to imported produce in safety or taste, it is critical for exporting countries to inform and convince them regarding the safety and taste of their produce in comparison with domestic rice. On the other hand, for Japanese rice farmer, they should make more effort to realize less input agricultural production. And the Japanese government may support such farming practices to reduce chemical inputs by introducing the environmental payment scheme. Also, it is emphasized that resistance to foreign produce was smaller for fresh vegetables than imported rice. Only 11 % of respondents stuck to domestic produce. It might be for this reason that the majority of imported rice was routed for processing, and average Japanese had not viewed it in stores and tasted it. Hence, the results are merely reflecting the consumers’ current perception of foreign rice which is unknown to most of the Japanese consumers. Although the current results provide realistic WTP estimates judging from the estimated coefficients of interaction variables, the interpretation of the results can not be expanded to general Japanese consumers. In-depth follow-up surveys for more consumers at other regions will be important to reveal the general consumers’ attitude toward imported rice. 1) The estimation results of MNL model were not given in table 2 since estimated coefficients of MNL model were quite similar to those of RPL model. References Baker, G. A. “Consumer Preferences for Food Safety Attributes in Fresh Apples: Market Segments, Consumer Characteristics, and Marketing Opportunities.” Journal of Agricultural and Resource Economics 24, 1999: 80-97. Burton, M., D. Rigby, T. Young., and S. James. “Consumer Attitudes to Genetically Modified Organisms in Food in the U.K.” European Review of Agricultural Economics 28, 2001: 479-98. Greene, W.H. Econometric Analysis, Fourth Edition, Prentice Hall, 2000. Kodama, Y. “Analysis of Consumer Preference on Local Brands.” Journal of Rural Economics, Special Issue, 2001: 149-151. Louviere, J.J., D.A. Hensher, and J.D. Swait. Stated Choice Methods: Analysis and Application, Cambridge, 2000. Quagrainie, K. K., J. Unterschultz, and M. M. Veeman. “Effects of Product Origin and Selected Demographic on Consumer Choice of Red Meats.” Canadian Journal of Agricultural Economics 46, 1998: 201-19. Sato, K., H. Iwamoto, and K. Demura. “Using Choice Based Conjoint Analysis to Assess Competitiveness of Chemical-free Hokkaido Rice.” Journal of Rural Problem, 142, 2001: 37-49. Train. K.E. “Recreation Demand Models with Taste Difference Over People.” Land Economics 74(2), 1998: 230-39. West, G. E., B. Larue, C. Gendron, and S. L. Scott. “Consumer Confusion over the Significance of Meat Attributes: The Case of Veal.” Journal of Consumer Policy 25, 2002: 65-88.