Willingness to Pay Premium for Foods Produced in Taiwan and Country of Origin Labeling: Results from Auction Experiment

Chun-Yu Chang and Wen S. Chern Department of Economics National Chung Cheng University
E-mail: chern.1@ccu.edu.tw

Draft August 2008

Abstract
This research is aimed at investigating the consumer’s preference for food produced in Taiwan and the economic benefits for the country of origin labeling. The study uses both experimental auction and contingent valuation method (CVM) to investigate factors that affect the consumer’s willingness to pay (WTP) for products under country-of-origin labeling (COOL). Experimental auctions of Taiwan and China preserved olives as well as Taiwan, China and Vietnam oolong teas were conducted using the Vickrey’s second price sealed bid auction. For CVM, the study used the double-bounded dichotomous choice method in which we started assuming the same base price for all products in the first question and then varied the prices in the second CV question. The products not chosen in the first question were offered with a discount in a range from 10% to 50% in the following question. Based on auction data, the Tobit model shows that the estimated premiums are 58.1%, 78.15% and 98.13% for Taiwan products over their alternatives of China olives, China oolong tea, and Vietnam oolong tea, respectively. Base on the CVM, the estimated premiums for Taiwan over China olives from a logit model is 67%, and the premiums for Taiwan produced oolong tea should lie between 50% to infinity (unknown upper bound) over China and Vietnam alternatives as the multinomial logit model can not be successfully estimated due to too few choice switches with discounted prices. The study thus demonstrates the superiority of the experimental auction over the CVM in eliciting the WTP for foods produced in Taiwan. The study concludes that enacting a COOL law would increase economic benefits to consumers in Taiwan. Keywords: country-of-origin labeling, experimental auction, Tobit model, contingent valuation method, logit model, multinomial logit model, willingness to pay

1. Introduction
The increasing standards of living and concerns about food safety have raised the consumer’s demand for information about the safety, origin, and processes used to produce the food they consume. Since Taiwan joined the World Trade Organization (WTO) in 2001, its agricultural sector has faced great competition from foreign imports of food products, especially from China, the United States, and those from South East Asia like Thailand and Vietnam. The flooded foreign imports, especially those from developing countries are often with lower quality and higher probability of contamination than those produced domestically in Taiwan. Recent reported contaminated imported foreign food included 1,142 trunks of contaminated coconuts from Thailand in April, 2008 and the contaminated stuffed dumplings from China in February, 2008. Those imported food often contained too much pesticide and preservative, which may cause neurodegenerative diseases as well as other diseases on people. Country-of-origin labeling (COOL) of food products will give Taiwanese consumers more information about where are those foods from and will give more protection on food safety. The more information the consumer gets from food labeling, the more likely they can avoid the threat of contaminated food. Since food produced domestically in Taiwan is probably safer than those from South East Asia, Taiwanese consumers would be willing to pay a premium to avoid contamination of foreign food. Unfortunately, for a lot of foods without a country-of-origin label, Taiwanese consumers currently can not tell where the food is from. Therefore, to protect Taiwanese consumers away from imported contaminated food, Taiwan government should enact a COOL law. But before we jump to the conclusion that the COOL law should be enacted in Taiwan, it is important to show whether the COOL on the imported food such as preserved olives or oolong teas is beneficial to the society. If the willingness to pay (WTP) premiums for the Taiwan products are large enough, then the COOL law could incur economic benefits to the consumer. This paper is aimed at providing such an economic analysis of the COOL in Taiwan. Since consumers are becoming increasingly concerned with the safety and origin of their food, many countries already have enacted mandatory country-of-origin labeling (MCOOL) law. Those countries, such as Japan, Australia, the United States and New Zealand, often show great confidence on their domestic products. In the 2002 Farm Bill, the U.S. congress first introduced country-of-origin labeling on beef, lamb, pork, fish, perishable agricultural commodities, and peanuts. The bill states,
1

“…for a commodity to be labeled a USA product, it must be born, raised, and processed in the United States” (U.S. Senate, Farm Bill Conference Framework, 2002). It later became mandatory in 2004, which has become known as MCOOL. In August 2007, the U.S. congress enacted a legislation requiring MCOOL for meat products. Japan is another country that is mature in COOL. Japan’s Ministry of Agriculture Forestry and Fisheries (MAFF) mandated COOL on all the fresh foods (including fruit, vegetable, meat and seafood) and processed food according to the Japanese Agricultural Standard (JAS) in 2000. In 2004, the MAFF mandated eight kinds of processed food, including eel, mackerel, and others, to label both the origin of the raw materials and the processing location. The Japanese government believes that the difference in the origin of raw materials will cause the differences in the processed products. In 2006, they added more items, such as processed beef products, etc. to the COOL law. Taiwan also follows the trend of COOL. The Act governing of Food Sanitation in Taiwan mandated COOL on all packaged food in 2006 and the law became effective in January, 2008. But Taiwan only defines the origin of country as the country where the products are produced or processed. Therefore, the commodity, which is made by semi-processed foreign imports of food products but complete their process and packaging in Taiwan, may be labeled as Taiwan products. For example, preserved fruit and processed eel products imported from China or other countries but finished their process in Taiwan will be labeled as Taiwan produced products. Taiwan is introducing its COOL in food, but not fast enough. For example, the COOL of unpackaged food in Taiwan is not mandatory. Therefore, whether the government should also include the COOL of unpackaged food and make it compulsively would also be an important issue. As the COOL law becomes a world-wide trend, numerous arguments have emerged in terms of for and against the COOL law on food products. Proponents of COOL argued that “consumers have a right to know” where their food is coming from (Food Market Institute, 2002) and they believe that COOL will provide domestic producers with competitive advantage. They also suggest that the country of origin labeling has great impact on the products, which do not have strong brand names. In contrast, opponents of the law have argued that the costs incurred by producers, importers, packers, wholesalers and retailers to segregate and label the effected products would be substantial.
2

Our study has the following five objectives. First objective is to develop a methodology for soliciting the consumer’s willingness to pay for food products produced in Taiwan as compared with those produced in foreign countries. This methodology will be based on auction experiment and contingent valuation method (CVM). Secondly, we will analyze what factors may affect the consumer’s behavior on purchasing products under COOL. Third, econometric models will be developed to quantitatively estimate the premiums that Taiwanese consumers are willing to pay for food produced in Taiwan. Forth, we will analyze whether the order of the section and the order of the trail will affect respondent’s willingness to pay or not. Last but not least, we will evaluate the economic benefits for enacting the country of origin labeling legislation in Taiwan. The paper will also provide suggestions on COOL of the unpackaged food to the government based on our econometric results.

2. Literature Review
The 2002 U.S. COOL law has been studied by many economists. Schupp and Gillespie (2001) first conducted a research on the attitude of food handlers and restaurants on the country-of-origin labeling law for fresh and frozen meats in U.S. They found that the COOL would be supported as long as the producer believes that consumers would benefit from the labeling. Umberger et al. (2003) used both survey and the auction (forth-price, sealed-bid) to estimate the willingness to pay premiums of country-of-origin labeled steaks. The price premium estimated from the auction method for the steak labeled with “USA guaranteed, born and raised in the U.S.” was 19%, which was larger than those from contingent valuation (CV) survey (11%). This result may be because consumers would like to see the product they are bidding on. They also showed that food-safety concerns, preferences for labeling source and origin information, strong desire to support U.S. products and beliefs on U.S. beef were reasons why consumers preferred COOL. Loureiro and Umberger (2003) conducted a CV survey of shoppers in grocery stores. They found that the willingness to pay premiums for steak and hamburger labeled as “U.S. Certified Steak” and “U.S. Certified Hamburger” was 38% and 58%, respectively. Al though most authors suggested that legislation of COOL would generate price premium on the domestic products, having a price premium does not mean there will be enhanced social welfare. Schmitz et al. (2005) showed that there exists a break - even cost that maintains the welfare at its original level during the COOL shock (COOL may cause supply shift
3

and demand shift). So, when the labeling cost is less than the break-even cost, having the country-of-origin labeling will increase the welfare. On the contrary, it will not. Lusk et al. (2004) found that the cost of COOL could transfer from industry to industry. Costs of COOL could be shifted from producers to processors and retailers, and in this case, producers would be better off while consumers will be worse off. According to their study, an increase in aggregate consumer demand of 2% to 3% is likely sufficient to offset the losing producer welfare due to increased COOL costs. There are many arguments on how to design an experimental auction mechanism correctly. Hoffman et al. (1993) used auction experiment to estimate the willingness to pay premiums for the vacuum-skin (VS) packaged steaks over the traditional overwrapped Styrofoam tray (OST) steaks. Through the auction, they found that the auction order of the products did not affect the estimation results. They also suggested that to design learning trails and instructions that explain incentive compatible auctions carefully was very important. Specifically, learning trails could teach respondents how to bid and the explanation of the auctions could minimize the impact of strategic behavior. To improve the accuracy on the auction results, Vickrey (1961) suggested that the second-price auction, in which the highest bidder would be awarded the object by just paying the second-highest bid price, is relatively easy to implement and it is a weakly dominant strategy for the participants to reveal their true valuations. To show the reason why second- price auction is considered to be incentive-compatible, for example, assume you bid more than your willingness to pay, if you win and the second-highest bid price is less than you are willing to pay you neither gain nor lose. But, if you win the bid and the second-highest bid price is more than your true value, you would have to pay more than you are willing to pay. Suppose you bid less than you are willing to pay. If you lose and the second-highest price is more than you are willing to pay you neither gain nor lose. But, if you lose and the second-highest price is less than or equal to your reservation value, you have lost the opportunity to purchase the object at a price you are willing to pay. Second-price auction helps to reveal the true WTP for the respondents. Therefore, it is adopted in this study. Corrigan et al. (2006) suggested that the bids of the purchase auction (coffee mug) would be influenced by posted prices for unrelated goods in trail auction (candy bar). To avoid the posted effect, they also suggested that we should calculate the WTP of the bid premiums instead WTP of the bids. Therefore, we also estimate the WTP of the bid premiums of the products in this study. Since the
4

willingness to accept (WTA) measures of value are often larger than WTP, whether to use WTP or WTA is also important to experimental auction. However, recent research suggests that the difference between WTP and WTA can be consistent with economic theory (Hanemann, 1991; Hoffman and Spitzer, 1993). Thus, an observed difference between WTP and WTA is not a per se behavioral violation of the incentive compatibility of the auction mechanism. Shogren et al. (1994) even showed that for market goods with close substitutes, there was a convergence of WTP and WTA measures of value. In this study, various oolong teas as well as preserved olives are market goods with close substitutes, so we consider estimating only Taiwanese consumer’s WTP not WTA for the preserved olives and teas. Beside experimental auction method, CVM was first proposed in theory by Ciriacy-Wantrup (1947) as a method for eliciting market valuation of a non-market good and has been widely used to evaluate food safety attributes (Eom, 1994; Buzby, Skees, and Ready, 1995; Huang, Kan, and Fu, 1999).This is because food safety attributes are embodied in the food products and not sold separately in the open market; the researchers need to collect primary data because market-level data are usually nonexistent. Single-bounded dichotomous choice format, which provides the respondents with some threshold value and asks them if they are willing to pay that amount, was first considered to be a better method than open-ended format, because it was easier for respondents to answer the questions. But through this method, the researcher can only know the respondent’s WTP is greater or smaller than the threshold (Haab and McConnell, 2002). To solve this problem, Hanemann and Kanninen (1991) showed that double-bounded dichotomous choice format would improve the welfare measure. Although it has been pointed out that double-bounded dichotomous choice format may cause inconsistency problem 1 , it still retains efficiency (Cameron and Quiggin, 1994; Herriges and Shogren, 1996). Hanemann and Kanninen (1999) suggested that the gain in efficiency could override the problem of inconsistency. Alberini (1995) also showed that the format was robust for estimating the mean or median of the welfare measure. We therefore decided to use the double-bounded dichotomous in our CV survey.

3. Instruments of Auction and Survey
3-1 Choices of Products
Which means that the respondent’s willingness-to-pay function may not be identical between the initial and follow-up CV questions. 5
1

The reason why we chose preserved olives and oolong teas for the auction experiment and the CVM is because both products were often imported in bulk and unpackaged. Since the retailers often try to fool consumers as though they were produced in Taiwan, those products are usually sold without country-of-origin labeling. Although there are a lot of other unpackaged foods could be used for the auction and experiment, such as preserved mangos and dried mushrooms. But preserved mangos from Thailand look different from Taiwan mangos, which are moister and softer than Thailand mangos. Dried mushrooms from different countries are also easy to distinguish by their looks. Korean dried mushrooms are also bigger than Taiwanese ones. Therefore, it is not easy for us to find food products from different countries with identical looks. It took us much time to find that preserved olives and oolong tea products from different countries that look very much the same. Another reason is that there were newspaper reports about imported contaminated preserved fruit and oolong teas. Most of the imported contaminated food, such as China olives and Vietnam tea, has too much preservative and insecticide residuals, which may be harmful for human health. For these reasons, those products are chosen in our study. The preserved olives have been marinated with Chinese herb for a long time and color of the preserved olives is black. It is very hard for us to distinguish their country-of-origin by their appearances. Oolong teas are also very hard for us to tell their differences by their looks. 3-2 Experimental Design After the focus group session held in Chiayi (National Chung Cheng University), we conducted three formal experimental auctions in Taipei (Academia Sinica), on March 13, 2008, Taichung (National Chung Hsing University), on March 14 and in Kaohsiung (National Kaohsiung Normal University), on March 18. Two sessions were held in each location; the experiments were conducted at 5:30 pm and 7:30 pm each day. Each experimental session involved nearly 13 general public samples: The female ratio is set at 60%-76%. Although the ratio of female and male in Taiwan is almost 49% or 50% (Taiwan Ministry of the Interior (2008)), the particular sex ratio in the experiment was chosen because that female often plays the role of buying food for the household. Therefore, we recruited more female respondents in our study. We screened potential participants recruited by the Survey Center of Academia Sinica to get a desire all mix of sample by income and sex for each session. These respondents signed up through the internet or telephone to the Academia Sinica
6

and were chosen if they never participated in any kind of experimental session. The total sample includes 74 participants. Participants were paid $1,000 for one and a half hours of their time and each of them was assigned an ID number. Each experiment started with three trail auctions of candy bar. The trail auctions were followed by six real auctions: three for preserved olives; three for oolong teas. The preserved olive auctions used olives from Taiwan and China. The oolong tea auctions involved teas from three different countries: Taiwan, China and Vietnam. Both of the olives and teas from different countries were packaged into a 150g bag. Trail auctions were conducted to get participants acquainted with the mechanism of the Vickrey second-price sealed-bid auction. In each auction, participants bid for two or three products, but only one product is actually sold. After the auctions, a straw was randomly drawn to determine which trial was binding, and then another random drawing of a straw was made to determine which product was for sale in the chosen trail. The use of random draws was made to control for the wealth effect: since only one product was sold, each participant did not need to split the budget over the trails or over the products (Shogren et al., 1994). The selling price was the second highest bid price and the panelist would have won the auction if their bid exceeded the selling price. Therefore the winner would only have to pay the second highest bid to buy the product. After the auctions, participants were asked to complete a survey about their food-purchasing behavior, and a set of contingent valuation questions corresponding to the same products (preserved olives and oolong teas), personal background and socio-demographic characteristics. 3-3 Contingent Valuation Questions Design The contingent valuation questions are questions for simulated market. The study used the double-bounded dichotomous choice method to guide the participants to make hypothetical purchase decisions under given price scenarios. We asked the respondents to make a purchase choice among the selected products (preserved olives and oolong teas) labeled with various countries of origin given various prices. The study designed a double bounded CV survey in which we started assuming the same base prices in the first question and then varied the prices in the second CV question. The prices for the products not chosen in the first question were lower in the following question. For preserved olives, there were four options of choice: Taiwan, China, “both of them are indifferent” or “buy neither of them”. If one chose Taiwan preserved olive as more preferable than the China preserved olives in the first
7

purchase choice question, we would lower the price of the China preserved olive in the follow-up question by a certain amount of discount. If one chose China preserved olives in the first question, we would lower the price of Taiwan preserved olives in the second question. For oolong teas, there were five possible choices: Taiwan, China, Vietnam, “all of them are indifferent” or “buy none of them”. Similar to the case of preserved olives, if one chose Taiwan tea, we would lower prices of the China and Vietnam oolong teas at the same level of discount in the follow-up question. If one chose China tea, then Taiwan tea would be discounted in the second question. The same procedure would happen if one chose the Vietnam tea in the beginning. Table 1 shows the discount range with 10%, 20%, 30%, 40% and 50% used in the survey. The price scenarios were randomly distributed among the participants. If the one selected “indifference” in the first question, the following question will ask whether they are still indifferent or whether they would choose the cheaper alternative. There were also two versions of the experiment in this study, three sessions with a total of 38 respondents were given auction first and then the CV survey, the other three sessions with 36 respondents were given the CV survey first and then followed with the auctions.

4. Descriptive Analysis
4-1 Definition of Variables Dependent variables and independent variables used in the econometric models for the experimental auction and the CVM may be different. Table 2 shows the definitions of dependent variables and symbols that are different from auction and CVM. For the experimental auction, we denote respectively Ot and Oc as the respondent’s bids for Taiwan and China preserved olives, and Tt, Tc and Tv as the bids for Taiwan, China and Vietnam oolong teas, respectively. For the CVM, the dependent variables include binomial Y and multinomial Y for preserved olives and oolong teas, respectively. The symbol Y is based on the follow-up question in the CV format. Besides the dependent variables, Table 2 also shows definitions and coding of independent variables that are different between auction and CVM. Independent variables from the auction include behavior variables and trail variables. The behavior intention variables represent respondent’s choice based on the first question in the CV format when all products were offered at the same price. The reason why the trail
8

variables are included in the auction is to see whether the posted price would affect the respondent’s bid in the subsequent trails. For the CVM, we use △PRICE to denote the price differences between the non-Taiwan product and Taiwan product in the CV format. Although there are some independent variables which are specifically defined for either auction or CVM, most of the independent variables are the same in both regression models. Table 3 shows the common independent variables that are shared between auction and CVM, including information variables as well as the demographic variables, such as sex (SEX), age (AGE), marriage (MAR), monthly individual income (MINC), yearly household income (INC), household expenditure for food away from home (FAFH), household expenditure for the food at home (FAH) and the number of family members (NUM). We expect that information variables, such as INFO2 and RISK would have a positive effect on Taiwan products, because people who care about the information of food as well as the risk of purchasing contaminated food would be willing to pay higher for Taiwan products that may be perceived as relatively safer than imported food from China or South East Asia. We also separate food expenditure variables into food at home and food away from home, we may expect that people who spend more on food away from home would probably care less about where the food from than people who spend more on food at home. Therefore, we assume that respondents with higher FAH and lower FAFH would have higher willingness to pay on Taiwan products. Other coefficients of the demographic variables, such as sex and age, could be either positive or negative, and therefore we could not have any expected sign on them. Table 3 also includes the product characteristics variables, such as safety (SAFTY), nutrition (NUTRI), county-of-origin labeling and the purchasing behavior variables, such as how often do you or members of your household purchase preserved fruits and oolong teas (BUYPF and BUYTEA). The mean of the variable SAFETY is one2, which means that all of the respondents consider food safety a very or somewhat important factor when purchasing food. If we add the variable SAFTY in models, there will be a collinear problem. Therefore, the variable SAFETY is excluded from both Tobit and CVM regressions in our study. Last two columns of Tables 2 and 3 present the means and the standard deviations of all variables used in the models.

2

There are exactly 98.7% of respondents consider food safety is very important (Figure 11) 9

4-2 Socioeconomic Variables The total sample includes 74 respondents from Taipei, Taichung and Kaohsiung. The distributions of gender, age and education level in each area are shown in Table 2.a (Appendix 2). The respondents from Taipei have more even distribution of education levels than Taichung and Kaohsiung. Although we tried to control the distribution of the three zones by education level, but more of the respondents who signed up for the experiments had higher than elementary school level in Taichung and Kaohsiung. Therefore, only one respondent in Taichung is in elementary school level. As can be seen from the distribution of the total sample in Table 2.a (Appendix 2), only ten participants are at elementary school level, most of the participants are in high school level. The average age of the sample is 39.3 years old and almost 73% of respondents are female. Figures 1 – 6 show the sample distributions of the selected socioeconomic variables. Figure 1 shows the distribution of marital status of the total sample. It is shown that about 72.8% of participants are married. Figure 2 shows that over 70% of the respondents from Kaohsiung are overweight. With respect to income distribution, Figure 3 shows that about 56.06% of participants had their average household income in 2007 between $ 340,000 and $1,020,000. Figure 4 shows that the average monthly personal income in these three cities. The results show that the average monthly personal income in Taipei is $29,560, which is much higher than that of $20,791 ($25,041) in Taichung (Kaohsiung). Figure 5 shows that the average household expenditure for the food at home (FAH) of the total sample is $12,965 and the average expenditure of the food away from home (FAFH) is $5,344. Respondents from Kaohsiung have the highest average expenditure on both FAH ($14,041) and FAFH ($6541). These are consistent with the fact that over 70% of respondents from Kaohsiung are overweight. Figure 6 shows the distribution of education levels and the highest percentage of respondents is those who completed a high school degree (48.5%). 4-3 Purchasing Attitude and Behavior The distributive statistics related to food purchase behavior are shown in Figures 7 - 15. Almost all respondents in the sample preferred preserved olives and oolong teas labeled as Taiwan products than from other countries. Figure 7 shows that most respondents (51.3%) purchase food several times a week and 23% shop for
10

foods every day, but Figures 8 and 9 show that 74.4% of the total sample rarely purchase preserved fruit and 70.3% of respondents purchase oolong teas only once in more than two months or even never. In Figure 10, respondents from Taipei (76%) and Kaohsiung (76%) purchased organic foods more frequently than those in Taichung. Therefore, respondents from Taipei and Kaohsiung appear to be more concerned about diet and health. Figure 11 shows that respondents consider safety (98.7%), freshness (94.3%), nutrition (62.1%) and county-of-origin (54.1%) as the most important factor that will affect their choices on food purchase. Environmental consciousness, price, calorie, convenience, color and brand are less important for the respondents. Furthermore, most respondents have confidence in food produced and labeled as Taiwan food. Most of participants suggest that their choices for Taiwan labeled food are for food safety (37.1%) and freshness (37.1%) concerns as shown in Figure 12. As high as 44.7% of the total sample considers the existence of imported contaminated food is the most important food safety problem happening in Taiwan currently (Figure 13). Therefore, most of the respondents would support a country-of-origin labeling law to reduce the probability of buying foreign food products with contamination. Figure 14 shows that most respondents get their information about food from friends and family. Figure 15 shows that there are 24.3% of respondents who do not know about the country-of-origin labeling law.

5. Econometric Models
5-1 Test of Affiliation It is important to examine whether the announced winning bid would affect the bidding behavior in the subsequent trails during auction experiment. It is possible that respondents found out that their bids were different from the posted price and therefore changed their bids in the next trail. Participants may also increase their bids if they observed the second highest bid was higher than their last bid. If there was no affiliation then there would be no correlation between the bid adjustment and the difference between the winning bid and the participant’s last bid. Following Kaneko and Chern, the model to investigate the possible affiliation can be expressed as:

ΔBi,t =α +β1Pi,t-1 +β 2 t+ε i,t

(1)

where i and t index individual and trail, respectively; ΔBi ,t is participant i’s bid in trail
11

t minus his or her bid in trail t-1; Pi ,t −1 is the posted price in trail t-1 minus participant i’s bid in trail t-1. Hence, if β1 >0, the participant would increase his or her bid if his or her last bid was lower than the posted bid, and vice versa. Trail number t is added as an explanatory variable to see if there is a trend that the individual bid prices would have more experience on bidding in repeated trails. Table 4 shows the results of the affiliation test. For all of the products, the coefficients for variable P are all significant and positive. It shows that participants would tend to increase their bids for all products when their last bid was lower than the posted bid. Because respondents will adjust their bids during the repeated trails, we should include the trail dummy variables in the auction regressions. For China olives as well as the teas from all three countries, the coefficients of the variable t are significantly positive, indicating that participants tend to increase their bids in the second trail if the posted bid was higher than their bids in the first trail.

5-2 Experimental Auction

We construct regression models for analyzing participants’ biding behavior in experimental auction. Consider first the bids for preserved olives, we run regressions
o for preserved olives labeled Taiwan ( Bo ) and China ( B1 ). For oolong tea bids, we run 0

T T T regressions for Taiwan tea ( B1 ), China tea ( B2 ) and Vietnam tea ( B3 ) to investigate

how the variables affect the respondent’s bids for these auctioned products. Table 5 shows the unconditional mean bids for each product by country. There were over 10% of respondents submitted zero bid for every product. Because the hypothesis of the linearity can not be held when there are too many zeros in the data set. This is a typical censoring problem in econometrics. Therefore, we use Tobit model for analyzing the bidding behavior for Bio (i=0, 1), BT (j=1, 2, 3.). Since bided premiums j
o T T T of (Bo -B1 ) , (B1 -BT ) and (B1 -B3 ) for Taiwan products can be positive or negative, 0 2

we can simply use ordinary least squares (OLS) method to run those premium regressions. Tobit model can be used to describe the relationship between a non-negative dependent variable yi and a vector of independent variables xi . The model is specified that there is a latent variable ( yi* ) linearly dependent on xi via the parameter vector β. In addition, there is a normally distributed error term ui to capture random influences on this relationship. The observable variable yi is defined
12

to be equal to the latent variable whenever the latent variable is positive and zero otherwise. That is,
⎧ yi* if yi* > 0 ⎪ yi = ⎨ * ⎪0 if yi ≤ 0 ⎩ where yi* is a latent variable: yi* = β ' xi + ui , ui ~ N (0, σ 2 ) . When the respondent’s willingness to pay ( yi ) is more than zero, the probability density function (pdf) is shown as: (2)

⎛ β ' xi ⎞ prob( yi* > 0) = f ( yi | yi > 0) = Φ i ⎜ ⎟ ⎝ σ ⎠

(3)

⎛ β ' xi where Φ i ⎜ ⎝ σ

⎞ ⎟ is a cumulative standard normal distribution function. ⎠

However, when the respondent’s willingness to pay is less or equal to zero, the cumulative density function (cdf) is shown as: ⎛ β ' xi ⎞ prob( yi* ≤ 0) = F ( yi = 0) = 1 − Φ i ⎜ ⎟ ⎝ σ ⎠ (4)

Amemiya (1973) has shown that the likelihood estimator suggested by Tobin for the model can be used to the estimate parameter vector β and σ by Maximum Likelihood Estimator (MLE). The likelihood function for the Tobit model is
⎡ ⎛ β ' xi ⎞ ⎤ ⎡ ⎛ β ' xi ⎞ ⎤ L = ∏ ⎢Φ i ⎜ ⎟ ⎥ ∏ ⎢1 − Φ i ⎜ σ ⎟ ⎥ ⎝ σ ⎠ ⎦ yi = 0 ⎣ ⎝ ⎠⎦ yi > 0 ⎣

(5)

After we take the partial derivation on β and σ, we can obtain the estimated ˆ ˆ parameters β and σ .With the estimated parameters calculated by equation (5), we can calculate the expected willingness to pay for consumer i as follow:
E (y* )=E(yi | yi > 0) × f ( yi | yi > 0)+E(yi | yi =0) × F( yi =0) i = [ x ' β + σλ ( β ' x / σ ) ] × Φ ( = β ' xi Φ (

β ' xi β ' xi ) + σφ ( ) σ σ

β ' xi )+0 σ

(6)

where λ ( β ' x / σ ) = φ ( β ' x / σ ) / Φ ( β ' x / σ ) is called the inverse Mills ratio; it is a ratio
between the standard normal pdf and standard normal cdf. Following Wooldridge (2006), if x j is a continuous variable, its marginal
13

effect can be expressed as:

β ' xi ∂E(yi ) )β j . = Φ( σ ∂x j

(7)

where x j is the jth independent variable. If x j is a binary variable, the effect of interest is obtained as the difference between E ( yi |yi > 0, x j = 1) and E ( yi |yi > 0, x j = 0) . Marginal effects involving other discrete variables can be handled similarly. The following equations are specified as the empirical Tobit model for preserved olives labeled as Taiwan and China and for oolong teas labeled as Taiwan, China and Vietnam.
Bio = β 0 +β1info2+β 2 RISK+β 3BUYPF+β 4 BUYTEA+β 5 BUYNG+β 6 CHECK+β 7 BRAND+

β8 NUTRI+β 9 ENVIR+β10 COLOR+β11GOV+β12 AGE+β13SEX+β14 MAR+β15 EDU3+β16 KID + β17 NUM +β18 FAH+β19 FAFH+β 20 VEGE+β 21BMI+β 22 INC+β 23SECTION+β 24 ZONE2 + β 25 ZONE3+β 26 TRAIL1+β 27 TRAIL3, i =0,1 (8)
o B1 -Bo =β 0 +β1info2+β 2 RISK+β 3BUYPF+β 4 BUYTEA+β 5 BUYNG+β 6 CHECK+β 7 BRAND+ 0

β8 NUTRI+β9 ENVIR+β10 COLOR+β11GOV+β12 AGE+β13SEX+β14 MAR+β15 EDU3+β16 KID + β17 NUM +β18 FAH+β19 FAFH+β 20 VEGE+β 21BMI+β 22 INC+β 23SECTION+β 24 ZONE2 + β 25 ZONE3+β 26 TRAIL1+β 27 TRAIL3 (9) T B j = β 0 +β1info2+β 2 RISK+β 3BUYPF+β 4 BUYTEA+β 5 BUYNG+β 6 CHECK+β 7 BRAND+
β8 NUTRI+β 9 ENVIR+β10 COLOR+β11GOV+β12 AGE+β13SEX+β14 MAR+β15 EDU3+β16 KID + β17 NUM +β18 FAH+β19 FAFH+β 20 VEGE+β 21BMI+β 22 INC+β 23SECTION+β 24 ZONE2 j =1, 2, 3 (10) + β 25 ZONE3+β 26 TRAIL1+β 27 TRAIL3,
T B1 -BT = β 0 +β1info2+β 2 RISK+β 3 BUYPF+β 4 BUYTEA+β 5 BUYNG+β 6 CHECK+β 7 BRAND+ j

β8 NUTRI+β 9 ENVIR+β10 COLOR+β11GOV+β12 AGE+β13SEX+β14 MAR+β15 EDU3+β16 KID + β17 NUM +β18 FAH+β19 FAFH+β 20 VEGE+β 21BMI+β 22 INC+β 23SECTION+β 24 ZONE2 + β 25 ZONE3+β 26 TRAIL1+β 27 TRAIL3, j =2, 3 (11)
Note that in equations (8) and (9), the subscript i=1 is for Taiwan and i=0 is for China. In equation (10) and (11), j=1 is Taiwan while j=2, 3 are for China and Vietnam, respectively.
5-3 Contingent Valuation Method

The basic framework for analysis is based on the random utility model (Haab and McConnell, 2002). Let U ij denote consumer i’s utility from choosing alternative j. Then consumer i chooses alternative j if U ij > U ik for all k≠j. It is standard to assume that U ij = Vij + ε ij . Vij is the deterministic component of the utility of
14

respondent i from choosing j. ε ij is the error term. Since the data set is small in the study, it is not appropriate to discuss the distribution of the data. Therefore, we do not show the distribution of the error term in the study. According to the Haab and McConnell, the regression results using the probit model and the logit model (where the error term has a normal distribution) should be almost the same. For preserved olives, the appropriate model is the binary logit model because the choice is between preserved olives labeled Taiwan and those labeled China alternatives. For oolong tea, the choice is between Taiwan (j=1) and China oolong tea (j=2) and between Taiwan (j=1) and Vietnam oolong tea (j=3), which is better handled by a multinomial logit model. We assume that the random components ε ij are independently and identically distributed as a Gumbel distribution. Then the model for the choices of oolong tea is given by
pr (Yi = j ) = e
3 k =1 Vij Vik

∑e

j=1, 2, 3

(12)

where the respondent i’s observed choice ( Yi ) takes the value of 1 (Taiwan ),2 (China), or 3 (Vietnam). The log-likelihood function for the multinomial logit model is then given by ln L = ∑∑ dij ln
i =1 j =1 n 3 V

e ij

∑e
k =1

J

Vik

(13)

⎧1, if individual i chose alternative j Where dij = ⎨ ⎩0, otherwise

We further assume that the deterministic component is linear in parameters: Vij = β 0j +β1Pij +β 2j'x i (14)

where Pij is the price of the jth alternative and xi is a vector of consumer i’s demographic characteristics and other subjective and risk relative components. The log-likilihood function in equation (13) is maximized with respect to the parameters to obtain the maximum likelihood estimates of the parameters. Not all of the β 0j ’s and β 2j ’s are identifiable, so we adopt a normalization rule such that β 01 = 0 and

β 21 = 0 (see Greene, p.860).
15

Following the procedure used previously by Kaneko ad Chern (2005) (j=1 for Taiwan, j=0 for China) we have the following utility functions for preserved olives: U i0 =β1Pi0 +ε i0 U i1 =β 01 +β1Pi1 +β 21'x+ε i1 , For the oolong teas, we have the following utility functions: U i1 =β1Pi1 +ε i U i2 =β 02 +β1Pi2 +β 22 'x i +ε i2 U i3 =β 03 +β1Pi3 +β 23'x i +ε i3 (16) The model for the paired choice could be formulated by using the latent variable defined as
Yi * = U i1 − U i 0 = ( β 01 − 0) + β1 ( Pi1 − Pi 0 ) +( β 21 − β 20 ) x1 + ... + (ε i1 − ε i 0 ) = β 0 + β1ΔPi + β 2 ' xi + ε i

(15)

(17)

Then, the respondent i chooses Taiwan preserved olives ( Yi =1) if Yi* >0 and China alternative ( Yi =0) if Yi* ≦0. The latent variable approach would not work for the multinomial logit model, and hence the choice probably is not directly linked with the sign of the coefficient in equation (17). Once the parameters are estimated, we can compute the sample mean of WTP. Consider the case of oolong teas. Let WTPi 2 denote consumer i’s willingness to pay a premium on the Taiwan oolong tea. Then the following equations have to hold:

β1 ( Pi 2 + WTPi 2 ) + ε i1 = U i1 = U i 2 = β 02 +β1Pi2 +β 22 'x i +ε i2 . The left–hand side is the utility
from consuming Taiwan oolong tea at the price of Taiwan oolong tea plus the premium. The right-hand side is simply the utility from consuming the China oolong tea at its own price. We can than solve that the WTPi 2 =

β 02 +β 22 x i +ε i2 -ε i1 . β1

Taking the expected value of the WTPi 2 , we obtain the expected willingness to pay for Taiwan oolong tea in order to avoid China oolong tea as: E (WTPi 2 | xi ) =

β 02 +β 22 'x i β1

(18)

We can also obtain the willingness to pay for Taiwan oolong tea in order to avoid Vietnam oolong tea as:

16

E (WTPi 3 | xi ) =

β 03 +β 23'x i β1

(19)

6. Results
6-1 Descriptive Statistics of CVM and Experimental Auction

Although there were 74 participants in the sample, but there were only 70 participants who filled out the form carefully and completely. All of the respondents had handed in their bidding cards during the experimental auction. Therefore, we still have 74 bids for all selected product and the mean bids of the total sample are shown previously in Table 2. Table 5 shows the unconditional mean bids for each product by country, which is computed from the raw auction bids. Since there are four respondents without completing their forms or having irrational answers to our questions, exclude them in the econometric analysis and therefore the descriptive statistics of bids from the auction in Table 5 only include 70 bids. The mean bids for Taiwan preserved olives ($46.71) and Taiwan oolong tea ($200.27) are higher than those from other countries. Therefore, most of the respondents consider Taiwan preserved olives and tea are more preferable than China preserved olives as well as China and Vietnam oolong teas. Figure 16 shows the reasons why some respondents submitted zero bid for the selected products. For the Taiwan olives and tea, 66% of respondents who submitted zero suggested that they have never eaten preserved olives or drank oolong teas, and therefore they would never buy those products. For China and Vietnam products, most of the respondents who bid zero questioned the safety of China olives and tea as well as Vietnam tea. Therefore, most of the respondents in the sample have in general more confidence on Taiwan olives and tea. Let us examine the patterns of responses to the initial CVM question where the products were sold at equal price. Figure 17 shows the percentages of respondents answering each option. Obviously, no respondents chose China preserved olives in the initial question. As is evident, respondents preferred Taiwan preserved olives significantly more than China olives when their prices were the same. Almost all respondents preferred Taiwan oolong tea initial, only one participant chose China oolong tea at equal price. In Table 6, when the respondents facing 10% (A) and 30% (C) price discounts on the China olives, they still considered buying Taiwan preserved olives. When facing a 20% (B) discount on the China preserved olives, only one
17

respondent switched the original preference to China. When facing a 40% (D) discount on China preserved olives, there were only 14.3% of respondents turned to China olives. When facing a 50% discount on the China preserved olives, the percentage of respondent’s preference switching increased to 15.4%. There were 8 respondents who chose the neither option. For the preserved olives, most of them (50%) indicated that they never buy preserved olives and for the oolong teas, almost 50% of participants who chose the neither option mentioned that they never drink tea. Figure 18 presents more detailed results showing these patterns of responses. Table 7 shows that almost every respondent chose Taiwan oolong tea when all teas were sold at the same price, except one respondent chose China Oolong tea initially. All of the respondents facing a 20% (B) price discount on non-Taiwan oolong tea still chose Taiwan oolong tea. When facing a 30% (C) discount on the non-Taiwan Oolong tea, only one (6.6%) of respondents switched to China oolong tea. Moreover, 76.9% of respondents did not switch even when other teas were offered with a 40% (D) price discount. Only one person, who chose China tea in the first question under D, switched preference to Taiwan tea in the follow-up question. Only one respondent (7.7%) chose Vietnam Oolong tea at a 50% (E) price discount from initially chosen Taiwan tea. Therefore, almost all respondents chose Taiwan products at the same price, and over 85% of them did not change their preference under the selected price discounts. It shows that respondents would stick to Taiwan products even under 10%-50% price discount, their preferences on Taiwan products are very strong.
6-2 Regression Results of Experimental Auction

Table 8 shows parameter estimates of bid regressions for preserved olives and oolong teas using data from all trails. Since there are 70 respondents and each of them can bid three times, these regressions are based on 210 observations. For preserved olives, the Tobit results for the Taiwan bid equation in the first column show that the estimated coefficients for the variables INFO2, AGE, EDU3, FAH, GOV, BRAND and RISK are significant, those variables are also significant for the OLS results for the bid premium of Taiwan olives over China olives in the third column. Positive signs of the INFO2 imply that people whose information is mostly from food package labels will bid more on Taiwan olives. Negative signs of the coefficient of AGE show that younger respondents tend to bid more on Taiwan olives. Older people often care more about their blood sugar and show less interest on preserved olives than younger
18

people. Elders care about the unhealthy elements that may damage their health from preserved olives and therefore have no interest on olives as well as where the olives come from. Younger people care about the COOL of olives to avoid purchasing contaminated foods and bid high on Taiwan olives. Therefore, signs of the coefficient of AGE are negative. Positive signs of the coefficient for the EDU3 mean that respondents with college education tend to have higher bids on Taiwan olives, which also cause a higher bid premium for Taiwan olives over China olives. These results show that education helps people to raise their concerns about food safety as well as where their food come from. Therefore, more people with better education would increase the demand for food safety. Positive signs of the coefficients of FAH for the Taiwan olive equation and the olive premium equation show that respondents who spend more of food expenditure on food at home tend to bid higher on Taiwan preserved olives, which is consistent with the assumption that people who spend more on food at home would probably care more about where their foods come from than those who spend little on food at home. Positive signs of the GOV show that respondents who think the government’s regulatory performance is excellent will trust Taiwan’s food safety legislation and will probably believe that Taiwan produced food should be relatively safe under the government’s protection. Therefore, these respondents tend to place higher premium on Taiwan produced olives. Since the signs of the coefficients of BRAND and RISK are significant for both the olive premium equation and the tea premium equations, we will discuss them later. The second column in Table 8 shows the Tobit results for the China olive bids. The sign of the coefficient of the BUYPF is negative, which shows that respondents who purchase preserved food more often should have more experience in judging the quality of preserved olives and thus they may bid less on China olives than on Taiwan olives. Therefore, the positive sign of the coefficient for BUYPF in the bid premium can obviously be expected. Signs of the coefficients for CHECK, BMI and SECTION are the same in the first and second columns. Positive signs of BMI in Taiwan olive bid equation and China olive bid equation show that respondents with slim or normal figures tend to bid more on preserved olives. The reason may be that overweight people usually eat more than slim or normal people, and if under the same budget, overweight people will have to distribute their money for other real meals, and therefore tend to bid less on olives than respondents with normal or slim figures. Although signs of the
19

coefficients of CHECK are negative in both individual bid equations (for Taiwan and China), the coefficient in the China olive bid equation is still smaller than that in Taiwan olive bid equation. This result shows that respondents who consider checking the labeling of food as important factor on food purchasing would decrease their bids on all olives, but lowering their bids on China alternative more than Taiwan olives. Since signs of the coefficients of SECTION are significant in both the olive equations and the tea equations, we will discuss them later. To sum up, since the coefficients of info2, GOV, FAH, EDU3, BUYPF and RISK in the olive bid premium equation are all positive, suggesting that respondents whose information are mostly from food package labels, respondents who spend more on food expenditure at home, respondents with higher education level, respondents who purchase preserved olives frequently and respondents who think imported contaminated foods as the most important food safety issue in Taiwan tend to have higher price premium on Taiwan olives. Table 8 also presents the parameter estimates of bid regressions for oolong teas. The forth, fifth and sixth columns present parameter estimates for Taiwan, China and Vietnam oolong teas, respectively. Variables CALORIE, ZONE2, ZONE3, FAH and SECTION are either significant or marginally significant determinants for Taiwan, China and Vietnam oolong tea bids. In all three tea bid equations, positive signs of the coefficients for CALORIE show that respondents who consider calorie to be an important factor when making food purchasing decisions will increase their bids. Since respondents who were more concerned about how much Calorie they eat often pay more attention on keeping their figures slim and fit, they tend to have higher willingness to pay on teas. Positive signs of the coefficients for ZONE2 and ZONE3 also show that respondents from Taipei and Kaohsiung tend to have higher bid on teas than respondents from Taichung. Since signs of the coefficients for FAH are negative, respondents who spend most food expenditure on food at home tend to decrease their bids on all three teas. Food away from home is always too greasy than food we cook by ourselves, therefore respondents who spend more on food at home expenditure do not want to increase their bids on teas. For the preserved olives as well as the oolong teas, positive signs of SECTION suggest respondents who did the auction first and the CVM later would increase their bids on all olives and teas. In other words, respondents may lower their bids after seeing the offered prices we had given on the CV questions.
20

The coefficients of INFO2, EDU3, AGE and CALORIE in the Taiwan tea bid equation are significant and positive. Therefore, respondents whose food information is mostly from food package labels, and who are older and better educated may bid more on the Taiwan tea. The positive sign of AGE shows that older respondents like to drink more tea, and thus tend to bid higher than younger people and the negative sign of the BUYTEA in the Taiwan tea bid equation shows that the respondents who purchase tea more frequently would tend to decrease their bids for tea products. This is perhaps because most of heavy tea drinking respondents have their own tea suppliers, they would not be interested in the teas we offered in the auction. The last two columns in Table 8 present the estimation results for Taiwan tea premiums. Variables GOV, AGE, EDU3, VEGE, MAR, BUYPF, NUM, RISK and BRAND are either significant or marginally significant for Taiwan tea premiums over China and Vietnam alternatives. Positive signs of AGE and EDU3 show respondents with higher education level and older age will have a higher price premium for Taiwan tea. Since older people love to drink teas, they have ample experience on choosing them. Most of the older respondents said that their preferences on Taiwan tea were not only for the concerns of food safety but also for the good quality of Taiwan tea, they suggested that Taiwan tea tastes better than China and Vietnam teas (Even though we stressed that they tasted the same during the auctions). Therefore, they think Taiwan tea is better than those from other countries and are willing to pay more for it. Negative signs of GOV mean that respondents who think that the government is doing a god job on legislation tend to decrease their premiums. This is perhaps because respondents who think the government’s regulatory performance is excellent may trust that government will strictly restrain the import of contaminated food, and therefore no matter what country-of-origin labeling of the food, they are safe enough to eat or drink. We note that the coefficients of GOV have the opposite signs between the tea premiums and olive premium equations. This seemingly strange result is mainly caused by those who think the government’s regulatory performance being excellent bid significantly higher for Taiwan olives while this GOV variable is not significant in all other bid equations. Unfortunately, there is still no good explanation on the opposite results between olives and tea premiums. Coefficients of VEGE in all tea premium regressions are positive, suggesting that if respondents were vegetarian, their willingness to pay for Taiwan tea are more than those from China and Vietnam. Positive signs of the coefficients of MAR and BUYPF show that
21

respondents who are married and who purchase preserved olives frequently tend to increase their bid premium on Taiwan tea. Since variables BRAND and RISK are significant both in the Taiwan olive premium equation and Taiwan tea premium equations, we will discuss them together at the end of the paragraph. Table 8 shows that signs of the coefficients for BRAND are negative and the signs for RISK are positive in all Taiwan product premium equations. Negative signs for the variable BRAND mean that respondents who think brand is the most important factor that affect food purchasing tend to decrease their premiums on Taiwan products relative to others. This is perhaps because respondents who place more importance on brand believe that brand already gives them enough food protection. Therefore, if the brand of the food is reliable, then no matter what country-of-origin of the food is, they will consider the food to be safe enough to eat or drink. Positive signs of RISK in the Taiwan premium equations also show that respondents who consider imported contaminated food is very important issue in Taiwan currently will bid higher (lower) on Taiwan (non-Taiwan) tea and olives. Therefore, these respondents who care more about the labeling of food also care where the food produced and show greater preference on Taiwan tea and olives. In the affiliation test, we show that the participants would tend to increase their bids for all the products when their last bids were lower than the winning bid. Therefore, variables TRAIL should be significant in the bid or premium equations. However, the coefficients of TRAIL1 and 3 are not significant except in the Taiwan olive bid equation. Previous studies suggest that using only the data from the first trail to estimate parameters of bid regressions is better than using data from all trails, because the respondent’s bid for the next trail is often affected by the posted price from the previous trail. Therefore, we also estimate the parameters for bid regressions for preserved olives and oolong teas using data from the first trail. These results are shown in Table 9. Comparing the parameter estimates of regressions in Tables 8 and 9, the regression results based on data from all trails have more significant variables than those based on the first trail only. Especially, there are not many significant variables in the bid equations for olives and teas using data only from the first trail. For the preserved olives and oolong teas, although all of the signs of significant variables for bids equations from the first trail are the same as those from all trails, the standard errors of the significant variables from the first trail are larger than those from all
22

trails. This result shows that using bids from all trails is more efficient than using only data from the first trail. The reason is that there are only 70 data points in the first trail. In the affiliation test, we show that the TRAIL variables should play an important role for bid equations. But the regression result is different when including other variables in the equation. One possible reason is that when including other variables, the TRAIL variables become less important than other variables. Table 10 shows the marginal effect calculated from the parameter estimates of the bid regressions using data from all trails in Table 8. Since variables AGE, KID, NUM, FAH and FAFH are continuous variables, we use equation (7) to calculate the marginal effect. For other dummy variables, we use the difference between E ( yi |yi > 0, x j = 1) and E ( yi |yi > 0, x j = 0) to calculate the marginal effect. We do not discuss the marginal effect of Taiwan bid premiums over non-Taiwan alternatives, because the bid premium regressions are based on OLS. In this section, we discuss the marginal effects for significant variables only. For the variable INFO2, respondents whose information is from the food package labels will bid more than respondents who are not by the amount about $7.4 ($3.69) for Taiwan preserved olives (Taiwan oolong tea). The marginal effect on the variable RISK shows that respondents who think imported contaminated food is the most serious food problem in Taiwan are willing to pay $14.94 more on Taiwan tea and $8.59 less on China olives than those who do not. For the variable BUYPF, respondents who purchase preserved olives frequently would pay $11.27 ($37.11 and $40.44) lower on China olives (China tea and Vietnam tea) than who do not. Since most respondents who purchase teas frequently would have a specific tea dealer, they may care less on COOL of teas and pay $18.1 less on teas we auctioned than those who purchase tea not as frequently. The marginal effect of the variable CHECK shows that respondents who often check the food labels would pay $47.66 ($35.39) less on non-Taiwan tea than those who does not check food labels frequently. Though for both olives, respondents who often check the food labels will bid lower than who do not, but their bids on Taiwan olives less than the China alternative. The marginal effect on Taiwan olives for the variable GOV is $4.7, which means that respondents who believe in Taiwan government are willing to pay $4.7 more on Taiwan olives than those who do not think that Taiwan government does a good job. The variables SEX and AGE show that males tend to pay $2.94 and $3.03 more on China olives and tea, respectively, than females and respondents who are one year younger will
23

increase (decrease) their bids by $0.51 ($3.52 and $1.3) for Taiwan olives (Taiwan tea and Vietnam tea). These results show that females are more concerned on COOL than males and respondents with different age shows different reaction on Taiwan olives and tea. The variable EDU3 shows that respondents who have college degree will pay $8.1 ($64.1) more on Taiwan olives (tea) than those whose education level is lower than college. Therefore, respondents with higher education level will bid more on the Taiwan products. The variable SECTION shows that respondents who did the auction first and CV question will increase his or her bid on Taiwan and China preserved olives (Taiwan, China and Vietnam teas) by $12.64 and $11.56 ($57.3, $46.13 and $41.05, respectively) more than respondents who did the CV survey first and auction later. The positive marginal effects of the ZONE2 (ZONE3) on Taiwan, China and Vietnam teas are $21.39, $37.4 and $32.14 ($23.76, $8.13 and $3.42), which means that respondents from Taipei and Kaohsiung tend to have higher bids on teas than respondents from Taichung. Table 11 compares the estimates of the willingness to pay (WTP) for preserved olives and oolong teas using data from all trails and those from the first trail. The estimated mean WTP of Taiwan preserved olives and China preserved olives were derived from auction regression using all trails and they are $46.81 and $29.61 respectively, which are higher than the estimated mean WTP calculated based on first trail only ($43.74and $27.42). Table 11 also shows the estimated WTP values for oolong teas. The estimated WTP using all trails for Taiwan, China and Vietnam oolong teas are $205.15, $115.15 and $103.54, respectively, which are also higher than the estimated based on the first trail only ($188.88, $108.28 and $97.56).

6-3 Regression Results of CVM

Table 12 shows parameter estimates for the contingent valuation model for preserved olives. These logit regressions are based on the responses from the follow-up questions. We exclude those who selected “indifference” and “neither of both products” in the first question, because the follow-up questions for the “indifference” and “neither” options were not asked to make further choices with specific price scenarios. For preserved olives, the logit results show that the estimated coefficients for variables CHECK, COOL, MAR, AGE, FAFH and △PRICE are significant. The negative sign of the △PRICE indicates that the higher the price is,
24

the lower the utility of both Taiwan and non-Taiwan alternatives is. Furthermore, it implies that respondents are more likely to choose the non-Taiwan alternative if it is less expensive than the Taiwan product. The positive signs of CHECK and COOL mean that respondents who consider checking the labeling of food and country-of-origin of food as important tend to have higher utility on Taiwan olives relative to non-Taiwan alternatives. This is reasonable, since people put emphasis on what they consider important, the positive signs show that respondents who read food labels and country-of-origin of food prefer Taiwan olives more than China olives. They appear to believe that Taiwan olives are better than China olives. The negative signs of MAR, AGE and FAFH imply that respondents who are not married, younger and spend less on food away from home have higher aversion on Taiwan olives. The signs of the coefficients for AGE and FAFH are the same as the signs for AGE and FAFH appear in the auction results in Taiwan olive bid equation. In our CVM survey design for oolong teas, the first question offered Taiwan, China and Vietnam tea options at the same price. If one chose Taiwan tea, then in the second question, all three teas were offered with Taiwan tea at a higher price while the other teas had the same prices offered in the first question. Under this set up, if the respondent still chose Taiwan tea in the second question, we considered only two effective pairs of choices – between Taiwan and China and between Taiwan and Vietnam. However, if the respondent switched from Taiwan to non-Taiwan alternatives in the second question, the effective two pair choices would be between Taiwan and China, and between Taiwan and Vietnam. Furthermore, if in the first question, the respondent chose non-Taiwan alternatives, then in the follow-up question, we only offered one pair choice between the chosen non-Taiwan alternative and Taiwan tea. Even though we did not offer all possible combinations of complete choices with three alternative teas, we still can model these choices with a multinomial Logit model. In order the estimate the WTP for teas under COOL, we would had have estimated a multinomial Logit model since there are three different teas – from Taiwan, China and Vietnam. Unfortunately the model has never converged even after we narrowed down to a few variables similar to those used for preserved olives. The main and perhaps the only problem for not getting convergence is that there are few choice switches when we discounted the prices for those teas which were not chosen in the first CV questions offering all teas with the same price. As noted earlier in the
25

descriptive analysis (Table 7), only one respondent chose China tea and the rest chose Taiwan tea in the first question. In the second question, 59 respondents still chose Taiwan tea and only six respondents switched to non-Taiwan teas. Even with a 50% discounts for China and Vietnam teas, only one respondent switched from Taiwan to Vietnam tea. Overall we have 118 paired choices and about 95% of observations did not change their preferences. With such a few respondents choosing non-Taiwan teas with discounts in the data set, the model just simply did not work. Note also the maximum discount that we offered in the CV questions is 50%. We can thus conclude that the willingness to pay premiums for Taiwan tea over both China and Vietnam teas are at least 50% of the base price used in the first CV question. However, we would not be able to determine the upper bound for the WTP premium. This finding suggests that in order to obtain usable results, we should have offered more discounts, perhaps up to 70% or even 90% to investigate whether the consumer has such a strong preference that they will never purchase China tea or Vietnam tea no matter how cheap they are. Of course, that should not be the case because we have observed that many of the participants in the auction experiments gave non-zero bids for China and Vietnam teas. The fact that the WTP for all three teas can be successfully estimated with experimental auction, but could not be estimated with the CVM would demonstrate the superiority of the experimental auction for this particular study. Table 13 presents all estimated premiums for Taiwan products. From auction experiments, for Taiwan olives and teas, estimated Taiwan product premiums based on the OLS parameter estimates are higher than the Tobit model and the unconditional auction. But it is not appropriate to compare in absolute value of Taiwan premium derived from the OLS regression, Tobit regression and the raw auction since there is no base price for auction bids. Therefore, we added a row of Taiwan premiums in percentage terms. The percentage premium of the Tobit model is calculated by the individual expected WTP using all trails in Table 11, the difference of the Taiwan product expected WTP and the non-Taiwan alternative is divided by the expected WTP of non-Taiwan alternative derived from the Tobit model. The estimated percentage premium of the CVM is shown to be higher by factors of 1.15 and 1.14 than those from the Tobit model and raw auction data, respectively, for the preserved olives. Based on these results, there appears to be a considerable hypothetical bias. For oolong teas, as indicated earlier, we could not successfully estimate the needed multinomial Logit model from the CVM survey data. We can only conclude
26

that the willingness to pay premiums for Taiwan tea over China and Vietnam teas would be more than 50%, but we could not get the mean estimate, nor the upper bound. Based on the data from the auction experiments, the Tobit model shows that the estimated WTP premiums for Taiwan tea over China and Vietnam alternatives are 78% and 98%, respectively which are higher than those unconditional means estimated directly from the raw data. The OLS estimates which are less credible are even higher. Overall these results show a strong preference of the Taiwanese consumer for both preserved olives and tea produced domestically over those of imported alternatives from China and Vietnam.

7. Conclusions
This study attempts to use the auction experiment and CVM to explore the Taiwanese consumer’s preference for several selected foods produced in Taiwan v.s their imported alternatives. The study has important implications for the country-of-origin labeling legislation in Taiwan. The econometric results show that Taiwanese consumers have higher willingness to pay for Taiwan preserved olives and tea than the imported alternatives from China and Vietnam. It is clear that respondents are willing to pay more to avoid China and Vietnam alternatives. The results also show that the estimated Taiwan tea premiums in percentage terms are more than the estimated Taiwan preserved olive premium. Since food safety is the most important factor for these preference differentials, the results suggest that Taiwanese consumers are more concerned about the safety of teas than preserved olives. Results from the experimental auction show that higher educated respondents and respondents who think food label reading as well as imported contaminated food problems in Taiwan are important tend to have positive price premiums for both Taiwan olives and tea. If the respondents considered brand as an important factor for food purchasing, they would rely on the brand and neglect the importance of COOL, and therefore they tend to reduce the price premiums for Taiwan produced olives and tea. For oolong teas, respondents from Taipei and Kaohsiung have higher bids on teas than those from Taichung. The study also shows that the order of auction and survey as well as the order of the trail will affect the respondent’s bidding behavior, but it would not affect the respondent’s bid premium. Respondents who did the CVM questions fist and then
27

auction later tend to lower his or her bids. The results from the affiliation test show that participants in the auction would tend to increase their bids for all the products when their last bid was lower than the posted second - highest bid. Therefore, it is very important to design the experimental auction carefully. It would be an interesting task for the followers to design the order of the section well and to discuss how does the posted price affect the bids. For the CVM questionnaire, almost every respondent chose Taiwan products when all products were offered at the same price. Even after a discount up to 50%, few switched to cheaper items in the follow-up question. The survey suggests that the Taiwanese respondents were not responsive to the price differences when they come to choices between Taiwan tea and those from China and Vietnam. Almost all Taiwanese respondents were not willing to accept Chinese or Vietnamese tea. Since the logit and the multinomial logit models require the data with sufficiently large proportions of respondents falling into each choice category, the parameters based on CVM are extremely difficult to estimate. To solve this problem, we should increase the price discount range to 70% or even 90% in future research. In order to elicit the WTP for the Taiwan tea as compared to its imported alternative in this study, the experimental auction is clearly superior to the CVM. It is tempting to conclude that Taiwan has a prestige on preserved olives and oolong teas. Since 59.5% of respondents think country-of-origin of food is a very important food attribute, the COOL legislation is important for Taiwan. The study also shows that if the government enacts a country of origin labeling law on preserved olives and teas, the consumer’s willingness to pay for the Taiwan olive and tea relative to those from China and Vietnam, can than be revealed in the market price. Taiwanese consumers would be definitely beneficial from those labeling legislation. Since Taiwan product premiums in percentage terms are more than 50%, we would offer suggestions on COOL to the Taiwan government based on our survey and econometric results. First, since only 5% of respondents are very well informed about COOL in the study, the government should spread the information on the existing COOL law more progressively to the public. It can be done through building website for COOL and giving propaganda and lectures to the public. Second, according to the study, respondents who think food label reading and imported contaminated food problems in Taiwan are important tend to have positive price premium on Taiwan olives and tea. Therefore, adding the country-of-origin labels on non-packaged food
28

would help the competitiveness of Taiwan products. Third, education is shown to be a very important factor that will affect people on their purchasing behavior. Specifically, respondents with college degree would bid at $8.1 more on Taiwan olives and $64.1 in Taiwan tea than respondents with lower education level. With increasing emphasis on education in Taiwan, the payoff for COOL legislation would increase in the future. Comparing to other countries, such as the United States and Japan, they have more rigorous COOL law than Taiwan. They enacted MCOOL not only on packaged food but also on unpackaged food as well as fresh products. Therefore, we should take them as a role model to improve our current COOL law. In summary, imported contaminated food has become a threat to food safety for Taiwanese consumers since we joined the WTO. Many foods or related products we consume are found to be contaminated, such as bleaching chopsticks from China, pesticide residuals from Vietnam tea. We should urge the Taiwan government to enact the COOL legislation for non-packaged food products as soon as possible.

29

References
Alberini, A (1995), Efficiency vs. Bias of Willingness-to-Pay Estimates: Bivariate and Interval-Data Models, Journal of Environmental Economics and Management 29: 169-180. Amemiya, Takeshi (1973), Regression Analysis When the Dependent Variable is Truncated Normal, Econometrica 41 (6): 997–1016. Buzby, J.C., J.R. Skees and R.C. Ready. (1995), Using Contingent Valuation to Value Food Safety: A Case Study of Grapefruit and Pesticide Residues, Valuing Food Safety and Nutrition. J.A. Caswell, ed. Boulder, CO: Westview Press. Cameron, T.A. and J. Quiggin (1994), Estimation Using Contingent Valuation Data from a ‘Dichotomous Choice with Follow-up’ Questionnaire, Journal of Environmental Economics and Management 27: 218-234. Colin Carter, B. Krissoff and A.P. Zwane(2006), Can Country of Origin Labeling Suceed as a Marketing Tool for Produce? Lessons From Three Case Studies, Canadian Journal of Agricultural Economics 54: 513-530. Corrigan, Jay R. and Mattew C. Rousu (2006), Posted Prices and Bid affiliation: Evidence from Experimental Auctions, American Journal of Agricultural Economics 88 (4): 1078-1090. Ciriacy-Wantrup, S.V. (1947), Capital Returns from Soil Conservation, Journal of Farm Economics 29: 1181-1196. Eom, Y.S. (1994), Pesticide Residue Risk and Food Safety Valuation, American Journal of Agricultural Economics 76: 760-771. Food Marketing Institute (2002), Mandatory Country-of-Origin Labeling, FMI Backgrounder, http://www.fmi.org/media/bg/COOLabeling_02.pdf. Accessed February 24, 2008. Greene, W.H. Econometric Analysis, 4 th ed. Upper Saddle River, NJ: Prentice Hall, 2000. Haab, T.C. and K.E. McConnell (2002), Valuing Environmental and Natural Resources: The Econometrics of Non-Market Valuation, Cheltenham, U.K. and Northampton, MA, U.S.A.: Edward Elgar. Hanemann, M., J. Loomis and B. Kanninen (1991), Statistical Efficiency of Double-Bounded Dichotomous Choice Contingent Valuation, American Journal of Agricultural Economics 73: 1255-1263. Hanemann, M. and B. Kanninen (1999), The Statistical Analysis of Discrete Response CV Data, Valuing Environmental Preferences, New York: Oxford University Press.
30

Hanemann, W.M. (1991), Willingness to Pay and Willingness to Accept: How Much Can They Differ? , American Economic Review, 81 (3): 635-647. Herriges, J.A., and J.F. Shogren (1996), Starting Point Bias in Dichotomous Choice Valuation with Follow-Up Questioning, Journal of Environmental Economics and Management 30: 112-131. Hoffman, E.M. and M.L. Spitzer (1993), Willingness to Pay v.s Willingness to Accept: Legal and Economic Implications, Washington University Law Quarterly, 71 (1): 59-114. Hoffman, Elizabeth, Dale, J.M, Dipankar Chakravart, Ray A. Field and Gle D. Whilpple (1993), Using Laboratory Experimental Auction in Marketing Research: A Case Study of New Packaging for Fresh Beef, Marketing Science, 12 (3): 318-338. Huang, C.L., K. Kan, and T.-T. Fu (1999), A Generalized Binary-Ordinal Probit Model of Consumer Willingness-to-Pay for Food Safety in Taiwan, Journal of Consumer Affairs 33: 76-91. Kaneko, Naoya. and Wen S. Chern (2005), Willingness to Pay for Genetically Modified Oil, Cornflakes, and Salmon: Evidence from a U.S. Telephone Survey, Journal of Agricultural and Applied Economics 37: 701-719. Kaneko, Naoya, Wen S. Chern, and Gulay Babadogan Tarakcioglu (2005), Willingness to Pay for a New Orange Juice: An Application of Auction Experiment, submitted to Food Quality and Preference. Loureiro, M.L., J.J. McCluskey, and R.C. Mittelhammer. (2002), Will ConsumersPay a Premium for Eco-Labeled Apples?, Journal of Consumer Affairs 36: 203-219. Loureiro, M.L. and W.J. Umberger (2003), Estimating Consumer Willingness to Pay for Country-of-Origin Labeling, Journal of Agricultural and Resource Economics 28: 287-301. Lusk, Jayson L. and John D. Anderson (2004), Effect of country of origin labeling on meat producers and consumers, Journal of Agricultural and Resource Economics, 29(2):185-205. Schmitz, Andrew, Charles B.Moss and Troy G. Schmitz (2005), A Differential Examination of the effect of Mandatory Country of Origin Labeling on the Beef Sector, Paper presented at the AAEA meetings in Providence, Rhode Island. Schupp, A. and J. Gillespie (2001), Handler Reaction to Potential Compulsory Country-of-Origin Labeling of Fresh or Frozen Beef, Journal of Agricultural and Applied Economics 33: 161-171. Shogren, Jason F., Seung Y. Shin, Dermot J. Hayes and James B. Kliebenstein (1994), Resolving Differences in Willingness to Pay and Willingness to Accept, American Economoc Review 84: 255-270.
31

Umberger, Wendy J., Dillon M. Feuz, Chris R. Calkins, and Bethany M. Sitz (2003), Country-of-Origin Labeling of Beef Products: US Consumers’ Perceptions, AgDM Newsletter Article. U.S. Senate Farm Bill Conference Framework (2002), Farm Bill Conference Summary, http://www.senate.gov/~agriculture/Briefs/2001FarmBill/conframe.htm. Accessed February 29, 2008. Vickrey, W. (1961), Counterspeculation, Auctions, and Competitive Sealed Tenders, Journal of Finance 16:8-37. Wooldridge, Jeffrey M. (2006), Introductory Econometrics, 3 th ed. Thomson Higher Education.
台灣內政部內政統計資訊服務網 (2008),內政統計年報,2005 年人口年齡分配 http://www.moi.gov.tw/stat/ Accessed March 3, 2008.

32

Table 1. Survey Design and Sample Distribution of Discount Prices
Item Oolong Tea (150g/Package) Preserved Olives (150g/Package) Sample Distribution (Heads)
a b b

Base

a

A (10%off) $135 $18 14

B (20% off) $120 $16 15

C (30% off) $105 $14 15

D (40% off) $90 $12 15

E (50% off) $75 $10 15

$150 $20 74

Used for the initial question in the contingent valuation survey Shown as the number of participants

Table 2. Variable Definitions and Sample Mean from Experimental Auction and CVM
Variable Name Definition and coding Mean S.D Dependent Variables Experimental Auction The bids of respondents for the Taiwan preserved olives (NT$). 48.56 25.44 The bids of respondents for the China preserved olives (NT$). 31.46 23.07 The bids of respondents for the Taiwan oolong tea (NT$). 206.58 120.6 The bids of respondents for the China oolong tea (NT$). 118.54 86.78 The bids of respondents for the Vietnam oolong tea (NT$). 107.97 75.54 Contingent Valuation Method 1 if Taiwan preserved olives are chosen; 0.833 0.375 0 if China preserved olives are chosen. 1 if Taiwan oolong tea is chosen; 2 if China oolong tea is 1.077 0.553 chosen; 3 if Vietnam oolong tea is chosen. Independent Variables Experimental Auction Behavioral Intention (first question, given the same price) 1 if non-Taiwan products are chosen; 0 otherwise. 1 if Taiwan product is chosen; 0 otherwise. 1 if non-Taiwan and Taiwan products are equally good initially; 0 otherwise. Trail 1if the bid is the first bid of the respondent; 0 otherwise. 1 if the bid is the second bid of the respondent; 0 otherwise. 1 if the bid is the third bid of the respondent; 0 otherwise. Contingent Valuation Method Price Difference between price of Taiwan products and non-Taiwan products.

Ot Oc Tt Tc Tv Y (binomial) Y (multinomial)

NTW TW IND Trail1 Trail2 Trail3

△PRICE

33

Table 3. Common Independent Variables, Definition and Coding
Variable Name INFO1 INFO2 Risk Definition and Coding Information 1 if very well/somewhat informed about COOL; 0 otherwise. 1 if information is from the food package; 0 otherwise. 1 if imported contaminated food is the most important issue; 0 otherwise. Attitude and perception Frequency of preserved fruit 0 if never or rarely; 1 if once a purchasing. month, once in two weeks or if once a week. Frequency of tea purchasing. Frequency of organic food purchasing. 0 if never or rarely; Check the labeling of the food. 1 if sometimes, often or always. Brand. 1 if the characteristics of food is Price. considered to be very / Freshness. somewhat important for the Food Safety. respondents, Country-of-origin labeling. 0 if not very important, not Nutrition. Environmental consciousness. important or so-so. Convenience. Color. Calorie. 1 if the government’s regulatory performance is excellent or good; 0 otherwise. Demographic 1 if male; 0 otherwise. 1 if marriage; 0 if single. Age of respondent as of 2007. Number of kids under age 15. Number of persons per household. 1 if the respondent is a vegetarian; 0 otherwise. 1 if BMI≦24 (Normal or slim figure); 0 if otherwise. 1if Elementary school; 0 otherwise. 1 if Completed high school; 0 otherwise. 1 if >college; 0 otherwise. Monthly Income of each respondent ÷1000. 1 if Average income of the household in Year 2007≧680,000; 0 if otherwise. Monthly food expenditure at home per household÷1000. Expenditure of food away from home per household÷1000. 1 if respondent is from Taichung; 0 otherwise. 1 if respondent is from Taipei; 0 otherwise. 1 if respondent is from Kaohsiung; 0 otherwise. Survey versions 1 if respondents do the auction first and CVM later; 0 otherwise. Mean S.D

0.769 0.187 0.782

0.424 0.392 0.416

0.923 0.538 0.462 0.667 0.705 0.449 1 0.987 0.872 0.859 0.679 0.628 0.423 0.641 0.256

0.268 0.502 0.502 0.474 0.459 0.501 0 0.113 0.336 0.35 0.469 0.486 0.497 0.483 0.439

BUYPF BUYTEA BUYOG CHECK BRAND PRICE FRESH SAFETY COOL NUTRI ENVIR CONVEI COLOR Calorie GOV SEX MAR AGE KID NUM VEGE BMI EDU1 EDU2 EDU3 MINC INC FAH FAFH Zone1 Zone2 Zone3 SECTION

0.303 0.744 44.077 0.77 3.718 0.104 0.423 0.051 0.551 0.398 41.026 0.513 12.635 5.262 0.324 0.34 0.34 0.526

0.446 0.439 10.71 1.092 1.553 0.307 0.497 0.222 0.3 0.493 71.32 0.503 7.051 4.664 0.47 0.48 0.48 0.503

34

Table 4. Affiliation Test Results
Variable Preserved olives Taiwan Constant t P N -8.47*** (2.76) 1.75 (2.02) 0.17*** (0.04) 140 China -8.45*** (3.13) 3.55** (1.95) 0.13*** (0.04) 140 Taiwan -45.47** (20.1) 29.36** (14.32) 0.16*** (0.05) 140 Oolong tea China -51.08*** (11.54) 24.16*** (8.16) 0.2*** (0.04) 140 Vietnam -38.18*** (10.14) 1.84 (7.09) 0.2*** (0.05) 140

Notes: The symbols ***, **, and * indicate that the variables are significant at the 1%, 5%, and10% levels, respectively. Numbers in parentheses are estimated standard errors. Blank space indicates that the variable is not applicable. N is the number of observations.

Table 5. Unconditional Auction Results, Total Sample
Item COOL Mean Bid (NT$) S.D Max Bid (NT$) Number of Respondents Biding Zero Number of Irrational Respondents Sample Size
a a

Preserved Olives Taiwan 46.71 25.44 200 9(12%) 4 70 China 29.43 23.07 160 39(52%) 4 70

Oolong Tea Taiwan China 200.27 113.22 120.59 720 4 70 86.78 500 4 70 Vietnam 110.01 75.54 400 4 70

9(12%) 30(40%) 24(32%)

Participants who did not fill out their form carefully and completely. Therefore, we can not use their data.

35

Table 6. Survey Results of the Respondent’s Choice Behavior on Preserved Olives by Discount Level
Discount Level First Question Products Number of Respondents (percentage) 12(100%) 14(100%) 13(100%) 14(100%) 13(100%) 8 (TW,CH) Prices
a

Follow-up Products Number of Respondents (percentage) 12 (100%) 13(92.9%) 1(7.1%) 14(100%) 12(85.7%) 2(14.3%) 11(84.6%) 2(15.4%) -

Taiwan A (10% discount) B (20% discount) C (30% discount) D (40% discount) E (50% discount) All
a

(20,18) (18,20) (20,16) (16,20) (20,14) (20,12) (12,20) (20,10) -

China Taiwan China Taiwan China Taiwan China Taiwan China Neither

Taiwan China Taiwan China Taiwan China Taiwan China Taiwan China Taiwan China Taiwan China Taiwan China Taiwan China Taiwan China

TW and CH represent Taiwan and China.

36

Table 7. Survey Results of the Respondent’s Choice Behavior on Oolong Teas by Discount Level
Discount Level First Question Products
a b

Follow-up (TW, CH, V) Prices Products Number of Respondents (percentage) 12(85.8%) 1(7.1%) 1(7.1%) 14(100%) 14(93.4%) 1(6.6%) 10(76.9%) 2(15.4%) 1(7.7%) 12(92.3%) 1(7.7%)

Number of respondents (percentage)

A (10% discount) B (20% discount) C (30% discount) D (40% discount) E (50% discount) All All
a

Taiwan

14(100%)

(150, 135,135) (150, 120,120)

Taiwan

14(100%)

Taiwan China Vietnam Taiwan China Vietnam China Vietnam Taiwan China Vietnam Taiwan China Taiwan China Vietnam

Taiwan

15(100%)

(150, 105,105)

Taiwan China

12(92.3%) 1(7.7%)

(150, 90,90) (90, 150, 90)

Taiwan Neither indifference

13(100%) 5 0

(150, 75, 75)

Since almost all selected Taiwan product for the first question, the other products are not listed except for 40%discount case. TW, CH and V represent Taiwan, China and Vietnam.

b

37

Table 8. Regression Results on Bids and Premiums for Preserved Olives and Oolong Teas, All Trails, Experimental Auction
Variable Constant Preserved Olives
Taiwan (TW) China (CH) Premium (TW-CH) Taiwan (TW) China (CH)

Oolong Tea
Vietnam (V) Premium Premium (TW-CH) (TW-V) -115.55** -109.37*

69.88*** (15.27) INFO2 8.7** (4.02) RISK -7.84* (4.56) BUYPF 2.62 (5.18) BUYTEA -3.2 (3.63) BUYOG 1.31 (4.53) CHECK -14.24*** (4.57) BRAND -5.94** (3.49) NUTRI 1.78 (4.5) ENVIR 11.19*** (3.93) COLOR -11.96*** (3.76) CALORIE 2.88 (3.36) GOV 9.42** (4.55) SECTION 12.02*** (3.03) TRIAL1 -5.51* (3.12) TRAIL3 2.83 (3.04)

58.62*** (15.64) 0.63 (4.04) -16.11*** (4.64) -11.34** (5.12) -3.10 (3.83) 5.94 (4.65) -21.13*** (4.7) -1.35 (3.47) -4.802 (4.66) 11.02*** (4.21) -3.25 (3.97) -7.24** (3.58) 2.31 (5.05) 13.91*** (3.09) -3.27 (3.2) 2.38 (3.18)

23.86** (10.3) 7.89*** (2.72) 4.86* (2.92) 11.10*** (3.8) 0.3 (2.52) -1.37 (3.12) 3.56 (2.98) -6.06** (2.48) 3.36 (3.25) 2.47 (2.81) -7.76*** (2.66) 6.01** (2.36) 5.55* (3.12) 0.031 (2.11) -2.16 (2.21) 0.91 (0.68)

-105.21 (73.17) 45.5** (18.85) 49.25** (22.56) 23.93 (25.13) -47.7*** (17.89) -35.82 (23.61) -11.46 (21.84) -10.18 (17.83) -27.68 (21.34) -6.05 (19.09) -16.23 (18.89) 46.35*** (16.55) -31.01 (23.76) 86.95*** (15.24) -19.9 (16.06) 8.36 (15)

18.77 (58.81) 3.8 (14.65) 23.79 (16.9) -47.38** (19.18) -16.45 (14.11) -28.64 (19.14) -31.94* (17.32) 3.38 (13.45) -10.75 (17.93) 30.28* (15.72) 19.45 (14.73) 24.05** (12.74) -1.89 (18.92) 67.94*** (11.63) -17.22 (12.32) -1.7 (11.68)

44.84 (47.74) 15.77 (12.49) -3.24 (13.78) -41.19*** (15.82) -17.45 (11.43) 3.46 (14.92) -32.18** (13.8) 11.74 (11.07) -35.02** (14.97) -4.613 (12.79) 4.87 (11.87) 29.56*** (10.66) 3.7 (15.77) 49.66*** (9.51) -9.16 (10.21) 5.46 (9.61)

(51.24) 40.24*** (13.52) 39.78*** (14.54) 60.42*** (18.87) -28.18** (12.56) -3.05 (15.54) 18.46 (14.82) -20.97* (12.34) -16.7 (16.16) -31.01** (14.01) -21.87 (13.26) 24.71** (11.77) -30.56* (15.52) 22.67** (10.5) -6.07 (10.98) 14.29 (10.98)

(57.45) 16.13 (15.16) 56.04*** (16.3) 57.89*** (21.16) -21.42 (14.08) -22.05 (17.42) 11.22 (16.61) -25.57** (13.83) 10.24 (18.12) 0.17 (15.71) -13.44 (14.87) 21.4 (13.19) -32.63* (17.4) 33.99*** (11.78) -13.06 (12.31) 7.63 (12.31)

Notes: The symbols ***, **, and * indicate that the variables are significant at the 1%, 5%, and10% levels, respectively. Numbers in parentheses are estimated standard errors. Blank space indicates that the variable is not applicable. N is the number of observations.

38

Table 8. Regression Results on Bids and Premiums for Preserved Olives and Oolong Teas, All Trails, Experimental Auction (Continued)
Variable AGE Preserved Olives
Taiwan (TW) China (CH) Premium (TW-CH) Taiwan (TW) China (CH)

Oolong Teas
Vietnam (V) Premium Premium (TW-CH) (TW-V)

-0.53** (0.23 SEX -2.03 (3.3) MAR 1.35 (5.11) EDU3 11.39*** (4.42) KID -5.34 (2.06) NUM -1.54 (1.45) FAH 0.88*** (0.26) FAFH -0.26 (0.37) VEGE 6.86 (6.2) BMI 11.74*** (3.54) INC -16.24*** (3.52) ZONE2 -0.257 (6.48) ZONE3 1.02 (4.12) σ 25.46 (20.33) N 210

-0.17 (0.23) 8.52** (3.44) 2.19 (5.28) -1.02 (4.71) -6.04*** (2.06) 2.84 (1.58) 0.068 (0.28) -0.08 (0.38) -0.45 (6.38) 12.47*** (3.45) -10.94*** (3.58) -7.53 (6.94) 4.22 (4.4) 23.02 (20.24) 210

-0.55*** (0.16) -8.43*** (2.36) 6.45** (3.21) 10.31*** (3.13) 0.11 (1.401) -4.31*** (1.04) 0.79*** (0.18) -0.23 (0.26) 4.76 (4.39) -0.46 (2.39) -4.99** (2.47) 3.92 (4.58) -3.5 (2.8) 210

4.11*** 0.99 1.49** 2.72*** 2.03** (1.07) (0.85) (0.69) (0.78) (0.88) -10.17 29.89** 18.11 -30.57*** -24.89* (16.39) (14.09) (11.44) (11.74) (13.16) 74.46*** 12.58 21 78.16*** 65.18*** (23.16) (18.44) (15.6) (15.97) (17.9) 73.41*** 19.42 30.19** 47.26*** 36.71** (21.95) (17.45) (14.13) (15.56) (17.44) -14.13 -18.97** -6.24 4.22 -5.52 (10.55) (8.58) (6.61) (6.97) (7.82) 2.01 10.79* 9.3** -13.46** -11.95** (7.57) (5.6) (4.61) (5.18) (5.87) -5.69*** -5.49*** -4.56*** -1.25 -1.64 (1.29) (1.12) (0.93) (0.92) (1.03) 4.65** 3.79** 1.76 2.01 2.97** (1.89) (1.54) (1.2) (1.301) (1.46) 70.99** 1.55 -11.36 64.08*** 71.83*** (31.24) (26.24) (21.46) (21.85) (24.5)13.05 15.71 31.08*** -0.6 -19.14 (17.73) (13.01) (10.85) (11.87) (13.31) 16.15 -9.39 -22.53** 31.9** 43.89*** (17.5) (13.65) (10.86) (12.29) (13.78) 66.90** 84.36*** 37.28* -23.16 -1.4 (33.96) (26.67) (20.73) (22.81) (25.57) 50.83** 60.58*** 30.18** 2.30 18.64 (20.2) (16.44) (13.85) (13.95) (15.64) 122.03 87.63 75.85 (103.95) (73.1) (62.87) 210 210 210 210 210

Notes: The symbols ***, **, and * indicate that the variables are significant at the 1%, 5%, and10% levels, respectively. Numbers in parentheses are estimated standard errors. Blank space indicates that the variable is not applicable. N is the number of observations.

39

Table 9. Regression Results on Bids and Premiums for Preserved Olives and Oolong Teas, First Trail, Experimental Auction
Variable
Constant INFO2 RISK BUYPF BUYTEA BUYOG CHECK BRAND NUTRI ENVIR COLOR CALORIE GOV AGE SEX MAR EDU3 KID NUM FAH FAFH VEGE BMI INC SECTION ZONE2 ZONE3 σ

Preserved Olives
Taiwan (TW) 96.62*** (25.69) 11.23 (7.72) -10.89 (7.76) -2.54 (9.39) -5.55 (6.59) 2.28 (8.26) -22.26*** (7.44) -2.55 (6.13) 5.11 (7.96) 12.31* (6.98) -14.82** (6.78) -1.32 (5.81) 8 (7.86) China (CH) 67.43*** (26.05) 4.17 (7.36) -16.94** (7.62) -11.51 (8.6) -4.18 (6.42) 7.12 (8.09) -25.24*** (7.8) 0.68 (5.98) 2.55 (7.84) 9.43 (7.23) -4.44 (6.6) -12.4** (6) -5.35 (8.17) Premium (TW-CH) 32.29* (18.71) 8.24 (4.98) 4.35 (5.35) 6.93 (6.94) -1.65 (4.62) -1.84 (5.72) 0.7 (5.45) -4.77 (4.54) 1.45 (5.95) 5.24 (5.15) -8.4* (4.88) 6.15 (4.33) 6.25 (5.71) Taiwan (TW) -79.18 (134.21) 19.15 (35.72) 18.99 (41.89) 41.46 (47.47) -32 (34.59) -8.09 (43.82) -4.99 (42.35) 17.98 (34.38) -17.89 (40.57) -22.52 (37.26) -40.94 (36.99) 65.24** (31.71) -34.08 (49.17) China (CH) 63.7 (110.85) 5.43 (28.23) 10.59 (32.1) -25.76 (36.77) -17.68 (29.55) -20.43 (34.8) -35.46 (34.26) 11.97 (26.82) -25.79 (36.42) 32.18 (29.7) -12.34 (30.31) 25.75 (26.18) 1.19 (40.11)

Oolong Teas
Vietnam (V) 74.99 (87.85) 17.84 (23.56) -18.49 (26.42) -21.07 (29.73) -19.73 (23.3) 8.35 (27.6) -36.25 (26.61) 27.03 (21.55) -42.87 (28.66) -12.93 (24.97) -8.46 (23.16) 27.37 (20.66) 5.38 (32.61) Premium (TW-CH) -128.66 (91.76) 28.64 (24.4) 22.12 (26.24) 63.34* (34.06) -20.89 (22.66) 9.66 (28.04) 25.74 (26.74) -8.13 (22.26) -4.94 (29.17) -57.71** (25.28) -20.92 (23.93) 28.52 (21.23) -27.9 (28) Premium (TW-V) -130.94 (99.56) -3.62 (26.48) 38.53 (28.47) 56.16 (36.95) -11.64 (24.59) -13.02 (30.42) 25.4 (29.01) -13.35 (24.16) 25.13 (31.64) -6.15 (27.43) -24.24 (25.96) 25.78 (23.04) -34.86 (30.38)

-0.91** (0.39) -2.76 (5.73) 7.74 (9.37) 12.39 (7.61) -3.89 (3.57) -2.19 (2.53) 0.79* (0.44) -0.29 (0.65) 0.77 (10.89) 12.87** (6.12) -18.71*** (6.1) 0.78 (5.27) -5.01 (11.48) 4.74 (0.53) 26.86 (25.98)

-0.48 (0.38) 7.48 (5.63) 11.51 (8.84) 2.27 (7.83) -3.99 (3.43) 1.25 (2.7) 0.23 (0.45) -0.07 (0.65) -9.61 (10.22) 17.86*** (5.77) -14.89** (6.07) 5.1 (5.19) -8.23 (11.58) 11.47 (7.42) 23.45 (25.24)

-0.62** (0.29) -7.41* (4.32) 7.06 (5.88) 10.17* (5.73) 0.18 (2.57) -3.94** (1.91) 0.61* (0.34) -0.25 (0.48) 6.87 (8.07) -1.85 (4.37) -4.6 (4.52) -1.93 (3.87) 3.13 (8.4) -2.4 (5.13) -

3.17 (2.04) -13.57 (31.59) 86.44* (45.95) 64.7 (43.49) 1.82 (20.89) 2.34 (14.09) -3.77 (2.37) 2.37 (3.62) 76.58 (68.82) 35.4 (33.33) 14.68 (33.2) 27.87 (28.99) -21.59 (65.78) -3.78 (40.09) 118.07 (127.8)

0.58 (1.65) 21.57 (27.75) 23.53 (38.29) 11.41 (35.77) -12.41 (17.23) 5.16 (11.17) -3.79* (2.18) 1.89 (2.91) 12.13 (55.4) 17.72 (26.54) -15.45 (28.14) 48.39** (22.92) 33.7 (52.05) 41.86 (34.04) 90.3 (101.3)

1.5 (1.33) 13.01 (22.05) 32.46 (31.42) 32.13 (28.74) 1.8 (13.15) 8.14 (8.71) -4.39** (1.71) -0.02 (2.33) 2.4 (46.79) 25.88 (21.08) -25.27 (21.17) 24.68 (18.36) -8.49 (39.92) 5.75 (26.96) 77.79 (86.56)

3.13** (1.41) -27.29 (21.18) 64.19** (28.82) 33.22 (28.08) 8.87 (12.58) -11.11 (9.35) -0.077 (1.66) 1.91 (2.35) 79.84** (39.44) 7.93 (21.43) 37.02 (22.19) -8.05 (18.96) -51.39 (41.16) -32.37 (25.17) -

1.82 (1.53) -19.38 (22.98) 58.59* (31.27) 28.85 (30.46) 0.05 (13.65) -9.91 (10.15) -0.03 (1.8) 2.88 (2.55) 71.63 (42.79) -4.84 (23.25) 44.34* (24.07) 9.31 (20.57) -25.71 (44.66) -8.88 (27.31) -

70 70 70 70 70 70 70 70 N Notes: The symbols ***, **, and * indicate that the variables are significant at the 1%, 5%, and10% levels, respectively. Numbers in parentheses are estimated standard errors. Blank space indicates that the variable is not applicable. N is the number of observations.

40

Table 10. Marginal Effect for Olives and Teas, All Trails, Experimental Auction
Variable INFO2 RISK BUYPF BUYTEA BUYOG CHECK BRAND NUTRI ENVIR COLOR CALORIE GOV AGE SEX MAR EDU3 KID NUM FAH FAFH VEGE BMI INC SECTION ZONE2 ZONE3 TRAIL1 TRAIL3 Unit of measurement 1 or 0 1 or 0 1 or 0 1 or 0 1 or 0 1 or 0 1 or 0 1 or 0 1 or 0 1 or 0 1 or 0 1 or 0 Year old 1 or 0 1 or 0 1 or 0 Head Head NT$ NT$ 1 or 0 1 or 0 NT$ 1 or 0 1 or 0 1 or 0 1 or 0 1 or 0 Preserved Olives Taiwan (TW) China (CH) 7.4 -0.99 -1.69 -6.48 2.86 -0.61 -6.42 -4.62 4.76 2.81 -10.53 -1.75 4.70 -0.51 -1.73 -5.69 8.10 -5.18 -1.49 0.85 -0.25 5.37 11.31 -3.58 12.64 1.13 2.02 -5.66 4.62 -8.59 -11.27 1.21 3.39 -9.37 -3.11 4.81 2.24 -1.72 -3.18 -0.63 -0.16 2.94 -0.95 -0.96 -5.76 2.71 0.06 -0.08 -0.95 6.43 -1.70 11.56 4.85 0.01 -2.97 2.72 Oolong Teas Taiwan (TW) 3.69 14.94 -8.48 -18.1 17.37 -35.38 -5.78 2.47 -1.36 -24.72 27.9 -55.65 3.52 -18.10 33.29 64.1 -12.10 1.72 -4.87 3.98 6.67 9.28 10.35 57.30 21.39 23.76 -18.40 14.13 China (CH) -5.31 -4.73 -37.11 4.74 21.29 -47.66 -2.76 19.04 9.44 0.91 16.82 -7.57 0.93 3.03 3.45 8.74 -17.80 10.12 -5.15 3.56 -32.50 -0.60 -12.10 46.13 37.40 8.13 -10.98 4.61 Vietnam (V) -0.81 -6.05 -40.44 6.98 24.79 -35.39 -5.25 -1.89 -8.64 0.18 12.81 5.15 1.30 -1.11 6 16 -5.45 8.12 3.98 1.54 -21.56 11.41 -15.19 41.05 32.14 3.42 -8.20 7.00

Table 11. Expected Willingness to Pay for Olives and Teas, Experimental Auction
Model

Item
Taiwan

Preserved olives
China Premium (TW-CH) Taiwan China

Oolong teas
Vietnam Premium (TW-CH) Premium (TW-V)

All Trails First Trail

Mean (NT$) S. D Mean (NT$) S. D

46.81 16.23 43.74 16.04

29.61 13.31 27.42 13.18

19.78 9.99 18.02 9.52

205.15 69.49 188.88 58.05

115.15 53.13 108.28 39.37

103.5 45.23 97.56 35.70

98.97 47.95 92.00 53.01

109.07 45.22 97.9 45.28

41

Table12. Parameter Estimates for Preserved Olives and Oolong Teas, CVM
Variables Intercept RISK PRICE BRAND CHECK COOL CALORIE MAR AGE FAFH MINC △Price N -0.67** (0.31) 69 -1.85** (0.93) -0.24** (0.09) -0.66** (0.27) 2.56** (0.87) 1.98** (0.76) The multinomial Logit regression can not converge due to the number of switching choices under different price scenarios (the max discount of 50%) being too few. Preserved Olives (Logit) TW v.s CH 3.89 (3.41) Estimated Coefficients Oolong teas (Multinomial Logit) TW v.s CH TW v.s Vietnam

65

Notes: The symbols ***, **, and * indicate that the variables are significant at the 1%, 5%, and10% levels, respectively. Numbers in parentheses are estimated standard errors. Blank space indicates that the variable is not applicable. N is the number of respondents. TW, CH and V represent Taiwan, China and Vietnam.

Table 13. Estimated Premiums for Taiwan Products over Alternatives
CVM Item Logit or Multinomial Logit 13.4 (6.34) 67% 69 CH > 75 50%~ ∞
c

Auction
Unconditional
a

Tobit

OLS

Preserved olives Mean (Std. Dev) % Premium N
b

17.28 (16.48) 58.35% 70

17.2 (9.33) 58.1% 70

19.78 (9.99) 67.5% 70 CH 98.97 (47.95) 87.4% 70

d

Mean (Std. Dev) % Premium N
a b

V > 75 50%~∞ 65

CH 87.32 (80.81) 77.12% 70

V 90.26 (84.46) 82.05%

Oolong teas CH V 90 101.61 (39.8) (35.43) 78.15% 98.13% 70 70

V 109.07 (45.22) 99% 70

The unconditional auction mean is the mean computed from the raw auction bids. N is the number of respondents.

42

c

Percentage premium of Contingent Valuation is computed by dividing the point of willingness to pay by the base prices in the questionnaire. Percentage premium of OLS is computed by dividing the average raw auction bid for non-Taiwan product.

d

80.0% 60.0% 40.0% 20.0% 0.0%
Single

72.8%

19.1% 6.7%
Married Divorce

1.4%
Companionate marriage

0.0%
Separation

Figure 1. Distribution of Marital Status, Total Sample
80.0% 60.0% 40.0% 20.0% 0.0% Taipei Taichung Kaohsiung 70.8% 37.5%

60.0%

Figure 2. Proportions of Overweight Respondents by City

30.00% 25.00% 20.00% 15.00% 10.00% 5.00% 0.00%

25.93% 26.03% 26.03% 12.40% 5.45% 0.00% 1.33% 1.40% 0.00% 1.40%

Figure 3. Distribution of the Household Average Annual Income, 2007, Total Sample
NT$35,000 NT$30,000 NT$25,000 NT$20,000 NT$15,000 NT$10,000 NT$5,000 NT$0 NT$29,560 NT$25,041 NT$20,791

les

st

Figure 4. Comparison of Average Monthly Personal Income by City

NT $3 40 00 34 0 00 00 -6 80 68 00 00 0 00 -1 02 00 10 20 00 00 013 60 13 00 60 0 00 017 00 17 00 00 0 00 020 40 20 00 40 0 00 023 80 23 00 80 0 00 027 20 27 00 20 0 00 034 00 34 00 00 0 00 050 00 00 0
Taipei Taichung Kaohsiung

ha n

43

NT$15,000 NT$10,000 NT$5,000 NT$0

NT$12,125

NT$12,729

NT$14,041

NT$12,965 FAH FAFH

NT$4,604

NT$4,887

NT$6,541

NT$5,344

Taipei

Taichung

Kaohsiung

Average

Figure 5. Comparison of FAH and FAFH by City

60.0% 50.0% 40.0% 30.0% 20.0% 10.0% 0.0%

48.5%

5.0% Elementary school

8.0%

12.3%

19.0%

14.0%

Some high school

Completed high school

College

University

Graduated school

Figure 6. Distribution of Education Levels, Total Sample

60.0% 50.0% 40.0% 30.0% 20.0% 10.0% 0.0%

51.3% 23.0%

20.2% 1.4% 1.3% 2.7%

Every day

several times a Once a week week

Once in two weeks

Once a month Once in two months or even longer

Figure 7. Frequency of Food Purchasing
74.4%

80.0% 70.0% 60.0% 50.0% 40.0% 30.0% 20.0% 10.0% 0.0%

17.5% 1.0% Once a few day 0.0% Once a week 4.1% Once in two weeks Once a month Rarely 2.7% Never

Figure 8. Frequency of Preserved Food Purchasing
44

50.0% 40.0% 30.0% 20.0% 10.0% 0.0%

47.2%

20.4% 8.3%

20.1% 3.0%

Once aweek or two weeks

Once a month

once in two months or even longer

Rarely

Never

Figure 9. Frequency of Oolong Tea Purchasing

80% 60% 40% 20% 0%

76% 50%

76%

Taipei

Taichung

Kaosiung

Figure 10. Frequency of Organic Food Purchasing

10. Frequency of Organic Food Purchasing 120.0%
100.0% 80.0% 60.0% 40.0% 20.0% 0.0% 35.1% 23.0% 94.3% 98.7% 54.1% 62.1% 41.9% 27.1% 24.3% 34.9%

CO OL En vi ro N nm ut rit en io ta n lc on sc io us ne ss Co nv en ie nc e

Figure 11. Distribution of the Importance of Food Attributes of Total Sample

Fr

es hn es s

45

Ca lo rie

Pr ice

Sa fe ty

Br an d

Co lo r

40.0% 30.0% 20.0% 10.0% 0.0%

37.1%

37.1%

13.6%

12.2%

Food Safety

Freshness

High quility

Patriotism

Figure 12. Reasons for Choosing Food Labeled Taiwan
50.0% 45.0% 40.0% 35.0% 30.0% 25.0% 20.0% 15.0% 10.0% 5.0% 0.0% 44.7% 32.4% 13.4% 2.7% Preservative problems in foods Pesticides on fruits and vegetables 2.7% 0.0% Too much pigment on Foods 4.7% Food labeling is not clear enough

Hygiene GeneticallyImported contaminated modified (GM) problems on food away from foods food home

Figure 13. Distribution of What is the Most Important Food Safety Problem in Taiwan Currently, Total Sample 31.07% 35.00%
30.00% 25.00% 20.00% 15.00% 10.00% 5.00% 0.00%
ap er s/M ag az in es /B

18.13%

20.07% 12.63% 1.40%

19.67%

0.00%
an ck ag es ou rc es Ot h er S it i D iet

oo ks

ad io

am i ly

/R

nd s/ F

TV

In t

er ne t

ni st /

Figure 14. Source of Information on Food, Total Sample

Ne ws p

80.0% 60.0% 40.0% 20.0% 0.0% Very Well 5.4%

72.7%

Do cto r /N ut r it io

24.3%

Somewhat

Not at All

Figure 15. How Well You Are Informed About Country-Of-Origin Law in Taiwan?
46

Fo od

Fr ie

Pa

90% 80% 70% 60% 50% 40% 30% 20% 10% 0%

77% 66% 66%

83% 75% Don't like it Never eat it Doubt of the food safety Don't Know

22% 12% 0% Taiwan olives 13% 5% 5% China olives

22% 12.50% 12% 10% 7% 4.50%8% 0% 0% Taiwan tea China tea Vietnam tea

Figure 16. Reasons for Bidding Zero

100.0% 80.0% 60.0% 40.0% 20.0% 0.0%

89.2%

91.9% Taiwan China Vietnam Neither

10.8% 0.0% 0.0% 0.0% Olives

1.4% 0.0% Tea

6.7%

0.0%

Indifference

Figure 17. Initial CV Response at Equal Price

60% 50% 40% 30% 20% 10% 0%

50%

50% 34% 25% 25% 16% 0 0 Both Too Don't Know Expensive Olives Tea

Don't Buy Never Eat It the Product

Figure 18. Reasons for Choosing Neither of the Products

47

Sign up to vote on this title
UsefulNot useful