PaPer Ital. J. Food Sci., vol. 28 - 2016 107 - Keywords: dichotomous contingent valuation, food safety, organic potatoes, probit model, reduced pesticides - Consumer fair priCes for less pestiCide in potato C. serefoglu1* and s. serefoglu2 1Ankara Development Agency, Aşağı Öveçler Mah. 1322. Cad. No. 11, 06460 Çankaya, Ankara, Turkey 2The Ministry of Defence of Turkey, 06100, Bakanlıklar, Ankara, Turkey *Corresponding author: cserefoglu@gmail.com AbstrAct this study estimates turkish citizens’ willingness to pay (WtP) for reduced pesticides on pota- toes. these estimates rely on data collected from 393 persons covering all regions in turkey through an online survey during the period from June 22 - July 21, 2014. the average WtP was found to be about tL 1.68 for all observations including zero bids and tL 2.91 excluding zero bids. the re- sults of the probit model show that cosmetic defects, free-pesticide potatoes with insect damages, age, and gender were identified by the model to have significant impacts on the probability of WtP. mailto:cserefoglu%40gmail.com?subject= 108 Ital. J. Food Sci., vol. 28 - 2016 INtrODUctION Pesticides are defined by the European com- mission (Ec) (2009) as substances or mixtures of substances including chemical compounds intended for killing, destroying, or mitigating any pest. the use of pesticides has tragically and rapidly increased since 1960’s due to the green revolution (cArvALhO, 2006). As explained by hOPPIN et al. (2007), pesticides could cause some respiratory diseases to farmers. similarly, ALAvANJA et al. (2004) stated that indirect ex- posures which occur by way of drinking water, food or air happen more frequently than direct exposures occurring to individuals who apply pesticides in agriculture. the consumption level of pesticides in tur- key increased to 54,000 tonnes in 2002 but dur- ing the last decade the level notably decreased to 40,000 tonnes (MFAL, 2012). the amount of pesticides used in turkey seems quite low when compared with countries such as Germany and France in Europe according to the FAO statistics. stated and revealed preferences are the meth- ods that are often used to measure the WtP of consumers. As stated by EbErLE and hAyDEN (1991), each individual`s valuation of a non- market good is reflected through a direct ques- tionnaire approach. thus, our research is main- ly based on the contingent valuation Meth- od (cvM) and food safety issues through the responses which come from an online survey which covers the whole of turkey. the food safe- ty issue plays a crucial role for both policy mak- ers and consumers, with fast dissemination of information through social network. As under- lined by rOWELL (2004), food safety and sustain- able food supply are on the agenda of developed countries to develop diets that are fundamental- ly affordable and health-enhancing. the overall objective of this study is to assess turkish consumers’ attitudes towards purchas- ing reduced pesticides that are guaranteed not to be risky to human health. the specific objec- tives are: determine consumers’ attitudes and concerns toward pesticide use in potatoes and ascertain consumers’ willingness to pay high- er bid amounts for reduced pesticides in pota- toes by ensuring no pesticide residues, and es- timate consumers’ mean WtP for reduced pes- ticides potatoes. there are several reasons why the potato prod- uct is chosen. Firstly, potato is one of the most consumed vegetables in turkey and its consump- tion increases yearly even if its price increases, this is according to the data extracted from the database of the turkish statistic Institute (tUrK- stAt). second, it is a traditional food that has a wide usage with different vegetables. Last but not least is the over-use of pesticides used on po- tatoes and pesticide residues in it (bIrINcI and UzUNDUMLU, 2009; AyAz and yUrttAGUL, 2008). Following the Introduction, methodology will be covered in detail in section 1. Within the framework of the methodology, there is discus- sion of: sample size, and data analysis covering both questionnaire design and descriptive sta- tistics including the socio-economic character- istics of the respondents and consumer prefer- ences with respect to health risks and why the cvM is used. the second section will compre- hensively focus on the econometric results and their interpretations. regarding the economet- ric results, descriptive statistics and regression analysis will be included, and also, the assump- tions made to perform the study is included. the paper ends with a brief conclusion. MAtErIALs AND MEthODs the online survey as mentioned in the pre- vious sections randomly covered all of turkey through the social network. surveys’ results in table 1 clearly demonstrate that the rate of par- ticipation in survey in the North East region (NE) is proportionally higher than other regions while some regions such as Aegean region (AEG) and the south East region (sE) has a lower partici- pation rate considering their population. A high rate of responses in some regions might be ex- plained with a fast spreading of surveys linked with the help of respondents. the analysis was based on applying the cvM that is defined as “any approach to valuation of a commodity which relies upon individual respons- es to contingent circumstances posited in an arti- ficially structured market” (sELLEr et al., 1985). this method was first proposed by cIrIAcy-WAN- trUP in 1947 in order to estimate the benefits of the prevention of soil erosion (KONtOLEON et al., 2005; cAMErON, 1992). the cvM, which is basi- cally based on a survey-based methodology for elic- iting consumers’ valuations of non-market goods and services, has been widely applied by research- ers and policy makers in health economics and food safety for several decades and received con- siderable attention in the literature. It was stated by JEAN et al. (1995) that benefit estimates that are comparable to estimates from market-based approach can be produced by the cvM. there are a number of studies which have been used in surveys with discrete answers that have been an- alysed with logit and probit techniques (bUzby et al., 1995; AKGUNGOr et al., 2001; GArMING and WAIbEL, 2006; KALOGErAs et al., 2009). Determining sample size the sample size is defined by considering the current turkish population and calculated ac- cording to the formula provided by FINK (2003): Ital. J. Food Sci., vol. 28 - 2016 109 Where n is the sample size determined, N is the population size, p is level of precision. the sample size is 400 at 95% confidence level and a 5% margin of error. but 393 samples were used after the first elimination due to the in- completeness. Survey and data generation before moving further through online survey, the first draft was shared with 10 turkish con- sumers by using face to face interview method in order that the perspective of a consumer side is truly reflected in the format of questions. Af- ter receiving some positive and negative feed- back, the questionnaire form was finally rear- ranged in a short and clearer way as the first draft shared with consumers was found slight- ly longer and unclear instructions. Particular- ly, open-ended questions were not preferred by these consumers. Instead, options were in- cluded in some of the questions. Also, the an- swer choices were re-organized according to the consumer`s expectations. Following pre-test with turkish consumers, the link to the online survey was shared with turkish consumers via the social networks such as in general e-mails, Facebook, Linked- In and forums, and in particular regional de- velopment agency network covering all turkey for one month as from June 22 until July 21, 2014. the survey mainly comprised of three parts. the first part covered the questions to elicit perceptions that are related to pesticide residues. the consumers were asked about their perceptions of pesticide residues in pota- toes as well as the cosmetic defects. the ques- tion on cosmetic defects intended to measure whether or not the consumers would be willing to purchase fresh produce with insect damage, such as worm holes or irregular shape of the potatoes. the second part included WtP ques- tions. the survey asked consumers the maxi- mum WtP for reduced pesticide residues in po- tatoes. socioeconomic questions were inserted in the third part. For simplicity, the survey was designed to simulate consumers’ potato pur- chasing behaviour of their respective house- holds under alternative prices on reduced pes- ticides in potatoes. the scenario was built on the consumers that were provided with a label that guarantees that the potatoes were tested and certified that they do not contain pesticide residues harmful to human health by assum- ing no change in quality. by doing so, we were able to see if the consumer`s WtP is enough to justify these increased costs of production with a reduction in pesticide use. Regression Models of the CVM Probit and logit which are known as non-lin- ear functions of unknown coefficients in liter- ature are widely applied in binary choice mod- els. though both models may give similar re- sults, there are slight differences because of table 1 - Percentages of survey and turkish Population at the level of NUts 1 regions (source: Primary data extracted from survey). 110 Ital. J. Food Sci., vol. 28 - 2016 the tail of observations. AMEMIyA (1981) ex- pressed that the samples with heavier tails are more appropriate for logit models. A sim- ilar stance was made by cAKMAKyAPAN and GOKtAs (2013). they observed that logit mod- el is generally preferred for large sample sizes (500 and 1000) and probit model is usually for smaller sample sizes. so, probit model will ul- timately be employed for estimations because of the sample size. Alternatively, tobit model will be applied to measure WtP amounts that are obtained through single bounded dichoto- mous questions since the endogenous variable includes zero values. Probit model the Probit model is defined by WOOLDrIDGE (2006) as: zn=Xnβ+un. Where β is a vector of parameters including the intercept term; xn is a vector of covariates; u is the error term which either has the stand- ard logistic distribution or the standard normal distribution. In either case, u is symmetrically distributed about zero. zn is the unobservable amount that respondents are willing to pay for the reduced pesticides in potatoes. WtPi is the observed dichotomous variable stating whether the individual pays or not. It can be defined as follow: WtPn=0 if WtPn*≤0 WtPn=1 if WtPn*>0 As it is indicated by WOOLDrIDGE (2006), the main goal in binary responses is to explain the effects of x on the response that follows the probability P(y=1|x). P(WtP=1|x)=P(WtPn*>0|x)=P[e>-(β0+xβ|x]= =1-G[-(β0+xβ0]=G(β0+xβ). the direction of the effect of xj on E(WtP*|x)= β0+xβ and on E(WtP|x)=P(y=1|x)=G(β0+x β) is similar to each other. It is not possible to apply OLs due to the non- linear nature of E(y|x). Maximum likelihood methods thus must be used in order to esti- mate limited dependent variable models. the maximum likelihood can be written as follows (WOOLDrIDGE, 2006); ƒ(WtP|xi;β)=[G(xiβ)]y[1-G(xiβ)]1-y, WtP-0,1, It can easily be seen that when y=1 results in G(x, β) and when y=0, we get 1- G(xiβ). the func- tion of log likelihood for observation is a function of the parameters and the data (xi, yi) li(β)=WtPilog[G(xiβ)]+(1-WtPi)log[1-G(xiβ)]. Tobit model the general formulation of the tobit model can be expressed in the following way (GrEENE, 2000; WOOLDrIDGE, 2006); WtPn* =Xiβ+ui. WtP=0 if yn* ≤0. WtP=WtP* if WtPn*<0 E[WtPn*|xnβ] is xn`β. Wherefore, the nth in- dividual, Xn is a vector of explanatory variables, ui is a random disturbance term, and β is a pa- rameter vector common for each individual. by assuming the random error is independent and normally distributed among respondents, the expected WtP for an observation drawn at ran- dom from the population is E[WtP|xn]= ϕ(Xn`β/σ)+ xn`β+σλn) Where ϕ (Xn`β/σ)/Φ(Xn`β/σ); Where ϕ represents the normal distribution function and σ represents the standard devi- ation. Moreover, the expected value of WtP for observations above zero, which will be called E(WtP*), is simply Xβ plus the expected value of the truncated normal error terms. the expected WtP can be expressed as E(WtP)= ϕ(Xβ/σ)E(WtP*) WOOLDrIDGE (2006) points out that the func- tion of the tobit model which is based on max- imum likelihood estimation can be shown as: Ln L (β, σ)= (WtPn=0) ln[1-G(xn β/σ)]+(WtPn>0)ln{(1/σ)g [(WtPn-xn β)/σ]} Where G(.) is the standard normal cumulative distribution function; g(.) is the standard nor- mal density function; and σ refers the standard deviation of the error term. by maximising the log-likelihood function, the tobit estimator is obtained. rEsULts AND cONcLUsIONs As indicated in table 2, 63.10 % (248) of the 393 respondents that were considered in the study are males, and 36.90 % (145) are females, which represents all of turkey. It is also shown that 54.20 % (213) of the surveyed respond- ents are 31-45 years old, followed by individu- als of 18-30 and 46-64 years old, representing 38.17 % (150) and 7.38 % (29) of the sample re- spectively. the educational attainment of the respondents is in favour of higher level of edu- cation, 53.94 % (212) acquired a university de- gree followed by 42.24 % (166) of post graduate degree. When comparing the above figures with Ital. J. Food Sci., vol. 28 - 2016 111 the data of tUrKstAt as in table 3, our sample has higher income and education levels, and a higher percentage of males. regarding working status, a great majority of respondents (70.74 %) are employed in the pub- lic sector, while only 18.32 % and 4.33 % of the respondents work in the private sector and are unemployed respectively. taking into consider- ation income level of respondents, it was found that the middle income group was overwhelm- ingly predominant. respondents from low, me- dium and high income level consisted of rough- ly 12 %, 66 % and 32 % respectively. the aver- age size of the household of respondents is 3 in- dividuals per household and their age distribu- tion reflected 31-45 years old. table 2 - characteristics of the sample. Sample Size:393 Freq. % Gender 393 100 Male 248 63,10 Female 145 36.90 Age 393 100 18-30 150 38.17 31-45 213 54.20 46-64 29 7.38 >64 1 0.25 Employment Status 393 100 Public sector 278 70.74 Private sector 72 18.32 Retired 5 1.27 Unemployed 17 4.33 Housewife 5 1.27 Student 13 3.31 NGO 3 0.76 Education 393 100 Pri&High School 15 3.82 Graduate 212 53.94 Post Graduate 166 42.24 Household Size 393 100 1 person 47 11.96 2 people 63 16.03 3 people 123 31.30 4 people 107 27.23 >4 people 53 13.49 Monthly Income (1 TL=£0,28) 393 100 849 TL or less 16 4.07 850 TL – 1449 TL 29 7.38 1500 TL – 2149 TL 43 10.94 2150 TL – 2799 TL 69 17.56 2800 TL – 3449 TL 44 11.20 3500 TL – 4149 TL 64 16.28 4150 TL or more 128 32.57 Place of residence during the first 15 years of life 393 100 City or suburb 251 63.87 Small town 96 24.43 Village 46 11.70 table 3 - comparison of sample sociodemographics versus turkey’s Population. Sociodemographies Sample Turkey’s Population* Female (%) 36.9 49.8 Household Size 3.1 3.7 Graduates (%) 96.2 12.0 Median Income (TL) 3150 1838 Median age 40 31 *Elaborated from data extracted from TURKSTAT. table 4 fundamentally indicates the basic preferences stated by turkish consumers for pesticides and food safety issues. survey results showed that approximately 75 % of respond- ents have no idea about the pesticides and their harmful effects whereas only 20 % indicated lim- ited knowledge about pesticides. respondents aged 46-64 showed a higher degree of knowl- edge about pesticides. A great majority of those having pesticide knowledge specified mass media as a source of knowledge on pesticides. When a cross check question about the pesticides in potatoes was later asked, more than 50 % of respondents again indicated no idea about it; while 32 % of those have an opinion of “there are pesticide, hormone and other chemicals that are harmful for health”. regular shapes of potatoes are pre- dominantly remarked by respondents (around 56 % of respondents). A similar viewpoint comes from another question to observe how cosmetic defects are important for individuals. More than 86 % of respondents pointed out that they are not willing to pay for potatoes with insect dam- ages even though they are pesticide-free pro- duce. this finding might be interpreted that for those who are willing to pay more for pesticide- free products, suppliers should ensure that they can be provided with satisfactory quality stand- ards. Ott and MALIGAyA (1989) quoted in WEAv- Er et al. (1992) found that 88 % of the respond- ents would be unwilling to accept those defects. Apart from cosmetic defects, Independent sci- ence-based advice is one of the most important critical issues in food safety in the European Un- ion. European Food safety Authority (EFsA) as an independent body is responsible for carry- ing out risk assessment from risk management (EFsA, 2014). conflict of interest inevitably ap- pears when the same institutions both control and monitor the same findings. this is a crucial issue for turkey as well. therefore, a question was asked to observe the respondents’ opinions on “Who should carry out food safety control?”. the least frequent responses for this question are municipalities and public agents with rough- ly 4% and 12% respectively. the majority, 37%, of respondents preferred having an independent laboratory certification for more fair and trans- parent food safety control. 112 Ital. J. Food Sci., vol. 28 - 2016 table 4 - Pesticide concerns and purchasing preferences of turkish consumers. Source of Concern Freq. % Remember a serious incident 1 1.05 Heard concern expressed over one or more of mass media 48 50.53 Heard concern expressed by NGO`s 4 4.21 Heard concern expressed by Public agents 7 7.37 Other 35 36.84 Opinion about the pesticides in potatoes Freq. % There is no pesticide, hormone and other chemicals 17 4.33 There are pesticide, hormone and other chemicals, but residues are not risky for health 33 8.40 There are pesticide, hormone and other chemicals that are harmful for health 127 32.32 No idea 216 54.96 Purchasing preferences Freq. % No preservative including pesticide and hormones 21 5.34 Taste 78 19.85 Price 71 18.07 Regular shape 220 55.98 Brand 3 0.76 Purchasing place of potatoes Freq. % Open-air market 151 38.42 Greengrocer 46 11.70 Supermarket/Hypermarket/Shopping centre 174 44.27 Villagers 15 3.82 Others 7 1.78 Importance of cosmetic defects Freq. % Not important 0 0.00 Less important 53 13.49 More important 268 68.19 Highly important 72 18.32 Food safety control Freq. % Municipalities 16 4.07 Public agents 50 12.72 Universities 66 16.79 Independent agents 139 37.37 Producer Unions 13 3.31 Consumer Unions 109 27.74 based on the data in tables 1, 2, and 4, re- spondents aged 31 to 45 and having Master and PhD. degrees were found to be more willing-to- accept insect damage in reduced pesticides in potatoes than those aged 46 and older, and those having non-college and college degrees respec- tively. Males, lower income households and col- lege graduates were found to be less willing to accept cosmetic defects in reduced pesticides in potatoes than were females, high income house- holds and non-college graduates respectively. Finally, the survey results show that respond- ents considering pesticides in potatoes that are harmful for health and having no idea about it were found to be more willing-to-pay than were those considering no harmful pesticides in pota- toes and having no idea about pesticides respec- tively. this matter was comprehensively argued by rAvENsWAAy (1990). she mainly discussed that people with college degrees might be less concerned than those with non-college degrees since reaching knowledge for them is less cost- ly than others. they are, as a result of this de- duction, least willing-to-pay for the safe food. Additionally, it is possible to make regional comparisons at the level of NUts 1 regions. re- spondents from south East region (sE), Middle Eastern Anatolia (ME) and West Marmara re- gion (WMAr) were found to be less willing to pay for extra payment per kg for pesticide-free pota- toes than were other regions, while respondents from Ist (Istanbul) and East Marmara region (EMAr) which are largely industrialised parts of turkey were found to be more willing to pay for it than were other regions. Despite this result, it does not make sense at all to have any cor- relation between income and willingness to pay for the price increase per kg for potatoes has no major effect on individuals’ incomes. this is sup- ported by bUNtE et al. (2010). they showed that any reduction in organic prices for some prod- ucts such as potato has no considerable effect on demand. respondents from sE and EMAr were found Ital. J. Food Sci., vol. 28 - 2016 113 to be more willing-to-accept insect damage on pesticide-free produce than were other regions and respondents from West black sea (Wbs), and AEG are less willing to accept insect dam- age on reduced pesticides in potatoes than were other regions. variance Inflation Factor (vIF) should not ide- ally exceed rule of 4, rule of 10 in literature. If it exceeds the rule of thumb, it is regarded as casting doubts on the estimations of regression analysis. As attentively viewed from the results given in table 5, the vIF values among independ- ent variables change between 1.02 and 1.38 and mean vIF value is 1.14, which has sufficiently concrete evidence that there is no serious mul- ti-collinearity in the model. table 6 exhibits the estimation results pro- vided from the ordered probit model. As is illus- trated, cosmetic defects for consumer preferenc- es, free-pesticide potatoes with insect damages, indicating reasons of health for WtP questions, age, and gender were identified by the model to have significant impacts on the probability to WtP while spending the first 15 years in a vil- lage was found to negatively impact the proba- bility to WtP. however, income and education were not found to have a significant impact, pos- itively or negatively, on the probability to WtP. being female increases the probability of WtP by 21% as revealed in most of the studies (hEN- sON, 1996; GILL et al., 2000; LOUrEIrO et al., 2002; KONtOLEON et al., 2005; sUNDstrOM and ANDErssON, 2009). this can easily be explained as women are more sensitive to food safety problems than men. Also, those indicating health reasons for WtP question were found to increase the prob- ability to WtP by 43%. On the contrary, KA- LOGErAs et al. (2009) found that health aspect does not significantly influence the probability of WtP. similar effects were observed on cos- metic defects and age. considering cosmetic de- fects as an important feature for their purchas- ing preferences raises the probability to WtP by 12%. In much the same way, the age of our model had a positive impact (by 10%) on WtP as in most of the studies (MIsrA et al., 1991; KONtOLEON et al., 2005; DEttMANN and DIM- ItrI, 2010). contrariwise, the age of the con- sumers were found to have a negative effect table 5 - collinearity diagnostic. Variable VIF 1/VIF Working Condition 1.38 0.72475 Income 1.38 0.725141 Education 1.15 0.871934 Age 1.12 0.893866 Insect Damage 1.06 0.943308 Living in a village 1.06 0.945987 Reason for Health 1.04 0.96083 Harmful pesticides 1.04 0.960983 Cosmetic Defects 1.02 0.981626 Mean VIF 1.14 table 6 - the Probit Model. Dependent Variable: WTP Variable Coefficient Standard error Marginal effect Standard error Constant -2.70513*** 0.739544 - - Knowledge -0.25266 0.180227 -0.09928 0.07116 Cosmetic Defects 0.319988** 0.128319 0.124498** 0.0499 Insect Damage 0.41714* 0.223001 0.154209** 0.07671 Harmful pesticide 0.242785 0.159825 0.093258 0.06043 Reason for Health 1.151586*** 0.149467 0.434584*** 0.05094 Age 0.267368** 0.122852 0.104025** 0.04781 Working Condition 0.078549 0.063429 0.030561 0.02469 Gender 0.556782*** 0.152852 0.210075*** 0.05503 Education Level 0.046347 0.36927 0.018112 0.14491 Income 0.038651 0.131314 0.015038 0.05109 Living in a village -0.39952* 0.219629 -0.15794* 0.0866 ***Indicates significance at 1% level, **at 5% level, *at 10% level. Probit regression Number of obs = 393 LR chi2(11) = 93.65 Prob > chi2 = 0.0000 Log likelihood = -220.82775 Pseudo R2 = 0.1749 166 left-censored observations at pay <=0; 227 uncensored observations; 0 right-censored observations. 114 Ital. J. Food Sci., vol. 28 - 2016 table 7 - the tobit Model. Dependent Variable:MWTP Variable Coefficient Standard error Marginal effect Standard error Constant -5.15815*** 1.344697 - - Knowledge -0.27364 0.328352 -0.1237628 0.14583 Cosmetic Defects 0.681142*** 0.228425 0.3135761*** 0.10497 Insect Damage 0.607647 0.373106 0.2955922 0.19146 Harmful pesticide 0.361183 0.284776 0.1690033 0.13527 Reason for Health 2.334047*** 0.288372 0.9972112*** 0.11239 Age 0.529826** 0.220929 0.243915** 0.10148 Working Condition 0.166313 0.112667 0.076565 0.05181 Gender 0.903691*** 0.263868 0.4291193*** 0.12871 Education Level 0.379784 0.655299 0.1673579 0.2761 Income 0.028448 0.233872 0.0130964 0.10767 Living in a village -0.87384** 0.41634 -0.3699325** 0.16148 ***Indicates significance at 1% level, **at 5% level, *at 10% level. Tobit regression Number of obs = 393 LR chi2(11) = 98.80 Prob > chi2 = 0.0000 Log likelihood = -627.37228 Pseudo R2 = 0.0730 on the WtP for organic potatoes by LOUrEIrO and hINE (2002) and reduced pesticides in to- matoes by AKGUNGOr et al. (2007). Addition- ally, spending the first fifteen years in a village reduces the probability to WtP by 16%, ceteris paribus. the interpretation could be made that those people who spent their first fifteen years in a village might have a lower level of educa- tion, thus, less knowledge of pesticide impacts and less sensitiveness to the topic. table 7 summarizes the results of the tobit model concerning their marginal effects. Indi- viduals who considered cosmetic defects as im- portant features for potato preferences, who are female, and who were indicating health reasons for WtP questions have higher WtP. to put it in context, considering cosmetic defects as impor- tant features for potato preferences raises the WtP amount by tL 0.31, and similarly, being fe- male raises the WtP amount by tL 0,4 respec- tively, ceteris paribus. respondents who spent their first fifteen years in a village have signifi- cantly lower WtP. the mean WtP amount was estimated for the reduced pesticides in potatoes in turkey on the basis of cvM study. the survey covering all of turkey showed that respondents, representing different geographical areas, on an average are willing to pay extra tL2.90 if the zero respond- ents corresponding to approximately 42% are not included in the models. If it was included, the mean would be extra tL1.67. these absolute numbers can be given in percentages as 48% and 83% price premium for reduced pesticides in po- tatoes per kg, respectively. the average market price for potato was found as tL 3.50 based on the virtual turkish super-market prices for those dates. the estimations could be likely interpret- ed that demand for organic food among turkish consumers is growing. In a similar study, GIL et al. (2000) presented that spanish consumers living in Navarra and Madrid would be willing to pay 17 % and 5.6 % more for organic potatoes, respectively. this big gap between turkish and spanish consumers can be explained mainly by the organic markets in turkey that are not suf- ficiently saturated yet. A similar result was found by AKGUNGOr et al. (2007) that turkish consumers would be will- ing to pay 36% price premium for organic prod- ucts or certified products. Also, WEAvEr et al. (1992) found that 26% of respondents in Penn- sylvania were willing to pay more than 15% for organic tomatoes. As seen from the values and percentages, there are no extreme prices that are accepted by consumers. this situation was argued by rAWENsWAAy (1990) that consumers would be willing to pay modest amounts to re- duce perceived health risks in food. two important caveats can be placed on any discussion drawn from the survey results. First, actual WtP cannot be observed as it is solely based on stated preferences. second is the ho- mogenous distribution of individuals with re- spect to income and education. In spite of the fact that education and income are found to be significant factors for many WtP studies, no re- lationship was found in our model. the first one seems more important while income has a minor impact on an individu- al’s budget as indicated by bUNtE et al (2010). however, there is no consensus in literature in- Ital. J. Food Sci., vol. 28 - 2016 115 dicating a certain effect of education on WtP amount. though DEttMAN and DIMItrI (2010) found a positive relation between education and WtP for organic products, MIsrA et al. (1991); bUzby et al. (1995); thOMPsON and KIDWELL (1998); bOrcELEttI and NArDELLA (2000) and sUNDstrOM and ANDErssON (2009) found a negative relation. It was also affirmed by van rAvENsWAAy (1995) that the people with high- er education level may be less concerned about pesticides because they might be better able to reach reliable information. these results might help to affirm why there is no significant im- pact of education on WtP in our model con- sidering an outstandingly high rate of educat- ed respondents. Lastly, survey results show that the respond- ents overwhelmingly indicate that they have no idea about the level of pesticide residues used in the food. roughly 32% of respondents consid- ered that there are serious pesticide residues in potatoes, which are harmful to human health. An interesting finding from the survey results comes from the question “who should be respon- sible for controlling and monitoring of residues in food”. Approximately 37.4% of respondents were in favour of independent laboratories while only 12.7 % went for public agents as an answer to this question. this clearly demonstrates that there is a high demand from consumers’ side to independent agents for neutral decisions rath- er than public institutions. As a result, this study stresses the consumer attitudes for pesticides in potatoes by employing cvM and single-bounded probit and tobit mod- els. One of the drawbacks of the survey is based on the stated preferences rather than revealed preferences. the respondents might answer the questions with overestimation if compared with real situations. It would thus be better as a fu- ture research agenda to conduct another study in order to observe if similar results were truly provided by respondents. rEFErENcEs Akgungor s., Miran b. and Abay c. 2001. consumer Will- ingness to Pay for Food safety Labels in Urban turkey. Journal of International FoodandAgribusiness Market- ing, 12(1):91. Akgungor s., Miran b. and Abay c. 2007. consumer Will- ingness to Pay for Organic Products in Urban turkey”. 105th EAAE seminar, International Marketing and Inter- national trade of Quality Food Products, bologna, Italy. Alavanja M.c., hoppin K.A. and Kamel F. 2004. health Ef- fects of chronic Pesticide Exposure: cancer and Neuro- toxicity. Annual review Public health, 25:155. Amemiya t. 1981. Qualitative response Models: A survey. Journal of Economic Literature, 19:1483-. Arrow K., solow r., Portney P.r., Leamer E.E., radner r. and schuman h. 1993. “report of the NOAA Panel on contin- gent valuation,” Federal register, January 15 (58):4601. Ayaz A. and yurttagul M. 2008. “toxic substances in nutri- tion-1”. University of hacettepe, Department of nutrition and dietetics, Ankara. birinci A. and Uzundumlu A. 2009. An assessment of pro- ducer’s approach to agricultural pesticide usage in po- tato production: A case study in Erzurum, turkey, sci- entific research and Essay vol.4 (11):1225. boccaletti s. and Nardella M. 2000. consumer willingness to pay for pesticide-free fresh fruits and vegetables in It- aly. International Food and Agribusiness Management review, 3:297. bunte F.h.J., van Galen M.A., Kuiper W.E. and tacken G. 2010. Limits to Growth in Organic sales. De Economist 158 (4): 387. buzby J., ready r. and skees J.r. 1995. contingent valua- tion in Food Policy Analysis: A case study of a Pesticide- residue risk reduction. Journal of Agriculture and Ap- plied Economics, 27 (2):613. cakmakyapan s. and Goktas A. 2013. A comparison of bi- nary Logit and Probit Models with a simulation study. the Journal of social and Economic statistics. vol. 2 (1), summer 2013. cameron t.A. 1992. combining contingent valuation and travel cost Data for the valuation of Nonmarket Goods. Land Economics, 68:302. carvalho F.P. 2006. Agriculture, pesticides, food security and food safety. Environmental science and Policy, 9:685. Dettmann r.L. and Dimitri c. 2010. Who’s buying Organic vegetables? Demographic characteristics of the Us con- sumers. Journal of Food Products Marketing, 16;1:79 Eberle W.D. and hayden F.G. 1991. critique of contingent valuation and travel cost Methods for valuing Natural resources and Ecosystems. Journal of Economic Issues, vol. XXv (3)649. Ec. 2009. “EU Action on Pesticides, Our food has become greener” Factsheet. European commission, brussels. EFsA 2014. “About EFsA” European Food safety Authority. Evans J.r. and Mathur A. 2005. the value of online sur- veys, Internet research, vol. 15 (2):195. Fink A. 2003. “the survey handbook”, 2nd Edition. sage Pub- lications Inc. London. Garming h. and Waibel h. 2006. Willingness to Pay to Avoid health risks from Pesticides, A case study from Nicara- gua, Woking Paper 2006,4. Gil, J.M., Gracia A. and sánchez M. 2000. Market seg- mentation and Willingness to Pay for Organic Products in spain. International Food and Agribusiness Manage- ment review 3:207. Greene, W.h. 2012. “Econometric Analysis”, 4th Ed. Pren- tice hall, New Jersey. henson s. 1996. consumer Willingness to Pay for reduc- tions in the risk of Food Poisoning in the UK. Journal of Agricultural Economics, 47 (3):403. hoppin J.A., Umbach D.M., Kullman G.J., henneberg- er P.K., London s.J., Alavanja M.c.r. and sandler D.P. 2007. Pesticides and Other Agricultural Factors Associ- ated with self-reported Farmer’s lung among Farm resi- dents in the Agricultural health study. Occupational En- vironment and Medicine, 64:334. Jean c.c. richard r. and skees J. 1995. contingent valua- tion in Analysis: A case study. residue risk reductions, J. Agr. and Applied Econ. 27 (2):613. Kalogeras N., valchovska s., baourakis, GandKalaitzis P. 2009. Dutch consumers’ Willingness to Pay for Organic Olive Oil. Journal of International FoodandAgribusiness Marketing, 21:286. Konteleon A., yabe, M. and Darby L. 2005. Alternative Pay- ment vehicles in contingent valuation: the case of Ge- netically Modified Foods. Munich Personal rePEc Archive. Lindhjem h. and Navrud s. 2011. Are Internet surveys an alternative to face-to-face interviews in contingent val- uation. Ecological Economics, 70(9):1628. Loureiro M.L. and hine s. 2002. Discovering Niche Markets: A comparison of consumer Willingness to Pay for Local (colorado Grown), Organic, and GMO-Free Products. Journal of Agricultural and Applied Economics, 34(3):477. 1 turkish Lira equals roughly € 0.34. http://www.emeraldinsight.com/action/doSearch?target=emerald&logicalOpe0=AND&text1=Evans, J R&field1=Contrib http://www.emeraldinsight.com/action/doSearch?target=emerald&logicalOpe0=AND&text1=Mathur, A&field1=Contrib 116 Ital. J. Food Sci., vol. 28 - 2016 Loureiro M.L., Mccluskey J., and Mittelhammer r.c. 2002. Will consumers Pay A premium for Eco-labelled Apples. Journal of consumer Affairs, 36(2):203. MFAL. 2012. statistics for Pesticides, the Ministry of Food, Agriculture and Livestock, Ankara. www.tarim.gov.tr. Misra s.K., huang c.L. and Ott s. 1991. consumer Willing- ness to Pay for Pesticide-Free Fresh Produce. Western Journal of Agricultural Economics, 16(2): 218. Ott s.L. and Maligaya A. 1989. “An Analysis of consumer Attitudes toward Pesticide Use and the Potential Mar- ket for Pesticide residue-Free Fresh Produce” south- ern Agricultural Economics Association Meetings, Nashville, tN. rowell A. 2004. Don’t worry It’s safe to eat. Earthscan pub- lications Ltd. UK. seller c., stoll J.r., and chavas J.P. 1985. validation of Empirical Measures of Welfare change: A comparison of Nonmarket techniques, Land Economics, 61 (2):156. smith r.D. 1997. contingent valuation: Indiscretion in the Adoption of Discrete choice Question Formats. centre for health Program Evaluation, Working Paper 74. sundstrom K. and Andersson h. 2009. swedish consumers’ Willingness-to-Pay for Food safety. Working paper 2009:1. thompson G.D. and Kidwell J. 1998. Explaining the Organ- ic choice of Organic Produce: cosmetic Defects, Prices, and consumer Preferences. American Journal of Agricul- tural Economics, 80:277. van ravenswaay E. and Wohl J. 1995. “Using contingent valuation Methods to value the health risks from Pes- ticide residues When risks are Ambiguous” In: J. cas- well (Ed.) valuing food safety and nutrition. chap.14:287. van ravenswaay E.v. 1990. “consumer Perception of health risks in Food” Michican state University. Weaver r. Evans D. and Luloff A.E. 1992. Pesticide Use in tomato Production: consumer concerns and Willing- ness-to-Pay. Agribusiness, vol. 8:131. Wooldridge J. 2006. “Introductory Econometrics” 4th Edi- tion. south Western. Paper Received October 10, 2014 Accepted May 5, 2015 http://www.tarim.gov.tr Ital. J. Food Sci., vol. 28 - 2016 117 APPENDIcEs Appendix 1. Survey Instrument Appendix 1.1. hypothetical scenario health risks resulting from pesticide use have made food safety a priority issue on the public policy agenda in devel- oped countries. A research made in U.s showed that pesti- cide residues were rated a serious risk by 68 of respondents attending in a survey. Pesticides can cause many types of health problems in humans. “Pesticides have been linked to a wide range of human health hazards, ranging from short- term impacts such as headaches and nausea to chronic im- pacts like cancer, reproductive harm, and endocrine disrup- tion (toxic Action center1)”. Ec Directive 2009/128/Ec determined the sustainable use of pesticides to reduce health risks resulted from pesti- cides. therefore, EU countries minimise or ban the use of pesticides for health reasons. turkey as a candidate coun- try for EU membership has to harmonize her own legisla- tions and directives. the amount of pesticide use in turkey has gradually in- creased since 2009 and it was over 40,000 ton in 2011 ac- cording to the data taken by the Ministry of Food, Agricul- ture and Livestock of turkey. Particularly, potato is one of the most consumed vegetables which seriously include pes- ticide residues in turkey. the scenario it is proposed for this survey is a price increase for reduced pesticides in po- tatoes per kg. the research project is aimed at evaluating your opinion of reduced pesticides in potatoes. reduced pesticides are in general valued for one or more of the following attributes: better taste, food safety, health, freshness, environment preservation and local production. Good Agricultural Prac- tices are “practices that address environmental, econom- ic and social sustainability for on-farm processes, and re- sult in safe and quality food and non-food agricultural prod- ucts” (FAO, 2003). More precisely the main aim of this study is to find out what would persuade you to buy reduced pesticides in potatoes. On this basis the questionnaire tries to find out your opin- ion of the quality and availability of the reduced pesticides in potatoes in turkey and the price that you would be will- ing to pay for these reduced pesticides in products. Finally, for the purposes of the study you are required to give truthful answers and we recommend that you think care- fully about the scenario previously mentioned, your dispos- able income and health concerns during the questionnaire survey. Furthermore, you should notice that this survey is completely anonymous and confidential. however, if you de- sire a copy of the final study, you should provide an email address so it can be sent to you. Appendix 1.2. Questionnaire Questions about qualifying candidates 1) Please indicate your current place of residence. 2) Please indicate whether you participate in the decisions regarding the payments in your household. a) yes b) No Questions about perceptions for food 3) Please indicate whether or not you have an idea regard- ing level of pesticides and hormones in potatoes, if you in- dicate choice a, please go question 5. a) No idea b) Little information c) sufficient information c) All information in detail 4) Please indicate your recalling of pesticide information as related to level of concern for human health. a) remember a serious incident b) heard concern expressed over one or more of mass media c) heard concern expressed by NGO`s d) heard concern expressed by Public agents e) Other 5) Please indicate the most important feature of potato for your purchasing preferences. a) No preservative including pesticide and hormones b) taste c) Price d) regular shape e) brand 6) Please indicate how cosmetic defects are important for your purchasing preferences in pesticide free products. cosmetic defects refer growth cracks and knobby or irreg- ular growth. a) Not important b) Less important d) More important c) highly important 7) Please indicate if you accept potatoes with insect damage, such as worm holes in pesticide free products. a) yes b) No 8) Please indicate your opinion about the pesticides, hor- mones, and other chemicals for potatoes. a) there is no pesticide, hormone and other chemicals b) there are pesticide, hormone and other chemicals, but residues are not risky for health c) there are pesticide, hormone and other chemicals that are harmful for health d) No idea 9) Please indicate what you generally do in order to allevi- ate your concern over pesticide dangerous in the potatoes. a) Nothing b) Washing it with plenty of water c) consuming by peeling off it d) cooking e) Other (Please specify) 10) Please indicate whether or not fresh fruit and vegeta- bles are as healthy as it was in the past with respect to health safety. a) Never healthy b) still healthy c) better healthy d) No idea Questions about willingness-to-pay At this stage, you should consider that the payment vehicle for the reduced pesticide in potato will lead to increases in potato prices if you favour the reduced pesticides in potato. Moreover, we strongly recommend you to consider your disposable income, health concerns, and possible positive and negative consequences of the reduced pesticide in potato when making your decision. 11) Would you be willing to pay extra 2 tL/per kg for reduced pesticides in potato? If answer is yes, please go to question 12, otherwise go to question 18. a) yes b) No 12) Would you be willing to pay extra 2.5 tL/per kg for re- duced pesticides in potato? a) yes b) No If answer is no, please go to question 17. 13) Would you be willing to pay extra 3 tL/per kg for re- duced pesticides in potato? a) yes b) No If answer is no, please go to question 17. 14) Would you be willing to pay extra 3.5 tL/per kg for re- duced pesticides in potato? a) yes b) No If answer is no, please go to question 17. 15) Would you be willing to pay extra 4 tL/per kg for re- duced pesticides in potato? a) yes b) No If answer is no, please go to question 17. 16) Would you be willing to pay above 4 tL/per kg for re- duced pesticides in potato? Also please indicate how much you would be willing to pay. a) yes (Please specify): tL b) No how much:........................................................................ 1 http://www.toxicsaction.org/problems-and-solutions/pesticides http://www.toxicsaction.org/problems-and-solutions/pesticides 118 Ital. J. Food Sci., vol. 28 - 2016 17) Would you please indicate the reason for the expressed amount? a) More healthy b) A reasonable price for my budget c) More tasty d) Protecting environment e) Protecting local producers f) Other (Please specify) Questions about social and economic factors 18) regarding your age, which of the following would you select? a) 17 or less b) 18-30 c) 31-45 d) 46-64 e) 65 or more 19) regarding your working condition, which of the follow- ing would you select? a) Public sector b) Private sector c) retired d) Unemployed e) housewife f) student g) Farmer h) NGO 20) regarding your gender, which of the following would you select? a) Male b) Female 21) regarding your marital status, which of the following would you select? a) Married b) single 22) regarding your family composition, which of the follow- ing would you select? a) have children b) Do not have children 23) regarding the size of your household, which of the fol- lowing would you select? a) One person b) two persons c) three persons d) Four persons e) More than four persons 24) regarding your education level, which of the following would you select? a) Primary school graduate b) secondary school graduate c) high school graduate d) bachelor’s degree graduate e) Master’s degree graduate f) Ph.D. ’s degree graduate g) Other:............................................................................ 25) regarding your monthly income, which of the following would you select? a) 849 tL or less b) 850 tL – 1449 tL c) 1500 tL – 2149 tL d) 2150 tL – 2799 tL e) 2800 tL – 3449 tL f) 3500 tL – 4149 tL g) 4150 tL or more 26) Please indicate the place of residence during the first 15 years of life? a) city or suburb b) small town c) Farm 27) Please indicate the place you are currently living? a) Less than 3 years b) 3-5 years c) 6-10 years d) 11-20 years e) More than 20 years 28) Please indicate from where do you generally purchase potatoes? a) Open-air market b) Greengrocer c) supermarket/hypermarket/shopping center d) villagers e) Others 29) Please indicate your preference about which agent should ideally and fairly be responsible for food safety control? a) Municipalities b) Public agents c) Universities d) Independent agents e) Producer Unions f) consumer Unions thank you for your time! Appendix 2. Summary and descriptions of variables Variable Obs Mean Std. Dev. Min Max pay 393 1.676845 1.621334 0 6 bid 393 0.5776081 0.4945699 0 1 Knowl 393 0.2417303 0.4286774 0 1 cosm_Def 393 3.048346 0.5626138 2 4 Insect_Dam 393 0.1399491 0.3473765 0 1 harmfulpes 393 0.3231552 0.4682776 0 1 Age 393 2.697201 0.6123035 2 5 Work_cond 393 1.608142 1.289334 1 8 Gender 393 0.3689567 0.4831373 0 1 Marital 393 0.4707379 0.4997793 0 1 hav_child 393 0.4274809 0.4953436 0 1 household 393 3.142494 1.197383 1 5 livingincity 393 0.6386768 0.4809963 0 1 livingindist 393 .2442748 0.4302041 0 1 livinvilage 393 0.1170483 0.3218877 0 1 Educ 393 0.956743 0.2036944 0 1 Income 393 2.223919 0.6273835 1 3 livinvilage 393 0.1170483 0.3218877 0 1 Ital. J. Food Sci., vol. 28 - 2016 119 Appendix 3. Multicollinearity analysis K nowl C os m_D ef I ns ect_D am H ar mfulpes R eas onH ealth A ge W or k _C ond G ender E ducdumy I ncome livinvilage K nowl 1.0000 C os m_D ef -0.0380 1.0000 I ns ect_D am 0.2519*** -0.0739 1.0000 H ar mfulpes 0.3469*** -0.0401 0.1604*** 1.0000 R eas onH ealth -0.0888 -0.0306 -0.0828 0.0709 1.0000 A ge 0.1435*** -0.0092 -0.0281 0.0129 -0.0121 1.0000 W or k _C ond 0.0057 0.0016 0.1114** -0.0179 (-)0.1438*** -0.0667 1.0000 G ender -0.0129 (-)0.1127* 0.0716 0.0129 0.0402 (-)0.1647* 0.0853** 1.0000 E ducdumy 0.0032 -0.0262 -0.0224 -0.0135 0.0508 -0.0439 (-)0.327*** 0.0707 1.0000 I ncome -0.0216 -0.0018 (-)0.109* 0.0309 0.0623 0.2566*** (-)0.4558*** (-)0.2059***0.2357*** 1.0000 livinvilage 0.1457*** 0.0955* 0.0356 0.0531 -0.0762 0.1803*** -0.0306 -0.0816 -0.0393 0.0467 1.0000 Appendix 4. Regression analysis Source SS df M S Number of obs = 393 F ( 9, 383) 10.79 M odel 14.56733 9 1.61859267 Prob > F 0 R esidual 57.46829 383 .150047753 R -squared 0.2022 A dj R -squared 0.1835 T otal 72.03562 392 .183764345 R oot M SE 0.38736 K nowl C oef. Std. E rr. t P>t [95 % C onf. I nterval] C osm_Def -0.01601 .0350985 -0.46 0.649 -0.0850178 0.053002 I nsect_Dam 0.233844 .0579889 4.03 0.000 0.1198276 0.34786 H armfulpes 0.290287 .0426197 6.81 0.000 0.2064893 0.374085 R easonH ealth -0.07965 .0416259 -1.91 0.056 -0.1614897 0.002198 A ge 0.100169 .0337963 2.96 0.003 0.0337199 0.166619 W ork_C ond -0.01019 .0178243 -0.57 0.568 -0.0452324 0.024859 E ducdumy 0.06667 .102861 0.65 0.517 -0.1355734 0.268913 I ncome -0.04631 .0366208 -1.26 0.207 -0.1183125 0.025693 livinvilage 0.126524 .062492 2.02 0.044 0.0036537 0.249395 _cons -0.01413 .1961847 -0.07 0.943 -0.3998678 0.3716 Appendix 5. Covariance matrix of coefficients of regress model e(V ) C osm_D ef I nsect_D am H armfulpes R easonH ealth A ge W ork_C ond E ducdumy I ncome livinvillage _cons C osm_D ef 0.0012319 I nsect_D am 0.0001513 0.00336271 H armfulpes 4.661E -05 -0.00041276 0.00181644 R easonH ealth 3.774E -05 0.0001939 -0.0001506 0.00173271 A ge 3.531E -05 0.00002271 6.14E -06 0.00001231 0.0011422 W ork_C ond 4.64E -07 -0.0000685 8.00E -06 0.00008935 -2.13E -05 0.00031771 E ducdumy 8.484E -05 -0.00015164 0.00011349 -9.37E -06 0.0003194 0.0004571 0.0105804 I ncome -1.32E -06 0.00014854 -5.893E -05 9.25E -06 -0.000321 0.00025825 -0.000478 0.00134108 livinvilage -0.0002205 -0.00013419 -0.0001421 0.00020032 -0.000363 0.00004616 0.000235 7.79E -06 0.0039053 _cons -0.0039642 -0.00104393 -0.0005647 -0.00142089 -0.002716 -0.00152239 -0.010951 -0.0020787 0.0008133 0.0384884 120 Ital. J. Food Sci., vol. 28 - 2016 Appendix 6. Correlation matrix of coefficients of regress model e(V ) C osm_D ef I nsect_D am H armfulpes R eason_health A ge W ork_C ond E ducdumy I ncome livinvillage _cons C osm_D ef 1 I nsect_D am 0.0743 1 H armfulpes 0.0312 -0.167 1 R easonH ealth 0.0258 0.0803 -0.0849 1 A ge 0.0298 0.0116 0.0043 0.0087 1 W ork_C ond 0.0007 -0.0663 0.0105 0.1204 -0.0354 1 E ducdumy 0.0235 -0.0254 0.0259 -0.0022 0.0919 0.2493 1 I ncome -0.001 0.0699 -0.0378 0.0061 -0.2594 0.3956 -0.127 1 livinvilage -0.1005 -0.037 -0.0534 0.077 -0.1719 0.0414 0.0366 0.0034 1 _cons -0.5757 -0.0918 -0.0675 -0.174 -0.4096 -0.4354 -0.5427 -0.2893 0.0663 1 Appendix 8. Statistic values of WTP before and after trimming outlier P ercentiles Smallest P ercentiles Smallest 1% 0 0 1% 0 0 5% 0 0 5% 0 0 10% 0 0 Obs 393 10% 0 0 Obs 374 25% 0 0 Sum of W gt. 393 25% 0 0 Sum of W gt. 374 50% 2 Mean 1.676845 50% 2 Mean 1.497326 L argest Std. D ev. 1.621334 L argest Std. D ev. 1.444224 75% 2.5 6 75% 2.5 4 90% 4 6 V ariance 2.628723 90% 4 4 V ariance 2.085784 95% 4 6 Skewness 0.424015 95% 4 4 Skewness 0.2272442 99% 4 6 K urtosis 2.16069 99% 4 4 K urtosis 1.633528 pay T rimmed data (5% ) pay Appendix 7. Logit model WTP Coef. Std. Err. Knowl -0.4106674 0.3074221 cosm_Def 0.5257642** 0.2167355 Insect_Dam 0.6643928*** 0.3706913 harmfulpes 0.4106739 0.269297 reasonhealth 1.886197*** 0.2523839 Age 0.4439607** 0.2078363 Work_cond 0.1279345 0.1060791 Gender 0.9363477*** 0.2617843 Educdumy 0.0939652 0.6196589 Income 0.0539704 0.2209946 livinvilage (-)0.681425* 0.3660309 _cons -4.444088 1.238931 Number of obs=393 Lr chi2 (11)=93.33 Prob>chi2=0.0000 Pseudo r2=0.1743 ***Indicates significance at 1% level, **at 5% level, *at 10% level.