The influence of incentives in eliminating hypothetical bias: Evidence from a choice-based conjoint experiment for beef products in Iringa and Mbeya regions in Tanzania

Journal of Agricultural Economics and Development Vol. 2(8), pp. 324-332, August 2013 Available online at http://academeresearchjournals.org/journal/jaed ISSN 2327-3151 ©2013 Academe Research Journals Full Length Research Paper The influence of incentives in eliminating hypothetical bias: Evidence from a choice-based conjoint experiment for beef products in Iringa and Mbeya Regions in Tanzania S. W. Nandonde1*, E. E. Msuya1, L. A. Mtenga2, F. T. Kilima1 and R. Alphonce1 1 Department of Agricultural Economics and Agribusiness, Sokoine University of Agriculture, P. O. Box 3007, Morogoro Tanzania. 2 Department of Animal Science and Production, Sokoine University of Agriculture, P. O. Box 3014, Morogoro, Tanzania. Accepted 6 June, 2013 Consumer responses were observed for a within-sample comparison of preference and willingness-topay (WTP) estimates for tenderness, leanness, freshness and hygiene for beef products from finished cattle and non-finished cattle (status quo). This comparison was conducted through two sessions of repeated choice-based conjoint experiments (CBC), starting with a hypothetical choice-based conjoint (HCBC) experiment that involved cheap talk only followed by a real choice-based conjoint (RCBC) experiment that involved the actual purchase of experimental products with real money. Consumers prefer more tender, less fatty, chilled beef and clean retailing premises, regardless of the choice session; however, the estimated coefficients were not equal (p<0.001). The selection was motivated by alternatives in HCBC where finished beef constituted 76% of all choices made. The selection of finished beef dropped to 67% in RCBC where consumers were sensitive to the price and quality content of alternative products. Consumers overestimate the WTP for hygiene in HCBC (p=0.014); however, there are no significant differences in WTP estimates for other attributes. Therefore, it is concluded that monetary incentives can reduce hypothetical choice bias and provide more trustworthy WTP estimates for all attributes. Key words: Beef, finishing, chilling, hygiene, tenderness, preferences, willingness to pay. INTRODUCTION The development and shaping of the beef industry depend on customer desires and preferences for various ranges of beef quality attributes. The importance of beef consumers in the entire beef industry worldwide has necessitated the measuring of information concerning consumer preferences and willingness to pay (WTP) for attributes of interest, such as freshness, tenderness, fat content and safety or hygiene, using the conjoint technique as an appropriate marketing research tool because it reflects the real market scenario (Hensher et al., 2005). The technique has recently been adopted in market research in developing countries, including Tanzania (Jabbar et al., 2010; Nandonde et al., 2013a). Intuitively, conjoint experiments have suffered from consumer bias to the extent of losing financial meaningful to marketers particularly when the experiment is basically too hypothetical for consumer to put cognitive effort into their decisions as they make choices out of habit (Dawnay and Shah, 2005; Lusk et al., 2008; Alfenes et al., 2009). This issue has led to the reframing of choice experiments by conducting field-oriented experiments where real subjects can be found instead of the typical laboratory experiments that involve mainly students (List 2011). In addition, such incentives as cheap talk scripts and real purchase arrangements of the products provide *Corresponding author. E-mail: [email protected]. Tel: +255787772222. J. Agric. Econ. Dev. chances for accruing more truthful information from consumers (Lusk, 2003; Alfenes et al., 2006; Lusk and Shroeder, 2004; Lusk et al., 2008; Genon et al., 2011). The use of purchase arrangements with real money provides a baseline to differentiate between hypothetical and real choices (Carlsson and Martinsson, 2001; Ding, 2007). According to Alfenes et al. (2006), using real money in choice-based experiments might be favoured because it reflects a common practise in most retail markets whereby sellers post the price and consumers choose which product to buy. Nevertheless, the reframing of the choice session and the extent of bias to CBC is relatively complicated because in certain cases, no bias has been observed, and in certain other cases, substantial differences have been observed (Carlsson and Martinsson, 2001; Chowdhury et al., 2009). These ambiguities could be attributed to the type of product in question, the exposure of respondents to choice experiments and the bargain technique to reflect the rational decision in the respondents’ choices (Carlsson and Martinson, 2001; Chowdhury et al., 2009; Ginon et al., 2011; Cecchi and Bulte, 2013). This issue has created the need for researchers to understand the potential differences that exist in hypothetical and real choice set ups (List, 2006). This quest is conducted not only for the sake of understanding the presence of choice bias and its influence on the implicit price measures but also to find the necessity of conducting such relatively expensive real choice experiments in countries such as Tanzania where beef consumers/respondents are not exposed to such a methodology compared to ordinary surveys using questionnaires. Previous conjoint studies from Tanzania have been dedicated to findings regarding monetary incentives (Alphonce and Alfenes, 2012; Nandonde et al., 2013a; Nandonde et al., 2013b) while studies regarding beef product in nearby countries mostly have been too hypothetical to address the differences (Jabbar et al., 2010). This paper is aimed at examining the presence of hypothetical bias and its consequences on the willingness-to-pay estimates for quality-improving beef attributes of freshness, leanness, tenderness and hygiene of retailing outlets in the Iringa and Mbeya Regions in Tanzania (Figure 1). MATERIALS AND METHODS Econometric modelling of choice experiments The conceptual foundation of choice experiments relies on Lancaster’s theory of value, which proposes that utilities for goods and services can be decomposed into separate utilities for their characteristics or attributes (Lancaster, 1966), and random utility theory, which explains the dominance judgments made between a pair of offering (Thurstone, 1927). Within this framework, 325 subjects (consumers) choose among alternatives according to a utility function that comprises two components (Equation 1). U ni  Vni   ni , (1) Where Uni is the utility obtained by individual n by selecting alternative i from a finite set of J alternatives in choice set C in situation t. Vni represents the systematic component of the utility (the indirect utility), which is a function of the attributes of an alternative [V ni = f (Xni)]; Xni is the vector of alternative i. εni is the random term. Therefore, the individual n will choose alternative i if it’s utility is higher than that of its alternative k within the same choice set; hence, his probability (P ni) is shown in Equation 2. Pni  Pr obVni   ni   Vnj   nk : k  C (2) Various logit probabilistic models are used in analysing choice experiment data, such as the hybrid conditional logit models, the random coefficient models, the hierarchical Bayes model and the latent class model, depending on the extent of the model to reveal more preference heterogeneity compared to ordinary or basic probabilistic models that assume that consumers are homogenous in their selection or preferences (Kallas et al., 2007 and Ortega et al., 2010). The hybrid condition logit model or the random effect logistic model are preferred as basic models in handling repeated choice data and allow flexibility through the interaction of the subjects’ socio-economic characteristics through the indirect utility function (Radder and le-Roux, 2005; Kallas et al., 2007; Hole 2008). Product alternatives and attributes for experimental design Our experiment was designed based on the on-going situation on the ground, where most beef is sourced from a local unimproved (non-finished) production system, which was treated as the status quo. The assumption was that both local and cross-breed cattle can be improved to produce the same quality beef with respect to tenderness; however, a difference might arise regarding the possibility of having higher adipose fat content from local cattle compared to crossbred cattle. 2 2 1 The full factorial of 2 × 3 × 5 = 180 choice combinations was reduced to 10 (Table 1). With the exception of hygiene of retailing outlet, attributes attached to beef itself that count for over 85% of all attributes were presented to consumers in their real forms using actual beef products stored in a refrigerator and a visible photo display of a whole carcass. The experiments employed the prevailing market price for the status quo beef for the Nandonde et al 326 Figure 1. Map of Iringa and Mbeya Regions with the specific study areas. meat-on-bone (Mchanganyiko) cut basis, while the price for finished beef was suggested on the same basis by key informants who were prominent beef dealers and cattle finishers in Iringa and Mbeya Regions. Choice implementation and data analysis Six experiments were conducted, specifically, one in each rural area (Mafinga and Tunduma Townships) and two in each urban centre (Iringa Municipal and Mbeya City) in February and March 2012 (Figure 1). The experiments were conducted in two sessions in one day starting with HCBC, whereby the cheap talk technique was used to ensure that consumers are sincere about what they are supposed to do and avoid haphazard responses (Lusk et al., 2007). The second session was conducted for RCBC whereby in addition to a cheap talk script; consumers were obliged to purchase one of the ten choices made using their own money (Ding, 2007; Nandonde et al., 2013a, Nandonde et al., 2013b). The experiments were conducted on weekends (Saturday and/or Sunday) and all sessions took 2 to 3 h with a 30 min break in between sessions. The filled-in choice forms were submitted to the researcher by the respondents before proceeding with the next session, and the subjects were not aware of the upcoming session. Consumers were given adequate time to concentrate on the choice by not asking them about their socio-economic background because this information was gathered after the completion of the choice experiments. Choice data J. Agric. Econ. Dev. 327 Table 1. Choice alternatives with attributes and attribute levels. Choice alternatives for beef with attribute levels Non-finished Local finished Crossbred finished Chilled Chilled Chilled Not Chilled Not Chilled Not Chilled Attributes 1. Chilling 2. Adipose fat content Medium High Medium High Low Medium 3. Tenderness Low Medium Medium High Medium High 4. Hygiene of outlet Clean Not clean Clean Not clean Clean Not clean 5. Price (Tshs/kg) 4,500.00 5,000.00 5,000.00 5,500.00 6,000.00 6,500.00 Table 2. Parameter (utility) estimates from choice sessions and pooled data. Choice Chilled Medium fat High fat Medium tender High tender Clean Price (x1000Tshs) Constant (ASC) Choice experiment session Hypothetical Real Coefficient P-Value Coefficient P-value 0.431 0.000 0.331 0.000 0.419 0.000 0.423 0.000 -0.117 0.178 -0.066 0.457 0.229 0.002 0.159 0.020 0.444 0.000 0.37 0.000 0.752 0.000 0.417 0.000 0.171 0.000 -0.317 0.000 -2.69 0.000 0.23 0.412 Log likelihood* Respondents (n) Observations -5588.58 308 9240 *Log likelihood Ratio test = -2[lnLpool – (lnLhypothetical+lnLreal)] = 146.97 -5724.8 308 9240 Preferences regarding product attributes and price The utility model estimates for all attributes in two choice Coefficient 0.378 0.415 -0.096 0.188 0.399 0.578 -0.074 -1.203 P-value 0.000 0.000 0.121 0.000 0.000 0.000 0.024 0.000 -11387 616 18480 . were stored in spreadsheets and analysed using a panel logistic regression in STATA 12 for utility model estimation. The model estimates were used to calculate the WTP and its standard error using the Jackknife procedure with the nlcom command in STATA (Hole, 2007). Detailed information on respondents’ socioeconomic characteristics can be found in Nandonde et al. (2013b). For the focus of the present paper we only considered interaction of product attributes with consumer’s gender, age, income and residence. RESULTS AND DISCUSSION Pool sessions and the pooled (combined) data are shown in Table 2. The preference directions for non-price attributes were the same in all choice sessions, indicating that consumers prefer chilled, medium adipose fat, medium tender and high tender beef sold in clean premises, regardless of whether the experimental set up was hypothetical or real. This result also implies the existence of one extreme favourite regarding eating and safety attributes. This finding might be attributable to the involvement of authentic consumers as opposed to typical laboratory experiments where students or less experienced consumers are involved in making choices (List, 2011). The market attribute (price) at hand had a positive and a negative coefficient in hypothetical and real choices, respectively, and both coefficients are significant (Table Nandonde et al 328 Table 3. Choice responses for different beef alternative sources. Statistic summary Hypothetical Observations Choices made % of choices made Real purchase Observations Choices made % of choices made Source of cattle for the carcass Local non finished Local finished Crossbred finished 3080 3080 3080 748 1187 1145 24.29 38.54 37.18 3080 1018 33.05 3080 1204 39.09 3080 858 27.86 Total 9240 3080 100.00 9240 3080 100.00 Precentage (%) Choice session Figure 2. Percentage of selected alternatives in two choice sessions with pooled data. 2). This result implies that beef product with higher prices had a higher chance of being chosen in hypothetical choices and that the price was affordable for most consumers. Despite the directional similarity and differences of non-price and price attributes, respectively, the log likelihood ratio test shows that the magnitudes of the estimated coefficients in two choice sessions were not equal (p<0.05), thus rejecting the null hypothesis of equal coefficients. This result is different from the findings obtained by Carlsson and Martinnson (2001) regarding public goods. The direction of the coefficient of the alternative specific constant (ASC) parameter (Table 2) does not have a meaningful interpretation in our case; however, its significance in HCBC implies that consumers made their choices motivated by alternative sources rather than by its quality contents. The ASC in connection with the positive price coefficient in HCBC indicates that finished beef was highly preferred to non finished beef. In HCBC, choices made for beef from local finished cattle and crossbred finished cattle outnumbered the status quo (Table 3 and Figure 2). This result was different in RCBC, where choices were made with respect to attributes and consumers were sensitive to price (Table 2), as also indicated by the sharp increase and decrease of the percentage of choices made for the status quo and crossbred finished cattle with a slight increase of finished local cattle (Figure 2). This finding is in keeping with reports from researchers (Ding, 2007; Lusk et al., 2008; Miller et al., 2011; Cecchi and Bulte, 2013), suggesting that consumers become more price sensitive and expose their true behavioural experience in the choice of the product and price bargaining under real market pressure. Cheap talk has been found not to exert adequate pressure on subjects to reveal their true preference for price compared to when real money was involved. This result is similar to the finding by Chowdhury et al. (2009). In the context of preference direction, it is the price that was adversely affected compared to other attributes that remain almost constant. This result implies that the utility of the price depends on the nature of the choice session and consumers do not appear to be genuinely concerned J. Agric. Econ. Dev. 329 Table 4. Implicit price (WTP) estimates for different sessions and the pooled data. Variable Attribute Chilled Medium fat High fat Medium tender High tender Clean outlet 1 WTP 2518 2453 -685 1339 2598 4396 Choice session Hypothetical 1 se P-value WTP 733 0.001 1059 613 0.000 1335 634 0.280 -208 650 0.039 501 1009 0.010 1168 1225 0.000 1316 Real se 211 372 264 203 227 241 Pool P-value 0.000 0.000 0.430 0.014 0.000 0.000 1 WTP 5089 5589 -1287 2534 5374 7781 se 2298 2854 693 1098 2160 3475 P-value 0.027 0.050 0.063 0.021 0.013 0.025 WTP1, estimates are in Tanzania shillings. Table 5. Comparison of WTP estimate differences. II Attribute Chilled Medium adipose fat High adipose fat Medium tender High tender Clean retail outlet Difference (%) I 57.98 45.58 69.63 62.58 55.04 70.06 Two sample t test t value P- value 1.9280 0.0562 1.5592 0.1175 0.6946 0.4876 1.2304 0.2189 1.3827 0.1673 2.4670 0.0139 III Pooled t test t value P- value 0.3174 >0.05 0.1959 >0.05 0.3441 >0.05 0.3816 >0.05 0.3310 >0.05 0.4431 >0.05 I Actual value difference for WTP estimates in HCBC and RCBC only. IIWTP estimate comparison for HCBC and RCBC only and as independent treatments. IIIt test determined with the inclusion of pooled variance. in no-purchase situations. Willingness to pay As expected, differences in coefficient estimates had a direct impact on the estimated WTP (Table 4). The decrease of the coefficients for attributes in RCBC had automatically lowered consumers’ marginal WTP (Table 5). However, in view of the t test for pooled variance with the assumption of independence of the choice session, none of the estimated WTP differences of an attribute was significant (Table 5), which implies a non WTP overestimation. This finding is similar to findings by Carlsson and Martinson (2001). By comparing two sessions as two independent treatments without considering the pooled variance, there was a different WTP estimation for hygiene (p=0.0139), implying an overestimation for the clean retail outlet attribute. This finding is different from that of Carlsson and Martinsson (2001) but is in line with other studies (Voelckner, 2006; Lusk et al., 2007; Alfenes et al., 2009), suggesting that consumers tend to overestimate their WTP in a hypothetical choice scenario. However, the overestimation of the clean retailing attribute in our case might be due to a too imaginary presentation of the hygiene attribute in the experimental set up because there was no visible reference for a clean and unclean outlet compared to other attributes that were directly attached to beef and that were presented virtually in real form through actual beef cuts (Nandonde et al., 2013a). The log likelihood ratio test in a choice model with the inclusion of socio-economic characteristics had shown coefficient inequalities (P<0.05) and a bias toward finished beef, as in the case when the analysis involves attributes and price only (Table 6). There was a general consistency in terms of the preference for chilled beef exhibited by female and more educated consumers, whereas urban residents had shown a preference for high fat beef in the hypothetical choice session; however, urban residents disfavoured the same choice in the real choice sessions. The observation on interactive utility estimates in real choices is similar to those addressed by Nandonde et al. (2013b), which implies that some behavioural differences are indeed contributed by consumer characteristics and might affect the willingness to pay for a specific attribute. CONCLUSIONS There is a choice bias in HCBC; however, this bias does not affect all estimated WTPs. Socio-economic characteristics of respondents contribute to the firmness Nandonde et al 330 Table 6. Choice estimates with socio-economic characteristics. Variable Choice Chilled Medium fat High fat Medium tender High tender Hygiene Chilled x Age Chilled x Gender Chilled x Education Chilled x Income Chilled x Residence Medium fat x Age Medium fat x Gender Medium fat x Education Medium fat x Income Medium fat x Residence High fat x Age High fat x Gender High fat x Education High fat x Income High fat x Residence Medium tender x Age Medium tender x Gender Medium tender x Education Medium tender x Income Medium tender x Residence High tender x Age High tender x Gender High tender x Education High tender x Income High tender x Residence Hygiene x Age Hygiene x Gender Hygiene x Education Hygiene x Income Hygiene x Residence Price (x 1000 Tshs) ASC Log likelihood* Participants Observations Hypothetical Coef. P>|z| 0.208 0.085 0.354 0.017 0.090 0.594 0.264 0.079 0.162 0.309 1.058 0.000 -0.106 0.253 0.355 0.000 0.205 0.043 -0.225 0.025 0.117 0.238 -0.012 0.913 -0.320 0.002 -0.076 0.515 0.029 0.798 0.395 0.001 -0.011 0.927 -0.451 0.000 -0.475 0.000 0.136 0.299 0.378 0.003 0.183 0.095 0.086 0.422 -0.020 0.866 0.058 0.626 -0.269 0.022 0.308 0.007 0.122 0.271 0.212 0.089 0.128 0.295 -0.185 0.127 -0.267 0.004 0.075 0.405 0.040 0.693 -0.038 0.705 -0.293 0.003 0.169 0.000 -2.704 0.000 Real Coef. 0.065 0.706 0.250 0.214 -0.081 0.435 -0.184 0.283 0.328 -0.153 0.142 -0.066 -0.012 -0.344 -0.019 -0.021 0.153 0.086 -0.516 0.119 -0.264 0.026 -0.229 0.034 0.022 0.039 0.134 -0.105 0.446 0.078 0.184 0.036 -0.042 0.052 0.054 -0.106 -0.322 0.239 -5545.8 308 9240 *Log likelihood Ratio test = -2[lnLpool – (lnLhypothetical+lnLreal)] = 98.29 of their decisions. RCBC can eliminate bias and exhibit the true preference for both price and non-price Pool P>|z| 0.587 0.000 0.133 0.140 0.608 0.000 0.045 0.002 0.001 0.122 0.145 0.531 0.909 0.003 0.867 0.849 0.187 0.447 0.000 0.343 0.031 0.811 0.029 0.773 0.850 0.736 0.239 0.343 0.000 0.523 0.131 0.695 0.638 0.604 0.587 0.273 0.000 0.395 -5680.91 308 9240 Coef. 0.133 0.517 0.159 0.233 0.038 0.737 -0.142 0.316 0.261 -0.184 0.127 -0.039 -0.161 -0.204 0.005 0.181 0.073 -0.171 -0.482 0.125 0.042 0.101 -0.071 0.005 0.036 -0.114 0.217 0.008 0.320 0.101 -0.002 -0.113 0.015 0.043 0.009 -0.196 0.000 -1.193 P>|z| 0.116 0.000 0.176 0.024 0.733 0.000 0.028 0.000 0.000 0.009 0.065 0.600 0.026 0.013 0.950 0.022 0.375 0.034 0.000 0.165 0.632 0.187 0.339 0.955 0.662 0.158 0.007 0.917 0.000 0.243 0.980 0.080 0.809 0.546 0.900 0.004 0.019 0.000 -11325 616 18480 . attributes. Further consumer behavioural studies should be conducted by employing advanced analytical models, J. Agric. Econ. Dev. such as latent class and mixed logit models, to capture more heterogeneity regarding consumers’ preferences and willingness to pay. ACKNOWLEDGMENTS The authors wish to acknowledge the German Academic Exchange Service (DAAD) in collaboration with the International Livestock Research Institute (ILRI) for their financial and advisory support. REFERENCES Alfenes F, Yue C and Jehsen H (2009). Cognitive th dissonance as a mean of reducing hypothetical bias. 4 Nodic Conference on behavioural and experimental economics. October 31-31, 2009. Oslo, Norway. Alfenes F., Guttormsen AG, Steine G and Kol-stad K (2006). Consumer willingness to pay for the colour of Salmon: Choice Experiment with real economic incentives. Am. J. Agric. Econ., 88: 1050-1061. Alphonce R, Alfnes F (2012). Consumer willingness to pay for food safety in Tanzania: an incentive-aligned conjoint analysis. Int. J. Cons. Stud., 36: 394-400. Carlsson F, Frykblom P, Lagerkvist C (2007). Preferences with and without price: Does the price attribute affect behaviour in stated preference survey? Env. Resource Econ., 38: 155-164. Carlsson F, Martisnson P (2001). Do hypothetical and actual marginal willingness pay differ in Choice experiments? J. Environ. Econ. Manage., 14: 179-192. Checci F, Bulte E (2013). Does market experience promote rational choices? Experimental evidence from rural Ethiopia. Econ. Dev. and Cult. Change., 61(2): 407-429. Chowdhury S, Meenakshi J, Tomlins K, Owori C (2009). Are Consumers in Developing Countries Willing-to-Pay More for Micronutrient-Dense Biofortified Foods? Evidence from a Field Experiment in Uganda: Contributed Paper prepared for presentation at the International Association of Agricultural Economists Conference, Beijing, China, August 16-22, 2009. [www.research4development.info/PDF/.../HarvestPlus_ Working_Paper_3.pdf] visited on 25.06.2010. Dawnay E, Shah H (2005). Behaviour economics: Seven principals for policy makers. New Economic Foundation. http://www.i-re.org/fiche-analyse97_en.html. Ding M (2007). An incentive align mechanism for conjoint analysis. J. Market. Res., 44: 214-223. Genon E, Chabanet C, Combris P, Issanchou S (2011). Are decision in real choice experiment consistent with reservation prices elicited with BDM auction? The case of French baguettes. Food Quality and Preference. Accepted manuscript. Hensher D, Rose J, Greene W (2005). Applied Choice Analysis A Primer. Cambridge, University Press. 331 Hole AR (2007). A comparison of approaches to estimating confidence interval for willingness to pay measures. Health Econ., 16: 827-840. Hole AR (2008). Modelling heterogeneity in patients’ preference for the attributes of a general practiotioner appointment. Health Econ., 27: 1078-1094. Jabbar M, Admassu S (2009). Assessing consumer preference for quality and safety attributes of food in the absence of official standards: The case of beef in Ethiopia. Contributed paper prepared for presentation at the International Association of Agricultural Economists Conference, Beijing China, August 16-22, 2009. Jabbar M, Baker D, Fadiga M (2010). Demand for livestock products in developing countries with focus on quality and safety attributes: Evidence from Asia and Africa. ILRI Research Report 24. Nairobi Kenya., p 168. Kallas Z, Gomez-Limon J, Arriazza M (2007). Are citizens willing to pay for agricultural multifunctionality? Agric. Econ., 36: 405-419. Lancaster K (1966). A new approach to consumer theory. J. Pol. Econ., 74: 132-157. List JA (2006). The behaviouralist meets the market: Measuring Social preferences and reputations effects in actual transaction. J. Political Econ., 114: 1. List JA (2011). Why should economist conduct field experiments and 14 tips for pulling one-off. J. Econ. Perspectives, 25(3): 3-16. Lusk J (2003). Effect of cheap talk on consumers’ willingness to pay for golden rice. Am. J. Agric. Econ., 85: 840-856. Lusk J, Schroeder T (2004). Are Choice Experiments Incentive Compatible: A test with quality differentiated beef steaks. Am. J. Agric. Econ., 86: 467-482. Lusk JL, Fields D, Prevatt W (2008). An incentive compatible conjoint ranking mechanism. Am. J. Agric. Econ., 90(2): 487-498. Lusk JL, McLaughlin L, Jaeger SR (2007). Stratergy and response to purchase intension questions. Market Lett., 18: 31-44. Miller KL, Hofstetter R, Krohmer H, Zhang JZ (2011). How should consumers’ willingness to pay be measured? An Empirical comparison of the state-ofthe-art approaches. J. Marketing Res., 48(1): 172-184. Nandonde SW, Msuya EE, Mtenga LA (2013b). Assessment of the influence of consumer characteristics on the choice of beef quality attributes. J. Agric. Econ. Dev., 2(3): 111-119. Nandonde SW, Msuya EE, Mtenga LA, Issa-Zakaria A (2013a). Assessment on consumer preference and willingness to pay for finished and chilled tender beef in Southern Highland Regions of Tanzania. Livestock Res. Rur. Dev., 25: 5. http://www.lrrd.org/lrrd25/1/nand25005.htm. Ortega D, Wang H, Wu L (2010). Modeling Heterogeneity in Consumer Preferences for Select Food Safety Attributes in China. Selected Paper prepared for Nandonde et al 332 presentation at the Agricultural and Applied Economics Association 2010 AAEA, CAES, and WAEA Joint Annual Meeting, Colorado, July 25-27, 2010. Radder L, le Roux R (2005). Factors affecting food choice in relation to venison: A South African example. J. Meat Sci., 71: 583-589. Thurstone I (1927) A law of comparative judgement. Psychol. Rev., 34: 273-286. Voelckner F (2006). An empirical comparison of methods for measuring consumer willingness to pay. Market Lett., 17: 137-149.