Table 4.
Censored quantile regression estimates (n = 147). Numbers in brackets are standard errors.
Variable Regressions
15% quantile 50% quantile 85% quantile
Intercept 0.276 4.533⁎⁎⁎ 6.415
(1.876) (0.515) (8.363)
Objective knowledge of food safety −0.112 −0.016 0.232
(0.250) (0.033) (0.827)
Salience −0.305⁎⁎⁎ −0.002 0.757⁎⁎
(0.091) (0.028) (0.388)
Fear of negative social evaluation 0.058⁎ 0.005 0.141
(0.032) (0.008) (0.078)
Trust 0.206⁎⁎ 0.028⁎ 0.295
(0.099) (0.017) (0.358)
Self-reported habit 0.027 −0.009 −0.173
(0.04) (0.006) (0.177)
Note: Dependent variable is auction bid per kilogram of kale. Censored quantile regression using the Portnoy (2003) estimator with jackknife samples (Portnoy, 2014).
⁎
P < 0.10.
⁎⁎
P < 0.05.
⁎⁎⁎
P < 0.01.
Table options
Across the quantiles, it can be seen that the bids were not significantly dependent on OKN and habit. The results also suggest that the influence of OKN was rather constant at a low level across quantiles. For the other determinants there was more disparity. For the habit measure, the estimates indicated a tendency for higher levels of habit to have a more pronounced negative effect in the top quantile. Salience, on the other hand, proved to have a relatively strong negative effect in the lowest quantile, but a strong positive effect in the top quantile. This suggests that the influence of salience on bids corresponds to an asymmetrically S-shaped relationship. The social pressure variable had a positive effect upon bids in each quantile, but the coefficient was significant only for the lowest quantile. The implied functional form was weakly quadratic, however. Finally, trust had a stronger positive effect upon bids for the lowest quantile and, although the implied functional form was U-shaped, the estimates for trust were non-significant except for the median quantile.