Table 1 Time Parameter, Correlation Coefficient, and Shift in WTP Parameter Estimates for Bivariate Probit Models
 Random Time Parametera (RTPw/o)Random Time Parameter with Observed Characteristics for Time Parameterb (RTPw)Fixed Time Parameter with Observed Characteristics for Time Parameter (FTPw)
Mean Time Parameter Estimate
Exponential Embedded Image0.068 (0.18)0.128 (0.03)**0.195 (0.00)*
Hyperbolic Embedded Image0.171 (0.79)0.269 (0.25)0.455 (0.06)***
Standard Deviation of Time Parameter Estimate from Unobserved Heterogeneity
Exponential (σρ)0.378 (0.00)*0.131 (0.00)*
Hyperbolic (σμ)1.128 (0.07)**0.119 (0.00)*
Correlation Coefficient between the Initial and Follow-Up Questions (τ)
Exponential−0.263 (0.10)***0.126 (0.50)0.319 (0.05)**
Hyperbolic−0.238 (0.20)0.078 (0.69)0.261 (0.09)***
Structural Shift in Utility with the Follow-Up Question (δ)
Exponential−0.315 (0.05)**−0.337 (0.02)**−0.281 (0.03)**
Hyperbolic−0.338 (0.05)**−0.381 (0.02)**−0.272 (0.04)**
Number of observed characteristics for the time parameter0817
LRI
Exponential0.1730.2140.287
Hyperbolic0.1730.2240.287
AIC
Exponential558.68552.96529.26
Hyperbolic558.78547.26529.49
  • Note: p-values for the estimates shown in parentheses next to the estimates. We round p-values < 0.001–0.00.

  • a The distribution of the random time parameter is normal. The FTPw/o mean time parameter estimate is zero, and the p-value is one in the exponential and hyperbolic models. The estimate of the time parameter in the FTPw/o model is only noise around zero without observed heterogeneity.

  • b A RTPw model with more than eight observed characteristics for the time parameter cannot solve with a negative semi-definite Hessian to recover the standard errors of the coefficients.

  • * p < 0.1;

  • ** p < 0.05;

  • *** p < 0.01.