Hedonic Regression Results Sample Clauses

Hedonic Regression Results. We estimate the hedonic regression model in equation 3.3 using ordinary least squares. We show the regression results in the section. Each coefficient estimate for the hedonic regression represents the marginal contribution of each characteristic to the log price of videoconferencing. The exponential of each coefficient can also be interpreted as the WTP for the said characteristic. For videoconferencing (see figure 3.B.1), it can be noticed that not all coefficients are sta- tistically significant. This suggests that the presence of some characteristics probably does not contribute substantially to the variations in prices across plan types and/or across ser- vice providers. We also observe the presence of negative coefficients that are statistically significant. If we were to interpret each coefficient as the marginal contribution of each char- acteristic to the price, it stands to reason that none of the variables should have a negative value for their coefficients. We offer two likely explanations for this. First, it is important to note that the coefficient estimates are partial elasticities and that we can only arrive at the marginal contribution of each characteristic by applying exponential transformations to the coefficients. In this case, the transformation would yield a positive value that is close to zero. Second, ▇▇▇▇▇▇▇▇ (2016) shows that it is possible for hedonic regressions to generate negative coefficient estimates if there are trade-offs between the characteristic with the nega- tive coefficient and other characteristics in the regression. For instance, the trade-off between horsepower and mileage could result in negative coefficient estimates in a hedonic regression for automobiles. In the case of this exercise, only Encryption yielded a negative coefficient that is statistically significant. An examination of the correlation between covariates (see appendix 3.D) shows that the presence of encryption is negatively correlated with some of the statistically significant explanatory variables in the hedonic regression19. One can argue that the presence of these features makes it difficult to make calls more secure. A major limitation of the panel hedonic regression is that it assumes that the marginal values of characteristics are fixed over time. It is possible that this assumption may not be true. From the descriptive statistics in table 3.4.1, we show that the average price per participant varies across years. We generate a second regres...