Working Paper: NBER ID: w1008
Authors: Manuel Trajtenberg; Shlomo Yitzhaki
Abstract: Studies of diffusion have traditionally relied on specific distributions-primarily the logistic- to characterize and estimate those processes.We argue here that such approach gives rise to serious problems of comparability and interpretation, and may result in large biases inthe estimates of the parameters of interest. We propose instead the Gini's expected mean differenceas ameasure of diffusion speed, discuss its advantages over the traditional approach, and tackle with it the problems of truncated processes, inter-group comparisons, and related issues. We also elaborateon the use of the hazardrate, and suggest some possible extensions. The diffusion of CT scanners is presented as an illustration.
Keywords: No keywords provided
JEL Codes: No JEL codes provided
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| Traditional methods relying on logistic distributions (C46) | Significant biases in estimating diffusion speed (C22) |
| Estimating a logistic model when the true distribution is exponential (C51) | Substantial downward biases in speed estimates (C51) |
| Gini's expected mean difference (D31) | More accurate measure of diffusion speed (C69) |
| Using the Gini measure (D31) | Consistent comparisons across different diffusion processes (C22) |
| Government regulations (L51) | Speed of adoption of CT scanners (O33) |