1. 1 Economics of Innovation Welfare gains from product innovation Manuel Trajtenberg 2005

2. 2 Reminder: social benefits from process innovation D MC 1 MC 0 P0P0 P1P1 P X X1X1 X0X0 Griliches: social returns from hybrid corn – 700%

3. 3 Product Innovation D1D1 MCP0P0 P X X1X1 X0X0 D0D0 D1D1 P0P0 P X X1X1 X0X0 D0D0 P1P1 But what if product differentiation, choose just one?

4. 4 The framework The “magnitude” of product innovation between two periods: the increment in consumer surplus associated with having the latest choice set (S t ) rather the previous one (S t-1 ) : If just one product x and one quality dimension z, i.e. additional area under demand function, brought about by  z.

5. 5 The MNL and the surplus function The surplus function (exponent: for linear utility):

6. 6 Estimation - Endogeneity Estimate MNL for differentiated products in the market. Issues: IIA, and: Serious problem: endogeneity of price, i.e. price (positively) correlated with unobserved quality (which goes into the error term), hence upward bias in estimation of  – critical (see surplus function)! In study of CT Scanners, finding of upward slopping demand curve, for that reason; correction using residuals from hedonic price function. But later on whole literature on how to cope with the problem – BLP etc.

7. 7 Key issue in Medicine: ability to visualize internal organs. X-rays (~1900), ultrasound (1950s), gamma camera (1960s). All of these: rather crude. Otherwise: cut open! Hounsfield began work on CT at EMI, in 1967. First installation: 1973. First only head, then also body scanners. Godfrey Hounsfield and Allan M. Cormack (Tufts University) got the Nobel Prize in Medicine in 1979. Since then: MRI, fMRI, PET, angio CT, etc. Computed Tomography (CT) Scanners

8. 8 First Generation CT Scanners

9. 9 Second Generation CT Scanners

10. 10 Third Generation CT Scanners

11. 11 Performance Characteristics of CT Scanners CT Scanners: highly complex systems, yet main attributes: Scan time: how long it takes to scan one thin cross-sectional “slice” (minutes at first, down to 1 sec) Spatial Resolution – size of smallest object that can be just visualized in best of conditions (but there is more to image quality). Reconstruction time: from end of scan to image display.

12. 12 Diffusion of CT Scanners by Hospital Size

13. 13 CT Scanners: Market Indicators

14. 14 Firms in CT

15. 15 Prices and characteristics of CT Scanners

16. 16 CT Scanners: sales, firms, and R&D (R&D: for US firms only, in constant 1982 millions $)

17. 17 Reminder: the framework

18. 18 Estimating the Nested MNL Estimate two clusters, one for Head (H) the other for Body (B) scanners, If = 1 then back to the MNL, if = 0 then separate “markets”.

19. 19 Estimates for Head Scanners

20. 20 How to compute  W Given changes in tastes over time, two ways to compute the welfare gains from year to year: “Dual inducement” mechanism, hence in practice found ex-ante > ex-post

21. 21 Annual incremental gains (  W) to representative user

22. 22 How to compute “total gains” from product innovation? One angle of the issue: Consumer buying PC today: should we take into account only the latest innovations? Or should we say that he/she is benefiting from the long sequence of innovations since say the first computers (the ENIAC) on? Or perhaps just since the first Apple? Another angle: Should we just multiply  W t times to number of buyers at t? Or perhaps times the number of projected buyers from t on?

23. 23 How to compute “total gains” – cont. Estimate diffusion as a function of e.g.  W t that is, innovation moves up the ceiling, bringing in new consumers. Assign to these additional consumers the benefits  W t. (but discount them back to t)

24. 24 Computing total gains

25. 25 Total Gains and R&D Social rate of return: 270%