That is the title of a new White Book of the USC of Darren Fineson, Karen Van Nuys, Darius Lakdawalla and Dana Goldman with the subtitle «How much do new medication development boost?»
What is the elasticity of innovation?
It measures the percentage change in innovation, using the flow of new medicines, or phase 1, 2 or 3 beginnings, caused by a percentage change in income, generally expected.
Future income.
In practice, what matters is the change in profits, but future income is much more observable and predictable than future profits. Therefore, the authors focus on the elasticity of innovation regarding income instead of profits.
How much do future income affect the probability of development of new drugs?
All studies conclude that elasticity is positive, the lowest income leads to less R&D, but estimates vary widely. However, we argue that a typical long -term elasticity associated with US income is within the range of 0.25 to 1.5, which implies that for each 10% reduction in the expected income, we can wait 2.5% to 15% less pharmaceutical innovation.
What is promoting variability in these estimates?
A key question is why there is such a large range in these estimates. Certainly, different study designs matter (see below). The authors also affirm that factors such as «the time horizon studied, the size of price change, the cost of drug development, barriers to prices based on value and other market factors» impact the magnitude of the elasticity of innovation.
What methodologies are used in the literature to estimate the elasticity of innovation?
- Cross: Exploit the variation in income between therapeutic classes (or some other unit of analysis) to estimate elasticity. For example, they can compare the «high» versus «income» income «classes to infer elasticity [Examples: Lichtenberg (2005) and Civan and Maloney (2009)].
- Aggregate Time Series: Exploit the variation in income at the industry level over time [Example: Giaccotto, Santerre and Vernon (2005)]
- Panel data approaches: Include the «fixed effects» of drug class and the reduction of difficult and persistent differences in class characteristics. In essence, this approach focuses on changing income within the class as a driver of innovation changes within the class. These analyzes generally require the use of «natural experiments» that cause a differential change in income in different market segments. Examples of natural experiments include future demographic changes or the advent of Medicare Part D. [Examples: Acemoglu and Linn (2004); Dubois et al. (2015); Blume-Kohout and Sood (2013)]
- Parameterized computational models (also known as structural models): Specify the objective functions of companies, sets of strategies and characteristics of the business environment, and when the model includes multiple companies, the model generally requires that the market be in balance. The parameters are selected to coincide with those of the real world (for example, the average costs of R&D) and are calibrated so that the exits of the model also coincide with the results of the real world (for example, average flow of new medications). [Examples: Abbott and Vernon (2007); Filson (2012); Adams (2021)]
The authors argue that panel approaches and parameterized computing models are preferred.
For studies with the preferred panel or calculation approach, what individual elasticity of innovation estimates came?
The authors have a good table that summarizes the findings that I hit below.
Great work of my colleagues at the USC! I certainly encourage you to read the full article here.
