This is the title of the new USC white paper written by Darren Filson, Karen Van Nuys, Darius Lakdawalla and Dana Goldman, with “How much revenue drives new drug development?”
What is the flexibility of innovation?
It measures the percentage change in innovation (the flow approved by new drug, or the beginning of phase 1, 2 or 3) is caused by the percentage change in income, usually expected
Future income.
In fact, what matters is the change in profits, but future revenues are more observable and predictable than future profits. Therefore, the author focuses on innovation’s elasticity to revenue rather than profits.
How much will future revenue affect the possibility of new drug development?
All studies concluded that resilience is positive – IE, with lower revenue leading to less R&D – but estimates vary widely. However, we believe that the typical long-term elasticity associated with U.S. income is in the range of 0.25 to 1.5, which means every 10% reduction in expected revenue, we can expect a 2.5 to 15% reduction in drug innovation.
What drives the variability of these estimates?
A key question is why these estimates are so wide? Of course, different research designs are important (see below). The authors also claim that factors such as “time of time frame research, the scale of price changes, drug development costs, barriers to value-based pricing, and other market factors” will affect the elasticity of innovation.
What methods are used in the literature to estimate the elasticity of innovation?
- cross section: Use income differences in treatment categories (or other units of analysis) to estimate elasticity. For example, they can infer elasticity by comparing the “high income” vs. the “low income” category. [Examples: Lichtenberg (2005) and Civan and Maloney (2009)].
- Summary time series: Take advantage of changes in industry-level revenue over time [Example: Giaccotto, Santerre and Vernon (2005)]
- Panel data method: Includes drug-grade “fixed effects” and is difficult to measure and persistently variance in class characteristics. Essentially, the focus of this approach is on changes in the classroom, which is the driving force for innovative changes in the classroom. These analyses often require the use of “natural experiments”, which lead to differences in revenue variations across market segments. Examples of natural experiments include future demographic changes or the emergence of the Medicare section. [Examples: Acemoglu and Linn (2004); Dubois et al. (2015); Blume-Kohout and Sood (2013)]
- Parameterized computing model (also known as structural model): Specify the company's target functions, strategy sets, and characteristics of the business environment. When the model includes multiple companies, the model usually requires the market to be in an equilibrium state. Select parameters to match real-world parameters (e.g., average R&D expenses) and calibrate so that the model output also matches the real-world results (e.g., average flow of new drugs). [Examples: Abbott and Vernon (2007); Filson (2012); Adams (2021)]
The author believes that panel methods and parameterized computing models are the first choice.
For the study of preferred panels or calculation methods, what is the personal elasticity of their innovative estimates?
The author has a nice table that summarizes what I have found pasted below.
My colleagues have an amazing job at USC! I certainly encourage you to read the full text here.
