U.S. Pharmaceutical Markets: Expenditures, Health Insurance, New Products and Generic Prescribing from 1960 to 2016

Abstract

Between 1960 and 2016 per capita expenditures on retail pharmaceuticals increased from $85.12 to $922.50 in 2009 prices, while the share of spending covered by insurance programs grew to 86%. The introduction of new molecular entities followed only a weak upward trend. Following passage of the Hatch Waxman Act in 1984, the share of retail prescriptions dispensed as generics has expanded from 19% to almost 90%. By employing a three-equation Vector Auto-Regressive (VAR) model and Granger causality, this paper assesses empirically the interdependence among these variables. Our most important finding is that increases in insurance coverage on pharmaceuticals contributed significantly to the spending growth, which in turn led to further insurance expansion. In addition, introduction of new drugs also has contributed to enhanced insurance coverage. While the expansion of generic prescribing has had a strong negative effect on spending, this effect has been largely offset by rapid increases in coverage.

Keywords:

• Insurance
• innovations
• pharmaceutical expenditures
• generics
• Granger causality

JEL CLASSIFICATIONS:

• C32
• C51
• I11
• L65

Notes

Acknowledgements

We would like to thank Ted Frech (the Editor), Joel Hay, and Jack Needleman, as well as seminar participants at the UCLA Seminar on Pharmaceutical Economics and Policy, and at the Leonard Davis Institute of Health Economics, University of Pennsylvania, for valuable comments.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1 These data were obtained by dividing the nominal expenditures, collected by the National Health Expenditure Accounts (NHEA; Center for Medicare and Medicaid Services 2018), and the annual population estimates (U. S. Census Bureau 2018), followed by applying the GDP deflators (Federal Reserve Bank of St. Louis, Economic Research Bureau 2018).

2 Because we set a maximum 8 year lag (to match the lag length in the full VAR model) in the bivariate cross-correlation analysis, the Dickey-Fuller test was performed on time series data from 1976 to 2016.

3 A significant caveat is in order. The longer term effectiveness of these interventions is uncertain. With the introduction of many novel and effective new drugs, particularly those developed and produced through genetic engineering, payers’ ability to negotiate lower prices for these drugs may be more limited (Frank and Newhouse 2008; Frank and Zeckhauser 2017).

4 Originally the U.S. Bureau of Labor Statistics (BLS) treated branded and the related generic drugs as separate products. Since 1995, however, the original branded product and its follow-up generic alternative are considered as different versions of the same product. As a result, the upward bias prevalent in earlier years has been significantly reduced since 1996 (Berndt and Aitken 2011; Bureau of Labor Statistics 2011).