Data Sources

To study the effect of each of the market entry determinants, we obtain data from the follow- ing sources:

(1) Annual editions of the Orange Book (also known asApproved Drug Products with Ther- apeutic Equivalence Evaluations) from 2000 to 2016;

(2) Clinformatics Data Mart;

(3) National Drug Code (NDC) Directory; (4) DrugBank database (Version 5.0); (5) FDA Inspections database;

(6) Implementation of the Generic Drug User Fee Amendments of 2012 (Woodcock 2016). The Orange Book is used to identify drug markets that are subject to potential generics entries. Clinformatics Data Mart is used to construct proxies for market profitability. We link these two databases using the National Drug Code from the National Drug Code Directory. The DrugBank database is used to obtain characteristics of drug products such as molar mass and indications of the active ingredient. The FDA Inspection database and the FDA report on Implementation of the Generic Drug User Fee Amendments are used to construct proxies for manufacturing quality and the ANDA backlog, respectively.

We obtain patent data, exclusivity data, and approved drug products data from annual editions of the Orange Book. The Orange Book is published by the FDA and identifies the complete set of branded innovator products and generic products approved by the admin- istration. We use patent and exclusivity expiration dates associated with the branded drug applications to identify generics entry opportunities. Each approved product in the Orange Book with application type “A” corresponds to a generics-related application. For these ob- servations, the FDA provides the applicant name, the approval date, the reference branded product, as well as characteristics of the drug product, including number of active ingredi- ents, route of administration, and strength. Based on its past approval history, we are also able to construct a firm’s experience with the ingredient and the firm’s experience with the route.

To measure market size, we obtain the annual prescription charge and quantity for the reference branded drugs from Clinformatics Data Mart, provided by Optum, Inc. The orig- inal data source of the Clinformatics Data Mart comes from a national US private health insurer. We cross validate the data with data from the Centers for Medicare & Medicaid

Services (CMS) and Drugs.com.5 According to the Drug Listing Act of 1972, drug products are required to be identified and reported using a universal product identifier, the National Drug Code (NDC). At the level of the nine-digit NDC, we retrieve the annual total claim counts and total charges from Clinformatics for 2001 through 2015. The nine-digit NDC can be used to uniquely identify a drug product. We rely on the NDC Directory to merge the Clinformatics data with the Orange Book.6

In order to better assess market profitability and production difficulty of each drug prod- uct, we obtain additional characteristics of the drugs from the DrugBank database. This database provides extensive biochemical and pharmacological information about drugs mar- keted in different countries (Law et al. 2013). Ayvaz et al. (2015) map all other drug databases to the DrugBank database to study the overlaps between various data sources, due to the DrugBank database’s broad inclusion. This latest version of the DrugBank database (Version 5.0) contains 2,021 FDA-approved small-molecule drugs, 233 FDA-approved bio- logical drugs, 94 nutraceuticals and over 6,000 experimental drugs (DrugBank 2017). For each identified active ingredient, the DrugBank database provides the molecule type (small- molecule ingredient or biological protein), the molar mass, the structured indications ex- tracted from the FDA drug labels and scientific publications, as well as the therapeutic class of the active ingredients, which is indicated using the Anatomical Therapeutic Chemical (ATC) classification code.

To obtain a proxy for manufacturing quality, we obtain facility inspection records since October 1, 1999 from the FDA Inspections database through a Freedom of Information Act (FOIA) request. We focus on inspections that received final classifications. An inspection classification reflects the compliance status of the manufacturer site at the time of the in- spection. The conclusions are reported as Official Action Indicated (OAI), Voluntary Action Indicated (VAI), or No Action Indicated (NAI). FDA concludes an inspection with OAI if significant objectionable conditions or practices were found and the firm must take regu- latory action to address the issues. Contrarily, a VAI classification indicates that the FDA revealed objectionable conditions, but the issues were not significant enough to warrant reg-

5_{We are unable to disclose the name or the market share of the insurer due to a non-disclosure agreement}

with the data provider. However, we cross validate it with CMS’s Part B and Part D data as well as the top 100 drug list from Drugs.com. We do not directly use the Part B National Summary Data File from CMS because there are only a limited number of branded drug products covered under the Medicare Part B program. The majority of the prescription drugs are covered under the Part D program; whereas the longitudinal annual revenue and quantity data for the drug products covered in the Part D program is not available. The top- 100 drug list (2003 – 2013 by sales amount and by sales units), available from Drugs.com, is compiled from QuintilesIMS, a company that provides proprietary data on the total sales and volumes of drug products. We find that the two data sources are consistent for those best-selling drugs.

6_{We retrieve the NDC for the branded drugs no longer marketed in the United States from past NDC data}

ulatory actions; whereas an NAI classification indicates that the FDA found no objectionable conditions.7 Besides the final classification, inspection records also provide us with the list of the Code of Federal Regulations (CFRs) a facility violates.

To study the impact of an ANDA backlog on a firm’s entry decision, we obtain the annual number of pending generics applications since 2000 from the FDA report on Implementation of the Generic Drug User Fee Amendments of 2012 (GDUFA) (Woodcock and Wosinska 2013). After the implementation of the GDUFA, more detailed ANDA submission informa- tion and progress became available since 2015; however, this level of information was not documented by the FDA and thus not available via FOIA request.