Introduction

We examine the nature of competition within the US market for prescription drugs. For generic drugs, one should generally expect a competitive market with multiple producers making each drug. Moreover, since many drugs can be offered and taken in multiple configurations (see example and definition below), competition must manifest itself at the configuration level (i.e., how many producers make each configuration). For instance, 21 firms offer the antidepressant drug Venlafaxine HCL which is available in 3 forms (Tablet, 24 Hour Extended Release Tablet, and 24 Hour Extended Release Capsule) and 8 dosage concentrations (from 37.5 mg to 225 mg). 12 configurations of this drug are offered in the market. However, each configuration has offerings from fewer than half of these 21 manufacturers, as seen below.

(** It might be nice to have an alternative visualization of the nature of competition – not just the number of manufacturers but WHICH manufacturer makes each configuration? **)

We seek to examine this market behavior at a larger scale using a dataset covering 389 distinct drugs. A “configuration” of a drug represents a specific product type in which it can be sold. We employ two alternate defintions:

The configuration set for a drug is all the tuples or configurations which are offered in the market. We use the two separate definitions because of is it ambiguous whether “Package Size” is a rigid element of a drug specification. Prescriptions are typically filled by a pharmacist who can, in principle, make any package size; however, only a handful of package sizes are commonly used in practice. Moreover, while the package size of many drugs can be customized, others (such as creams and ointments) are indeed available in pre-defined package sizes.

Data Sample and Properties

The research is based on product and price data for the years 2000-2020. Each observation in this dataset implies either a price change for an existing configuration from a specific producer, or the introduction of a configuration by a producer.

NDC Category Generic_Name Qty Dosage Mfr_Name Form LOB AWPPrice WACPrice AWPDate WACDate awp_per_30ds wac_per_30ds Label_Name Brand
43598021040 Dermatology SILVER SULFADIAZINE 400 1% DR.REDDY’S LABORATORIES, INC. Cream Commercial 0.16 0.13 2012-05-24 2012-05-24 68 56.9 SSD CRE 1% Brand
70010049205 Antidiabetics METFORMIN HCL 500 750MG GRANULES PHARMACEUTICALS Tablet Extended Release 24 Hour Commercial 1.08 0.07 2019-01-16 2019-01-16 53 3.6 METFORMIN TAB 750MG ER Generic
00002879959 Antidiabetics INSULIN LISPRO 3 100/ML LILLY Solution Pen-injector Medicare 39.36 32.80 2016-07-15 2016-07-15 534 445.9 HUMALOG KWIK INJ 100/ML Brand
16714045601 Antipsychotics QUETIAPINE FUMARATE 100 300MG NORTHSTAR RX Tablet Commercial 16.75 1.34 2017-03-17 2017-03-17 794 63.5 QUETIAPINE TAB 300MG Generic

Level of Competition and Market Separation

** We need to learn about the drug industry: what is the nature of drug approval for an off-patent off-exclusivity (and hence generic) drug. Does granting an ANDA (abbreviated new drug application) to a manufacturer automatically confer on them the right to make all configurations?

** Also, for some drugs if we notice there’s only one producer it could well be that there aren’t more than one approved manufacturers for that drug. It could even be that only the brand holding the patent is presently approved to make the now generic drug.

Typically, a generic (and therefore off-patent) drug is likely to be made by multiple approved manufacturers. The number of manufacturers – and hence degree of competition – can depend on several market and product factors that affect potential profit, including technical difficulty in producing the drug, variable and fixed costs, and market willingness to pay. In general, it is hard for us to comment on the correct degree of competition for any drug. However, conditional on the number of manufacturers for a drug (i.e., they make at least one configuration) we can still draw conclusions regarding the level of effective competition in the market. The idea is that if all manufacturers (of a particular drug) were to make all configurations, then they would be competing relatively more fiercely, implying lower profit for all. However if they make only a few of the possible configurations, then they can differentiate, leading both to higher prices and more price dispersion as manufacturers charge “whatever prices they want to charge in different settings”.

We can see below the number of drugs, number of manufacturers - and how they can produce “differentiated” products by varying elements of a drug configuration: the Package Size (Qty, e.g., 30 pills), Dosage (or concentration, e.g., 10 mg) and Form (e.g., capsule).

Number of Manufacturers and Drugs

For the time period considered (years 2000-2020) we have 365 unique drugs - and 247 unique manufacturers in the data set. Let’s look at how many manufacturers are making each drug (i.e., they make at least one configuration for that drug).

Of the 365 drugs in the dataset, we can see that just over half have only 1 or 2 producers making that drug (even though it is a generic drug); but also that almost a third of drugs have at least 5 manufacturers. Specifically, 34.25 percent of drugs have at least 5 manufacturers. Now, let’s look at the level of competition for each configuration of the drug.

Competition within Configurations

Let us examine the number of Manufacturers for each “drug configuration” - i.e., the same (Generic_Name, Dosage, LOB, Form). Note: the “Qty” (or package size) is not included as part of the “Configuration” because prescription drugs are often packaged for the customer (* do we know this to be true ? *); but at any rate we’ll check the results vs including “Qty” in the next step.

Specifically, 25.27 percent of drug configurations have at least 5 manufacturers.

** Actually it would be nice to have a CDF chart which shows the cumulative densities (number of competitors) for BOTH cases in the same plot (for Drugs and for Configurations).

Configuration = (Generic_Name, Qty, Dosage, LOB, Form)

Let us examine the number of Manufacturers for each “drug configuration” - i.e., the same (Generic_Name, Qty, Dosage, LOB, Form).

The above picture shows that there is a high degree of market segmentation and product differentiation by manufacturers. For a majority of drug configurations (i.e., fixing Dosage and Quantity and Form) there is only one manufacturer – even though multiple manufacturers can make the drug, and do in fact but in a different configuration.

IGNORE below here

Model of Competition and Prices for Prescription Drugs

Competition vs Pricing

There are 2 primary ways in which prescription drug prices are defined in our data set: 1) Average Wholesale Price (AWP), 2) Wholesale Acquisition Cost (WAC). Both AWP and WAC represent a baseline for price negotiations between sellers (pharma companies) and buyers. Although these are not the prices paid by consumers nor even necessarily by payors (insurance firms) they are considered industry benchmarks for drug pricing. In addition, there are two secondary measures of price, representing AWP (or, correspondingly, WAC) for a 30 day supply. Regardless of the price measure, the existence of a price observation is informative regarding the level of competition for a drug.
Now let’s compute the “median average deviation” for each group (i.e., for the same drug across multiple manufacturers).

Regression of dispersion against count. We should expect a negative slope: higher number of manufacturers should lead to less dispersion.

## 
## Call:
## lm(formula = dispersion ~ log(count), data = n.Mfr)
## 
## Residuals:
##    Min     1Q Median     3Q    Max 
##    -81    -40    -25    -25  10821 
## 
## Coefficients:
##             Estimate Std. Error t value Pr(>|t|)    
## (Intercept)    24.97       6.64    3.76  0.00017 ***
## log(count)     20.98       6.41    3.27  0.00108 ** 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 288 on 3440 degrees of freedom
## Multiple R-squared:  0.0031, Adjusted R-squared:  0.00281 
## F-statistic: 10.7 on 1 and 3440 DF,  p-value: 0.00108

Configuration = (Generic_Name, Dosage, LOB, Form)

Now suppose you define a configuration as just (Generic_Name, Dosage, LOB, Form), and count the number of manufacturers by including all manufacturers of a particular configuration even if they are selling in different package sizes.

## 
## Call:
## lm(formula = dispersion ~ log(count), data = n.Mfr.Qty)
## 
## Residuals:
##    Min     1Q Median     3Q    Max 
##    -86    -63    -61    -37  10785 
## 
## Coefficients:
##             Estimate Std. Error t value Pr(>|t|)    
## (Intercept)     61.3       12.9    4.76  2.1e-06 ***
## log(count)       8.5       11.1    0.76     0.44    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 416 on 2004 degrees of freedom
## Multiple R-squared:  0.000291,   Adjusted R-squared:  -0.000207 
## F-statistic: 0.584 on 1 and 2004 DF,  p-value: 0.445