Data Preparation and Constructed Variables

In this section, we provide details of data source, data cleaning, missing data and sample selection details, where appropriate.


Monthly sales data from May 2009 to June 2014 (62 months) of all SKUs of 105 regulated formulations was extracted from IMS Health (now IQVIA) (provided by the pharmaceutical firm). There were a total of 195 SKUs covering the 105 oral regulated molecules. We removed 16 SKUs due to reasons including – i) no sales values (8 SKUs), ii) no sales values in post DPCO period (3 SKUs), iii) missing values over more than 40 months and where available sales values vary very widely (2 SKUs), iv) list of brands under SKU inconsistent over time (2 SKU2), and v) change in classification resulting in inconsistent sales (1 SKU).

Type of Molecule (Acute)

This drug characteristic is at the molecule level (which is then applied to all SKUs falling under that molecule). The type of molecule (acute/chronic) indicates the type of illness for which the molecule is typically prescribed (categorized and provided by the pharmaceutical firm). This is a time invariant metric and is recorded at a molecule level. We have a total of 65 acute formulations and 40 chronic formulations in our dataset. 

Sales % from Tier1 and Tier 2 Cities (T1T2%)

We have data on this drug characteristic at the molecule level. T1T2% refers to the percentage sales percentage sales from tier 1 (metros) and tier 2 cities. IMS Health India (now IQVIA) categorizes Indian cities into four groups – metro, class 1, class 2-4 and rural. We sum the percentage sales for each molecule (105 molecules) in Tier 1 and Tier 2 to form T1T2%.

Prescription % from CP/GPs (CPGPRx%)

We have data on this drug characteristic at the molecule level. CPGPRx% refers to the percentage prescriptions by CP/GPs (consultant physician/general physician), indicating the type of doctors that typically prescribe the drug. For instance, Imatinib, a medicine prescribed to cancer patients has on average 4.7% CPGPRx%, whereas Oflaxacin, an antibiotic has on average CPGPRx% of 75%.

Sales Data from the Philippines

The quarterly sales data of all SKUs from the Philippines was extracted from IQVIA International. We have quarterly sales data on a total of 2,324 molecules amounting to 46,221 SKUs in the Philippines.

Sample Selection for Difference-in-Difference Analysis

We have sales data on 179 SKUs in India. We mapped the identical SKUs from the Philippines dataset. We dropped 95 SKUs from our sample – i) 12 SKUs – data was missing for several quarters, ii) 18 SKUs – either data was unavailable or the exact composition was not available in the Philippines, iii) 32 SKUs – while composition matched, the strength was not a match, iv) 16 SKUs – these drugs were classified as ‘other’ drugs in India (eg. other hiv drugs, other cns drugs, etc…), and v) 17 SKUs – Part of MDRP (Maximum Drug Retail Prices) and GMAP (Government Mediated Access Price) – price controls in Philippines. Hence, we use 84 SKUs in the difference-in-difference analysis.

Sample Selection for Generalized Synthetic Control Method

We used the entire Philippines sales data as a donor pool to create controls for the regulated SKUs in India. We have 2,324 molecules amounting to 46,221 SKUs in the Philippines. First, we dropped the molecules that were covered under MDRP and GMAP (17 SKUs). Second, we dropped all injections, ointments, syrups and other liquid drugs from the dataset, resulting in 22,429 SKUs. Third, we dropped SKUs that had missing values (for any quarter in our analysis period), resulting in 747 SKUs. Thus, we created a donor pool of 747 SKUs from the Philippines.

Data from Top Pharmaceutical Firm

Brand Level Data – Detailing and Sales

The pharmaceutical firm provided the detailing visit data for 27 regulated brands and 27 unregulated brands in its portfolio. The firm’s headquarters plans the total number of visits (termed as exposures) to be covered by its salesforce for each brand on a monthly basis. The number of visits are computed for – i) new doctors, ii) non-core doctors (these doctors are visited only once in a month), and iii) core doctors (more than one visit per month). The firm provided monthly data on detailing visits from April 2012 to June 2014 (15 months of pre-regulation and 12-months of post-regulation data) for the two sets of brands. The firm also provided brand level sales for the 27 regulated and 27 unregulated brands for the same time period. There were no missing data.

Prescriptions from physicians with no formal medical degrees (NonMBBSRx%)

The pharmaceutical firm has enlisted the services of a market research firm to provide prescription data on a monthly basis. The market research firm has a panel of doctors (those who prescribe allopathic medicines) across the country. The firm shared the percentage of prescriptions for a drug written by Non-MBBS physicians (physicians who prescribe allopathic drugs but do not have a formal medical qualification – NonMBBSRx%) from May 2009 to June 2014. Prescription data is available only for 51 of the regulated molecules (of the total 105 molecules under regulation). There was no missing data.