Using price elasticity of demand to optimize price increases

Question

I want to share some context with you I am working for a pharmaceutical company and the business side needed to update the price for some products once a year. So I thought to develop a model to estimate price elasticity of demand. The justification it that indicating which products have a inelastic behaviour can guide them to prioritize products to increase price or which products are not interesting to have a price increase.

So I developed a model to calculate the price elasticity of demand of the client's portfolio using Double/Debiased Machine Learning. The code uses historical data of the last 2 years moving window so I can bring always a fresh value of elasticity at each run.

The feedback was positive but they asked to have a recomendation of price increase as well. My idea is:

1. Select products that have elasticity coefficients below 1 cause they are inelastic and could receive price increases.
2. For these products, estimate the revenue effect for a variety of price increases for the next year.
3. Get the price that gives you the max revenue as price increase recommendation.

You think this idea is viable? The estimated decrease in demand is proportional to the price increases?

Answer 1

Let's assume that the company, as may well be the case in the pharmaceutical industry, has some degree of pricing power for its products (in other words, is not subject to so much competition that it is a price-taker). In principle it is appropriate for such a company to have regard to price elasticity of demand when setting a price for its product. However, several issues should be considered.

Firstly, is the company aiming to maximise revenue or profits? If the latter, then the change in production costs arising from any change in sales volume consequent upon a change in price also need to be considered. At least some change in sales volume is likely, since very few products are completely price-inelastic

Secondly, the available data from which to estimate elasticity may be limited. Using data from a moving 2 year window may seem sensible, to keep estimates up to date, but if, say, the price of a product is reviewed once a year, such a window will yield sales volume data at just two prices, which is a very limited basis from which to derive a curve as in your diagram. On the other hand if estimation is from data over a longer period, then the third issue below will become even more significant.

Thirdly, any conclusions drawn from price-volume data are subject to the qualification that all other things are equal, which may not be valid. Sales volume at any given price may be affected by, for example, changes in demand due to changes in the incidence of the medical condition for which the product is used, changes in public perceptions of the efficacy of the product, changes in consumers' income, and so on.

A pragmatic approach, given these complications, may be to adjust the price incrementally. In other words, instead of changing it directly to the level at which the estimated curve suggests revenue (or profit) would be maximised, it could be changed a fraction of the way towards that level for a trial period, and the effect on sales volume monitored. If the outcome were satisfactory, a further increment might then be considered.