The Problem with Price Elasticity
Most people working in finance, retail or pricing will likely have encountered the term ‘price elasticity of demand’ (PED) at some point. PED is simply a measure of how demand for a good changes when the price is altered. On the surface, this seems to be the panacea for optimal pricing allowing you to precisely set prices that maximise revenue or margin. It can, however, be incredibly difficult to apply in practice. This can lead to unwanted results that can dramatically affect your bottom line.
What’s it all about?
PED enables us to understand how much a change in price affects demand (volume sold), and by how much. The more elastic the good, the larger effect price has on volume (more price-sensitive), conversely the more inelastic a good, the less price influences volume (less sensitive). For example, essential goods tend to be inelastic as they are seen as a necessity and luxury goods tend to be very elastic. Great, we have a method of optimising the price we set for a given product and can also find goods we can increase price on without affecting the level of sales — increasing gross profit. Unfortunately, in practice, it’s not quite that simple…
In order to calculate PED (plot the above graph), we need;
- Historical (but recent) transactions for an elected product
- Multiple, different price points
- A ruler / steady hand
This gives us an overview of price level and respective quantity for the product, this can be useful to check for any outliers in the data and to get a quick understanding of the data we are using.
This is not PED however, in order to calculate ‘price point PED’ for each unique price point, calculate the change in quantity (ΔQ) as a percentage of the total quantity (ΔQ/Q) and divide it by the change in price (ΔP) as a percentage of the price (ΔP/P) . This denotes the percentage change of quantity to a percentage change in price, plotting the response in quantity to a change in price provides an average elasticity score.
The gradient of the line of best fit tells us about the elasticity of the good, the flatter the line, the more sensitive a good is to price. In other words: by how much will sales drop off when we increase the price, this denotes the ‘elasticity’ of a product and therefore to what extent is the price a factor in the consumer’s decision to buy a product. Price elasticity is almost always negative due to the innate relationship between price and volume (we don’t like buying overpriced things).
This is obviously a very simplistic view on the world; price is very much not the only factor that drives demand, and the points we plot will not be a straight line. It can be a powerful tool however if used in line with its limitations, there are many revenue and seasonality effects for example that can be explained well by PED.
What’s the problem then?
Data. Fundamentally our conclusions are only as good as the data we’re basing them on. The better the data describes the behaviour of the product the more reliable the price elasticity calculation. Here lies the crucial flaw in using PED, it is highly unlikely that the data is of high enough quality to carry out robust calculations.
In order to fully understand the price elasticity of a product, we need large amounts of historical data with multiple, different price points. In practice, most retailers, online or otherwise will not have accumulated enough price data to generate enough corresponding demand data. The difficulty is in obtaining data that correctly reflects the market the good is bought and sold within.
To achieve this one tends to minimise the time period over which price changes are monitored. This is to remove other contributing and potentially conflicting factors that affect demand such as seasonal effects. Unfortunately, the shorter the time period, the fewer the price points which can dramatically reduce our confidence in the measurement. Balancing quality against quantity is key to generating an accurate PED.
To illustrate this we can imagine plotting the graph above based on only two price points. The line we draw will almost certainly be wrong as we are making assumptions about all price points from just two data points. In practice, the actual relationship between price and quantity could be any curve between these two points. Therefore, in order to be confident in our elasticity calculation, we need to have an extensive distribution of price/volume data mapping out products’ behaviour.
This is exacerbated by the fact that such a simplistic view of PED doesn’t lend itself to accurate outcomes. There is an assumption here that price changes sales volume directly and proportionally (straight line). In practice there lies a ‘Goldilocks zone’ where a product is consumer viable, either side of this it is either too cheap to make any profit or just too expensive for consumers to consider purchasing.
The elasticity of a typical retail product, therefore, is more sigmoidal in shape. This leads us to ask some difficult questions, where on the sigmoid curve should we take a gradient measurement? How do we denote the PED of a nonlinear elasticity curve without very complex expressions? Fundamentally, fitting a line to a non-linear curve is problematic.
Another familiar term is cross elasticity price of demand (XED), this represents the change in demand of a product given a change in the price of another. This complicates product-level price testing as the change and effect is not isolated to the product in question. Substitute and complementary goods will antagonistically respond to a price change adding an additional layer of complexity to a simplistic PED calculation.
The preconception that all products will obey these rules is a potential oversight too, there are many determinants of PED that contribute to pricing behaviour. When we aggregate on a brand or sector level, this becomes less of an issue[more data, more general], this is why PED is usually cited in macroeconomic definitions. These pricing rules are correct when applied generally, products will be sensitive to price to a greater or lesser degree, PED indicates the extent to which this is the case.
The potential of measuring elasticity on a grander scale is more impactful simply due to its reliability, it describes a type of good at its demand level which is extremely useful for predicting market trends, forecasting P&L etcetera. The significance of the individual product is reduced when observed through this lens lending itself for overall business analysis not individual product pricing for typical samples of retail pricing data.
The final problem is very much a practical one. Often the datasets used to calculate PED are not representative of true market behaviour. Retailers often do not change prices over a long period of time and if they do they often do not change them by enough to accurately describe the price vs demand space. This leads to elasticity calculations that do not represent actual market conditions and as such can have a detrimental effect if used to inform pricing decisions.
So, what should I do?
Fundamentally, in order to understand the effect of price on demand we require as many different prices as possible, which simply takes time. Using different retail prices in different sales channels can speed up this process as long as the markets behave similarly. Without this data, we lack the key information required to optimise price. ‘Price testing’ enables us to gather data over a range of different prices to build up a reliable, high-quality data set to measure product level behaviour and make sounder decisions.
This can be challenging in the retail space, as there are issues restricting such testing as minimum advertised prices (MAP) to adhere to, catalogue/bricks and mortar price matching and nett margin targets which restrict the movement available. Even so, if the price is within a competitive range and is relatively elastic, the optimal price point might be small but highly impactful.
TLDR;
Using PED to derive product pricing decisions without enough high-quality data is ill-advised. Without frequent, good and diverse price changes there is limited information to be gained regarding elasticity.
Mapping out the price — demand space for products can be very useful, but practically very difficult, there are much more effective approaches for optimising price in a real-world retail space than PED driven price changes.
Regularly changing the price of products, even over a short period will start to generate robust data about the price sensitivity of demand, regardless of how you use this insight, the data itself contains a wealth of information on both a product and aggregated level which can be used to optimise price, drive customer behaviour analysis, inventory segmentation and more.
Learn more via our blog at BlackCurve: https://blog.blackcurve.com/
