Price on demand (Price Elasticity) has been much debated within the aviation industry. Despite being something of a hot topic, there’s yet to be an agreed understanding of how Price Elasticity can be effectively measured, or how airlines could gain significant commercial value by proactively managing it.
For those unfamiliar with the concept of Price Elasticity, it’s representative of how many more passengers would fly (as a %) if the price drops by a certain amount (as a %).
Assessing Price Elasticity, and the potential impact price changes have on demand, is extremely important because it can guide the choice of aircraft type when airlines place multi-billion dollar orders. While not the only consideration, we feel understanding Price Elasticity removes a significant risk from any airline’s business plan.
Using Skyscanner data, we are able to measure the Price Elasticity of a group of routes which have received a lot of industry attention lately - Transatlantic low cost.
The Price Puzzle
The main difficulty with Price Elasticity is that it needs to be derived from the number of people who did not end up travelling at a certain price point, rather than those who did. While the industry has developed some advanced algorithms to measure the number of passengers who have flown, most models have generally disregarded information on the number of passengers who did not fly. There have been some efforts by the industry, based on information on the number of passengers that have flown, to measure the unconstrained demand by looking at 'spill' models. However most of those theories are hard to validate, and we feel don’t offer the granularity necessary to truly measure Price Elasticity.
That’s why search data is the single largest advance in airline analytics in decades.
Measuring the number of travellers searching for a route who choose not to proceed further down the funnel is the best indicator of how satisfied users are with the options offered. When a traveller is presented with a consistent set of carriers for a specific route over a season, the only visible difference to the user is the price. Barring some noise and bias in the data*, the fluctuation in the number of unconverted searches related to the price displayed is therefore the Price Sensitivity.
However, search volumes can be influenced by destination marketing or user acquisition activities, which makes changes in search volume too unreliable, unless it is normalized. Fortunately, with Skyscanner data, we are able to go a step further and demonstrate intent by measuring the volume of 'click-throughs' on any given route as well. So, while a search may show “interest” from a user, our click through data shows “intent”, or at least a deeper level of interest. As a result, we'll make the assumption that a click-through correlates with a passenger travelling**, and that fluctuations in click-through rates on a specific route throughout the year will mostly be influenced by the price shown to the users, and therefore can be used to measure the first part of the equation: “how many more passengers would fly …?”. The equation for that would therefore be:
Modelling Price Elasticity
In order to measure Price Elasticity, we looked at a whole year of data. For each Origin and Destination (O&D) within the study, we needed to plot weekly aggregations (in term of travel dates) of the average price selected for travel during the period and the conversion rate for that same week.
We then needed to compare the change in conversion to the average conversion over the year and the average price to the average price during the year. The equation for this is:
For a year of data, and for each O&D in a large set, we plotted weekly aggregations (in term of travel dates) of the average price selected for travel during the period and the click-through rate for that same week. Once we had all the data points scattered, we ran a linear regression and measured the slope, which represented “the change in conversion rate for a change in the median price”. We then normalized the data, and the slope, so that it provided a ratio between the number of people travelling and the price displayed (“% change in traffic for a % change in price”).
While measuring the slope, it was also important to look at how much the linear regression matched the data. We used the R value (see full explanation here) to evaluate how reliable the measure of Price Elasticity is; we discarded O&Ds with an R value that was too small or positive (anything above -0.2).
To avoid having variance in the data due to small sample sizes, we limited our data to O&Ds that had at least 3000 searches and 100 exits per week for every travel week in the year 2016.
Low Cost & Demand Generation
One premise of the low cost model is the capacity to generate demand. This demand generation is composed of two phenomena: 1) the increase in interest through marketing and 2) a higher conversion of travellers interested due to cheaper pricing. We will ignore the former, since we are able to neutralize the influence of demand generating activities through the use of conversion rate data (as opposed to search data); and increasing demand can be measured separately through performance marketing (i.e. most online ads).
Looking at the potential for the model to impact revenue, it is important to consider that the revenue of a flight is the product of the number of people that booked, multiplied by the average fare per booking.
A strong Price Elasticity indicates that lower than average fares can be offset by more passengers on-board to result in higher revenue. It's important to note that from a profitability point of view, given that each additional passenger is associated with a unit cost (distribution, catering, insurance, compensations etc…), Price Elasticity needs to be significantly stronger than -1 to enable higher profit for the same cost base. It is also important to call out that each additional passenger is associated with a unit cost (distribution, catering, insurance, compensations etc…) and therefore for profit to be made with a consistent cost base, Price Elasticity needs to be significantly lower than -1 (negative and greater than 1 in absolute value).
Price Elasticity and Low Cost Long Haul : Challenge or Opportunity?
For a very long time, when accused of price dumping and uncompetitive behaviour, low cost airlines have argued that they don’t “steal” market share but rather create their own market. This is only half the picture, as most travellers are price sensitive and price is almost always the first criteria considered when comparing.
Low cost long haul has been tried without much success by a number of carriers in the past. Think Oasis Hong Kong airlines in Asia, Laker Airways in the UK, and Canadian based Zoom Airways. But now a new generation of aircraft and airlines like Norwegian, Scoot and WestJet seem to be successfully executing on the strategy. Only a couple of months ago, IAG announced the launch of a brand new BA sister airline, Level, promising flights from the UK to the US for as low as £99.
Transatlantic Low Cost: Price Elastic?
To evaluate if the Transatlantic market is Price Elastic we compared the Price Elasticity on Origins and Destinations (O&Ds) originating in Spain, United Kingdom, France, Italy, Germany and the Netherlands to Spain and the USA***. Spain was selected as it has the largest volume of inbound air tourism in Europe. The underlying assumption is that tourism to Spain is Price Elastic, but has already been stimulated over the last decade.
The Result: An Opportunity Ready in Waiting
X-Axis: The R-value is the correlation coefficient, which indicates how well a linear regression matches the data points. An R-value closer to -1 signifies better confidence in the Price Elasticity value.
Y-Axis: The Price Elasticity as explained above, the “% change in traffic for a % change in price”, a higher absolute value means that reducing prices on those O&D will result in a greater than proportionate increase of traffic.
Size: Number of searches used to calculate the values.
Shapes: Squares are for O&D with a US destination, which can be compared with circles, which are for Spanish (ES) destinations.
Colors: Country-to-Country O&D.
Our results show that Transatlantic leisure traffic to the US is indeed more Price Elastic than traffic to Spain. Since the market to Spain has already been stimulated by low cost airlines for decades this isn't particularly surprising. The Transatlantic low cost market on the other hand, has yet to be exploited with opportunities for carriers to take advantage.
Having said this, we haven't found any O&D with a Price Elasticity lower than -1. This indicates that overall Transatlantic revenue is questionable at best, and likely to decrease unless airlines significantly reduce costs, and are able to compensate for any losses with ancillary revenues.
In our analysis, ROM-NYC, has the highest Price Elasticity in absolute of all the routes. This means traffic is likely to increase dramatically with fare reductions. In fact, at the end of May 2017, Norwegian announced that they would launch a low cost service on this very route from Rome to New York Newark.
The Undeniable Importance of Price Elasticity
Price Elasticity is a valuable measure yet not fully part of airlines planning today. Comprehending this measure, however, is key to understanding where commercial opportunities lie - especially when offering low fares. With the growing number of carriers looking to expand their offering with low cost long haul, especially in the Transatlantic market, it now seems more pressing than ever to understand the data behind Price Elasticity.
Because Skyscanner's data suite, Travel Insight, is able to capture interest and intent data, the travel search engine is best placed to conduct advanced analysis.
Want to learn more about Price Elasticity and how understanding it could help your business?
* This can be affected by elements such as group size, which can fluctuate during the year and can influence how many searches are made to get to a result.
**This was confirmed by a third-party analysis comparing click throughs and GDS bookings and justifies the metasearch business model. To neutralize the fluctuations of the total search volumes we used the conversion rate defined as: number of click throughs divided by the number of searches.
***It's worth noting that since the pricing data comes from the fares clicked through airport-to-airport, this analysis does not take into account traffic leakage and large potential O&D may be discarded because of travellers choosing an alternate airport (such as LYS-NYC as shown here).