Will e-commerce make prices more flexible?

Yuriy Gorodnichenko, Oleksandr Talavera, Slavik Sheremirov 21 January 2015



Today, it’s hard to imagine the world without the internet. PewResearch Internet Project reports that the internet is used by 87% of American adults, up from 14% in 1995. Apart from changing the way we communicate, connect, or acquire information, the internet has also changed our shopping habits. With just a few clicks, one can buy almost anything online and get it delivered promptly! Not surprisingly, the growth of e-commerce has been phenomenal. Virtually non-existent 15 years ago, e-commerce sales stood at $263.3 billion and accounted for 5.6% of total retail sales in the US in 2013.

Furthermore, the internet offers seemingly limitless opportunities to the retail sector by enabling sellers to collect and process massive amounts of data to tailor prices and product characteristics to specific whims of consumers and ever-changing economic conditions. A popular view (e.g., Kannan and Kopalle 2001) holds that prices for goods and services sold online should approach—if not now, then eventually—the flexibility of auction or stock prices. Indeed, the internet makes it trivial to compare prices across sellers, the cost of posting a new price is minimal, the best price is just a few clicks away, the physical location of online sellers is largely irrelevant, and numerous services advise online shoppers on best time and location of the purchase. Should one expect extinction of sticky prices then?

In a recent paper (Gorodnichenko et al. 2014), we provide new evidence on the nature and sources of price dispersion and frictions in price adjustment using data from a leading online shopping platform on daily prices for more than 50,000 goods in 22 broadly-defined consumer categories in the US and the UK between May 2010 and February 2012. We document properties of online prices (frequency of price adjustment, price synchronisation across sellers and goods, size of price changes) and compare our findings to results reported for price data from conventional, brick-and-mortar stores. Our dataset is unique in a number of ways.

  • First, it covers an exceptionally broad spectrum of consumer goods (precisely defined at the level of unique product codes) and sellers, enhancing comparability with brick-and-mortar stores.
  • Second, it contains daily price listings over the period of nearly two years, allowing us to study high-frequency variation in prices, which is especially important for e-commerce.
  • Third, each price listing comes with data on the associated number of clicks, which serves as a proxy for demand and relevance to consumers.

New evidence on price stickiness and dispersion online

We find that despite small physical costs of price adjustment, the duration of price spells in online markets is about 7 to 20 weeks, depending on the treatment of sales. While this duration is considerably shorter than the duration typically reported for prices in brick-and-mortar stores, online prices clearly do not adjust every instant. We also find a low synchronisation of price changes by a seller across goods and for a good across sellers; by and large, price changes are independent from each other.

The median absolute size of price changes in online markets, another measure of price stickiness, is 11% in the US and 5% in the UK, which is comparable to the size of price changes in offline stores. Sales in online markets are about as frequent as in conventional stores but considerably smaller; the share of goods on sale is approximately 1.5–2% per week and the average size is 10–12% in the US and below 6% in the UK.

We observe ubiquitous price dispersion in online markets. For example, the standard deviation of log prices for narrowly defined goods is 23.6 log points in the US and 21.3 log points in the UK, which is similar to, if not larger than, price dispersion in brick-and-mortar stores (e.g., see Sheremirov 2013, Kaplan and Menzio 2014). Even after removing seller fixed effects, which proxy for differences in terms of sales across stores, the dispersion remains large, suggesting that lower costs of monitoring competitors’ prices do not necessarily lead to price convergence across sellers.

We also show that price dispersion cannot be rationalised by product life cycle. Specifically, a chunk of price dispersion appears at the time of product introduction, which then grows (rather than falls) as the product becomes older. Price dispersion is best characterised as spatial rather than temporal. In other words, if a store charges a high price for a given good, it does so consistently over time—rather than alternating between low and high prices.

To emphasise price listings that are more relevant to consumers, we also calculate and present all these measures weighted by clicks. Such weighting tends to yield results consistent with a greater flexibility of online markets; price rigidities decline, cross-sectional price dispersion falls, and the synchronisation of price changes increases. For example, using weights reduces the median duration of price spells from 7–12 to 5–7 weeks. Yet, even when we use click-based weights, online markets are far from being completely flexible.

Dynamic pricing: Anticipated and unanticipated demand shocks

To shed new light on the use of dynamic pricing – instantaneous price adjustment in response to a change in demand or supply conditions by online retailers – we consider different ways through which it can affect price flexibility.

  • First, we look at the reaction of prices to low-frequency anticipated variation in demand due to holiday sales such as Black Friday and Cyber Monday in the US or Boxing Day in the UK.

In each country and year, the number of clicks goes up and the average price goes down during the holiday sales. This finding is consistent with Warner and Barsky (1995), who find that brick-and-mortar stores choose to time price markdowns to periods of high-intensity demand. Uneven price staggering may affect the timing of optimal monetary policy response to changing economic conditions, similar to the Olivei and Tenreyro’s (2007) argument that uneven wage staggering makes monetary policy more effective in the first half of a year.

  • Second, we show that there is a large high-frequency variation in demand, proxied by the number of clicks, over days of the week or month.

For example, Table 1 reports that the number of clicks on Mondays is substantially larger than on Saturdays. Yet, online prices appear to have little, if any, reaction to these predictable changes in demand, which is inconsistent with the Warner-Barsky hypothesis.

Table 1. Intraweek variation in prices and clicks

  • Finally, we do not find strong responses of online prices or demand to the surprise component in macroeconomic announcements about aggregate statistics such as the GDP, CPI, or unemployment rate.

These findings are striking because online stores are uniquely positioned to use dynamic pricing.

Concluding remarks

In summary, our main result is that online prices (especially prices with a large number of clicks) are more flexible than prices in conventional stores. Yet, the difference in properties of online and offline prices is quantitative rather than qualitative. That is, despite the power of the internet, the behaviour of online prices is consistent with smaller but still considerable frictions, thus questioning the validity of popular theories of sticky prices and, more generally, price setting. By some metrics, prices of goods sold online could be as imperfect as in regular markets.

These findings have a number of implications:

  • Even if e-commerce grows to dominate the retail sector, price stickiness is unlikely to disappear because it does not seem to be determined exclusively by search costs and/or physical costs of changing a price sticker.
  • Policymakers should not disregard the effect of e-commerce on properties of the aggregate price level and inflation as pricing in online markets does differ from that in brick-and-mortar stores.
  • Macroeconomists should put more effort into developing theoretical models with alternative mechanisms that generate price stickiness, dispersion, and other imperfections.


Gorodnichenko, Y, V Sheremirov, and O Talavera (2014), “Price Setting in Online Markets: Does IT Click?”, NBER Working Paper 20819.

Kannan, P K, and P K Kopalle (2001), “Dynamic Pricing on the Internet: Importance and Implications for Consumer Behavior”, International Journal of Electronic Commerce 5(3): 63–83.

Kaplan, G, and G Menzio (2014), “The Morphology of Price Dispersion”, NBER Working Paper 19877.

Olivei, G, and S Tenreyro (2007), “The Timing of Monetary Policy Shocks”, American Economic Review 97(3): 636–63.

Sheremirov, V (2013), “Price Dispersion and Inflation: New Facts and Theoretical Implications”, Working Paper available at

Warner, E J, and R B Barsky (1995), “The Timing and Magnitude of Retail Store Markdowns: Evidence from Weekends and Holidays”, Quarterly Journal of Economics 110(2): 321–52.



Topics:  Industrial organisation Productivity and Innovation

Tags:  E-commerce, price flexibility, price setting, online markets

Associate Professor in the Department of Economics, University of California – Berkeley

Professor of Finance at the School of Management, Swansea University

Economist in the Research Department, Federal Reserve Bank of Boston