Frequently Asked Questions
- What categories do you cover?
- What are the names of the retailers?
- How many retailers do you monitor? How many items?
- What do you do with categories that you cannot monitor online, such as services? Do you ignore them?
- How do you decide which retailers to monitor & what data to use?
- How do you weigh different goods and categories?
- How often do you change the weights/retailers/categories in the index?
- Are the daily inflation series seasonally adjusted?
- What happens if a good appears or disappears from the online sample?
- Are you able to track core US CPI? If not, do you think that is an issue?
- Are online prices relevant?
- Are you trying to forecast monthly CPI announcements?
What categories do you cover?
Coverage varies from country to country, depending on the availability of online prices. In most countries, our index includes: food and beverages and supermarket products, household products, electronics, furniture, apparel, personal care items, over-the-counter medicines and gas prices.
Coverage is growing over time as more prices become available online. On average, the data we collect online can cover approximately 60% of CPI category weights.
Most of the categories that we are not able to cover are services. This, however, is not a problem for our goal to detect the main changes in inflation trends. Services are usually quite stable and not the main source of volatility.
What are the names of the retailers?
PriceStats does not share the names of individual retailers it monitors. We select our retailers carefully to include those with large market shares in relevant cities, and we focus on brick-and-mortar stores that also have an online presence (as opposed to retailers that sell exclusively online).
How many retailers do you monitor? How many items?
We do not disclose the specific number of retailers or products in the sample because this represents a key aspect of our methodology. Still, clients should know that the number of products included is never below 80,000, which is typically much more than what official indicators have (in the US, for example, the BLS uses 80,000 monthly prices while we use 800,000 daily).
What do you do with categories that you cannot monitor online, such as services? Do you ignore them?
Even categories whose prices we do not collect online are accounted for in our daily country-level index. The reason is that we proxy for them with goods or services that have been shown to behave similarly in the past. The details of how we do this are proprietary information, but the intuition is simple: for example, we may not have “education,” but books and office supplies tend to behave similarly to tuition (e.g. BLS, 1-2008 to 3-2011, books +16%, tuition +18%). That is why in general our indexes co-move closely with all-items CPI.
How do you decide which retailers to monitor & what data to use?
When we design an online index, we use the historical data we have collected over the years to evaluate the ability of specific retailers and categories of goods to match aggregate official statistics like the CPI. We study how our prices behaved in the past, select retailers and categories that have matched aggregate CPI trends in the past, identify patterns that are idiosyncratic to each retailer and separate them from more relevant aggregate patterns.
How do you weigh different goods and categories?
We try to use official CPI weights as much as possible, although we do make adjustments that account for the particularities of online data.
How often do you change the weights/retailers/categories in the index?
Once an index is created and released, we try not change the composition and weights. The reason is simple: we want clients to know that if they observe a new pattern in the data (e.g. prices start to rise faster than before), they can be confident that this is information coming directly from the prices—i.e. the state of the economy—and it is not something caused by an arbitrary change in methodology. Having said that, statistical offices change their baskets periodically, and we are also increasing our own coverage of categories over time, so we may have to make changes to some weights in the future. We will announce these changes to clients as they occur.
Are the daily inflation series seasonally adjusted?
PriceStats country series are not seasonally adjusted. There are two reasons for this. First, for some countries we still do not have enough of a historical record to identify seasonal patterns. Second, different statistical offices around the world use different methods for seasonal adjustments. If we adjusted our series, we would have to decide whether to match the method used in each country, or select one and apply it for all countries. In both cases this could lead to confusion. We think a better approach is to provide the non-seasonally adjusted index and let our clients apply their own adjustment if they wish to.
What happens if a good appears or disappears from the online sample?
While the retailers and categories of goods are not constantly changing, the mix of individual products may be doing so all the time. Within each retailer and category, we collect all available individual products we can find online each day. This means that new items may appear or disappear on a daily basis.
We treat all individual products as separate series, without making product substitutions or hedonic quality adjustments. Only consecutive price observations for exactly the same product are used to calculate price changes. So, for example, if a TV is replaced with a new, more expensive model, we do not have a price change in that category. Only when the new model starts changing its price will the index start to be affected by that product.
Similarly, when a product disappears from the sample, we assume it is temporarily out of stock for a set amount of time. After that period, the item is discontinued from the index.
Are you able to track core US CPI? If not, do you think that is an issue?
We focus on headline inflation for several reasons. First, we want to maintain consistency across countries. Not every country follows a core, and those that do use different definitions for what the “core” is. Second, in the case of the US, food and energy prices are well represented online, so the quality of our headline indicator is theoretically better than any core index that we could potentially construct. Finally, we think headline inflation will become increasingly relevant in policy making, even at the central bank level.
Are online prices relevant?
Our research shows that online prices are a successful measure of inflation, despite online price sources being different to those of official inflation estimates (baskets of online goods may not include, for example, service prices). In countries where internet retailing is a small share of transactions, retailers tend to tie their online prices to offline prices in order to minimize costs associated with making pricing decisions. In countries where online transactions have a significant market share, online and offline prices are closely related; online prices may lead offline prices since the demographic of computer-literate, middle-class, cash-rich and time-poor consumers to which online retailers sell is less price-sensitive than the average consumer. As purchasing habits change, these online price series capture prices and changes in product information and can provide real time information on consumer behavior.
Are you trying to forecast monthly CPI announcements?
No. Our indices are designed to anticipate major changes in inflation trends, not forecast monthly CPI announcements.
At any point in time, our indices will tend to be different from the CPI. We welcome those differences because they provide valuable information. In fact, they allow us to better anticipate changes in inflation rates, not only because we access the data sooner, but because our online data has repeatedly shown that it can react faster to shocks than the CPI.