Last week Google generated a flurry of interest about their idea to launch a Google Price Index (GPI) to track inflation. One example of the coverage is this article from the FT.
The idea has obvious appeal. Using real time shopping data, Google is able to publish data that is far more up-to-date than the government’s CPI.
Google was typically spare with details, but I see problems already. Let’s start with something simple. Say you are tracking the price of colas. (The beverage, not the cost of living adjustment.) Tracking price changes for colas is easy. But how do you account for the fact that when faced with rising prices, many consumers might shift from Coke to 7-Up? Or if prices for carbonated beverages rise, they might shift to fruit drinks. Or what if consumers just drink tap water.
What I am describing is substitution. It is not a new idea, but it is an idea that has yet to be fully incorporated into the government’s CPI; nor presumably would it be included in a GPI. (The fact that this is untracked is one reason the CPI is often seen as overstating inflation.)
The second problem is harder to resolve. Colas are simple products compared with, say, vehicles, which would surely be included in the GPI. Tracking the changes in sticker and transaction prices together with incentives is complex enough. How do you factor in shifting equipment levels? (Over the years, optional features tend to become standard.) And what about things like changes in quality and resale values?
The government makes an attempt to adjust for these factors, but I have to say I am skeptical about the results. Would Google—who are the master of the algorithm, but culturally opposed to manual manipulation—even try?
We have looked at an Edmunds Vehicle Price Index for years. So far, we are still looking. (What we are capable of doing today would not pass our data quality standards.)
Not to split hairs, but what Google is proposing is a measure of price trends, not an inflation index.