What really is inflation?

I understand that you measure inflation by first measuring what the average person buys. Then you compare the total cost of these goods in year x with year y to get the inflation (from x to y).

The problem: Not only the cost of goods, but also the quantity that people buy changes all the time. Things go out of fashion or are replaced by different (better?) things. When you calculate inflation, you implicitly assume that the amount of goods people buy is constant (during the measurement period, say one year).

I do not doubt that this is reasonable from the practical perspective. But I doubt that today's inflation measurement approximates a well-defined concept.

Assume you know for every day/month/year how much people buy from each product and what it costs, what would be the formula for inflation? How do you incorporate the fact that people might by apples at 1\$in May and oranges at 2\$ in June, i.e. that costs and quantities change all the time?

• There are whole books on this topic. As an introduction I recommend Eurostat's freely available publication European Price Statistics - An overview. – Giskard May 4 '18 at 13:24
• The weights are periodically updated, following several methodologies. There is a ton of resources on that, for example: bls.gov/cpi/questions-and-answers.htm – Tobias May 7 '18 at 12:34
• @Tobias Thank you, but the question is more about the problems of using last years basket with today's prices. – J Fabian Meier May 7 '18 at 12:57
• You can have a look at how much consumption shares change. The answer is not a lot. Granted, there are many technical challenges in measuring inflation. I'd say changing baskets is not one of these; It's more difficult to deal with new products, or products that are not priced. – Tobias May 8 '18 at 13:35