0
$\begingroup$

I'm looking to measure inflation in real time preferably minute to minute. What would be different feasible approaches to this? I know that the CPI measures inflation quite accurately but this is only on a month to month basis with the Bureau of Labour releasing their findings with a few month lag. The metric should be able to reflect a sudden inflationary spike like in the case of hyperinflation for example. Would much appreciate your help and thoughts!

$\endgroup$
1
  • $\begingroup$ What would you base your inflation numbers on? Most prices aren't updated minute to minute. Most prices aren't even updated daily. Hyperinflation is still quite visible in daily graphs, you don't need minute to minute date for this. Are you sure you're solving your actual problem? $\endgroup$
    – Mast
    Jun 16 '20 at 7:09
2
$\begingroup$

In addition to the +1 Brian's answer there is actually a way how to you could get hour by hour data. You can create your own dataset with webscraping.

Most of the datasets in the Bilion Prices Project mentioned by Brian are created through webscraping (See the details from Cavallo'spresentation on the project here). The way how they are able to get their daily data is that they webscrape them from a websites of retailers that are considered to be representative. You can do the same thing as well with python and enough computing power.

For a small country with a relatively small number of retailers doing so would not be prohibitively expensive and could be done even with a relatively small team on a budget. Although note that prices probably dont fluctuate much on an intraday basis so you should take that into account before investing too much effort and resources into such solution.

$\endgroup$
6
  • 5
    $\begingroup$ In the absence of hyper inflation, it would seem that hour by hour/week by week/etc "inflation" has no real macroeconomic meaning. $\endgroup$
    – Michael
    Jun 15 '20 at 18:44
  • $\begingroup$ @Michael I completely agree with you I do not necessarily think that doing this makes sense but who knows what kind of project the OP has in mind, the question just asked if it can be done so this would be a way of doing it. Whether it should be done is separate question and I completely agree as a macroeconomist that having data on such high frequency does not make much sense. Especially if other aggregates can’t be obtained at the same frequency. I can see use of the daily frequency for some univariate forecasting purposes though $\endgroup$
    – 1muflon1
    Jun 15 '20 at 18:49
  • $\begingroup$ Do you have any ideas that would not be as computationally expensive? $\endgroup$
    – MrClottom
    Jun 15 '20 at 19:43
  • $\begingroup$ @MrClottom i don’t think there is any other solution - although what I propose should not be outrageously computationally expensive. Some of my past colleagues are trying to copy the billion prices project for small EU country and they are all using their personal computers each focusing on different big online retailer and then they will add their data together. You will have more problems if you want to do it on hourly basis but I can’t think of any other way. Perhaps you could do it 4 times a day and then interpolate hourly data? I don’t expect that much interday variation in it so $\endgroup$
    – 1muflon1
    Jun 15 '20 at 19:48
  • $\begingroup$ that could work reasonably well. Or just try to consider lowering the frequency in general. I don’t know for what purpose you need the data but as @Michael pointed out for most macroeconomic applications hourly inflation data will be meaningless $\endgroup$
    – 1muflon1
    Jun 15 '20 at 19:50
5
$\begingroup$

The highest frequency real time data that I am aware of the Billion Prices Project, which is daily. Link to website. These are prices on the internet, which can be mapped to the US CPI relatively well.

Otherwise, consumer prices in bricks and mortar stores are Generally not set “minute by minute,” nor is there a way to sample them at that frequency.

The closest you can get is countries in hyperinflation, where prices end up being indexed to a foreign currency. That is, the price paid is a fixed price in a foreign currency, multiplied by the exchange rate. This mechanism results in rapid changes, even during the time to drink a coffee (anecdote from the Weimar hyperinflation). Foreign exchange rates are the usual way of estimating a hyperinflation, since there is no good way to survey prices.

$\endgroup$
2
  • $\begingroup$ How does the billion project calculate these values? $\endgroup$
    – MrClottom
    Jun 15 '20 at 17:42
  • $\begingroup$ They presumably have documents on their website? $\endgroup$ Jun 15 '20 at 18:34

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.