Are inflation figures true but misleading?

[stlrick]

With the inflation figures released today, the general tone of headlines is that inflation "cooled" to 2.8% in February from 3.0% in January. These numbers are correct, but misleading.

Here are the Consumer Price Index (CPI-U) figures that illustrate the problem:

January 2024.......308.417
February 2024.....310.326
December 2024...315.605
January 2025.......317.671
February 2025.....319.082

The 2.8% for February 2025 comes from comparing it to February 2024: 319.082/310.326 =1.0282 = 2.82%
The 3.0% for January 2025 comes from comparing it to January 2025: 317.671/308.417 =1.03 = 3.0%

The problem is that in comparing January and February 2025 to different values, it does not reflect the very high inflation in February 2024: 310.326/308.417 x 12 =12.0743. The annualized rate for February 2024 was 7.43%, so if February 2025 simply was less than that very high rate, inflation for this past month appears to be "cooling."

For the first two months of 2025, inflation is 319.082/315.605 x 6 =6.0661, an annualized rate of 6.6%.

[jebmke]

As far as I know, the reporting of inflation data has not changed recently.

What people choose to take away from the data is really just a personal choice.

Not sure if there are any reasonable actions to implement here.

Note, I should add that I don't look at historical data much -- even my own; it is generally too late to take any action in response to.

[stlrick]

I'm not saying it has changed.

What people take from the data depends on understanding how the data are calculated.

[rockstar]

I'm not saying it has changed.

What people take from the data depends on understanding how the data are calculated.

It’s what TIPS use, which is different from PCE. You can calculate your own to see how you compare to the headline numbers.

[delamer]

I'm not saying it has changed.

What people take from the data depends on understanding how the data are calculated.

Since when?

People take from the data what they think proves whatever economic or political point that they want to make. That’s been true forever and it is true today.

Sorry for the snark.

[SFAtty]

Inflation has a lot of seasonal factors. That is why the YOY data are generally highlighted as that avoids seasonality. Taking one or two months and then extrapolating that to 12 months is not going to give you an accurate (well, incomplete) inflation picture.

[runninginvestor]

You are referring to base effects. It's a common statistical/mathematical phenomenon when comparing differences. You can look into base effects and how it can impact numbers. You also hear that term be used from time to time in reporting.

[nisiprius]

Same problem as with all financial statistics: how to get a meaningful average of something that fluctuates, maybe fractally. Combined, of course, with all sorts of motives to spin the results.

My subjective impression is that inflation really was higher than normal during the period of "transitory" high inflation, and really has cooled to something that isn't way different from "normal" in my day-to-day life. For example, I'd anchored on $3/gallon as the "normal" price of gas and locally it's back to around that. Eggs are way up, obviously, but equally obviously they are an anomaly. Apples are running about the same as before, with of course considerable variation by variety.

Unless you keep very, very detailed personal records it's hard to be accurate about what's happen in your personal life, because of constant changes in what you choose to buy. Almost everyone pays very selective attention to the prices that have gone up, and not to those that have gone down.

[Everlast311]

That's known as the base effect. If you have a particularly high or low month rolling off, it might make the current reading look slightly better or worse, but it still reflects a rolling 12-month period. As for annualizing the Jan/Feb figures, the beginning of the year typically has higher monthly readings. Conversely, Nov/Dec are often low or even negative.

[stlrick]

Inflation has a lot of seasonal factors. That is why the YOY data are generally highlighted as that avoids seasonality. Taking one or two months and then extrapolating that to 12 months is not going to give you an accurate (well, incomplete) inflation picture.

You do not need to extrapolate to an expectation of what inflation will be for 12 months to use the annualized calculation to interpret what is happening presently. The headlines are stating that inflation is presently - this month - cooling. It is not presently cooling. At present, it has picked up.

[Thesaints]

The problem is that in comparing January and February 2025 to different values, it does not reflect the very high inflation in February 2024: 310.326/308.417 x 12 =12.0743. The annualized rate for February 2024 was 7.43%, so if February 2025 simply was less than that very high rate, inflation for this past month appears to be "cooling."

It makes little sense. Why are you saying Feb 24 inflation was very high ? With respect to what ?

For the first two months of 2025, inflation is 319.082/315.605 x 6 =6.0661, an annualized rate of 6.6%.

Yes and every time the s&p has a daily gain of 1%, your annualized return is more than 1,200% !

[stlrick]

The problem is that in comparing January and February 2025 to different values, it does not reflect the very high inflation in February 2024: 310.326/308.417 x 12 =12.0743. The annualized rate for February 2024 was 7.43%, so if February 2025 simply was less than that very high rate, inflation for this past month appears to be "cooling."

It makes little sense. Why are you saying Feb 24 inflation was very high ? With respect to what ?

For the first two months of 2025, inflation is 319.082/315.605 x 6 =6.0661, an annualized rate of 6.6%.

Yes and every time the s&p has a daily gain of 1%, your annualized return is more than 1,200% !

I don't understand why the OP has generated unpleasant responses - and such superficial misreading. First, your point based on belittling a calculation based on two months with a calculation based on one day is silly. Second, my point was not about predicting what the annual inflation will be for 12 months based on two months. My point was that the headlines say inflation is cooling. It is not.

[Thesaints]

First thing first: apologies if I sounded harsh.

Secondly, which inflation are you talking about ? If it is the annual inflation, that simply is the ratio of the CPI-U index measured 12 months apart.
If I read correctly, you are saying this leaves out the Feb24 index measurement (and all the measurements that cane before). That is as it should be: we are calculating the 12-month inflation, not the 13-month.
Comparing to the last 12-month measurement it went down. Comparing to previous 12 month periods, the number oscillates: 2.8% is higher than over some past 12-month periods and lower with respect to others.
Overall, notwithstanding the Fed declared 2% target, the number is on the low side.

As extrapolating an annualized value from any two consecutive months is concerned, that leads to a highly imprecise estimate; there are seasonal phenomena at play, as others have pointed out. There are non seasonal factors too: notably higher volatility index components. No doubt you have heard of the “CPI excluding energy and food”.

[toddthebod]

Inflation has a lot of seasonal factors. That is why the YOY data are generally highlighted as that avoids seasonality. Taking one or two months and then extrapolating that to 12 months is not going to give you an accurate (well, incomplete) inflation picture.

You do not need to extrapolate to an expectation of what inflation will be for 12 months to use the annualized calculation to interpret what is happening presently. The headlines are stating that inflation is presently - this month - cooling. It is not presently cooling. At present, it has picked up.

January 2023 - February 2023 had an annualized rate of 6.9%.
In 2022 it was 11.5%.

You are ignoring seasonality. This was the "best" January in at least the last 4 years.

This is why there are seasonally adjusted numbers. The annualized seasonal adjusted inflation for January 2025 was 2.6%.

Your post has generated "hostility" because you have accused the media or whoever of misrepresenting inflation by yourself misrepresenting inflation.

While I am not interested in calculating annualized rates for each and every month, if December's inflation was higher than the average December and January's inflation was lower than the average January, it would indeed be accurate to say "inflation is cooling."

[life_force_prana]

With the inflation figures released today, the general tone of headlines is that inflation "cooled" to 2.8% in February from 3.0% in January. These numbers are correct, but misleading.

Here are the Consumer Price Index (CPI-U) figures that illustrate the problem:

January 2024.......308.417
February 2024.....310.326
December 2024...315.605
January 2025.......317.671
February 2025.....319.082

The 2.8% for February 2025 comes from comparing it to February 2024: 319.082/310.326 =1.0282 = 2.82%
The 3.0% for January 2025 comes from comparing it to January 2025: 317.671/308.417 =1.03 = 3.0%

The problem is that in comparing January and February 2025 to different values, it does not reflect the very high inflation in February 2024: 310.326/308.417 x 12 =12.0743. The annualized rate for February 2024 was 7.43%, so if February 2025 simply was less than that very high rate, inflation for this past month appears to be "cooling."

For the first two months of 2025, inflation is 319.082/315.605 x 6 =6.0661, an annualized rate of 6.6%.

I agree with you. This makes sense to me because it is giving me a sense for how prices are changing month over month.

I dont care about all these statistically/technically correct ways to calculate (seasonality, base effects, etc.), what matters is how are prices change as one experiences it in one's daily life.

[Northern Flicker]

With the inflation figures released today, the general tone of headlines is that inflation "cooled" to 2.8% in February from 3.0% in January. These numbers are correct, but misleading.

Here are the Consumer Price Index (CPI-U) figures that illustrate the problem:

January 2024.......308.417
February 2024.....310.326
December 2024...315.605
January 2025.......317.671
February 2025.....319.082

The 2.8% for February 2025 comes from comparing it to February 2024: 319.082/310.326 =1.0282 = 2.82%
The 3.0% for January 2025 comes from comparing it to January 2025: 317.671/308.417 =1.03 = 3.0%

The problem is that in comparing January and February 2025 to different values, it does not reflect the very high inflation in February 2024: 310.326/308.417 x 12 =12.0743. The annualized rate for February 2024 was 7.43%, so if February 2025 simply was less than that very high rate, inflation for this past month appears to be "cooling."

For the first two months of 2025, inflation is 319.082/315.605 x 6 =6.0661, an annualized rate of 6.6%.

The change from 12 months prior is basically a 12-month moving average because it incorporates the full 12 months of changes. The motivation is that there is seasonal and idiosyncratic variance in the monthly inflation rate. Annual inflation over the previous 12 months smooths that out, but at the expense of not always reflecting what was happening most recently.

[bobcat2]

This is simply a problem in how the news media choose to report the inflation data. In addition to what they report they could also report the annualized 3 mo. and 6 mo. inflation rates, but they don't.

Here is the Cleveland Fed CPI inflation NowCast updated today for the first quarter. Keep in mind that over 70% of the NowCast quarterly inflation estimate is now known - all of January and February and some of March. This report gets to the core of the problem the OP is concerned about IMO.

QUARTERLY ANNUALIZED PERCENT CHANGE

Code: Select all

Quarter	     CPI  Core CPI	   Updated
2025:Q1	    3.86    3.74	     03/12

The NowCast estimate for March inflation is lower than the actual Jan-Feb inflation rate and reduces the annual inflation rate for the quarter from the rate in the first two months. This can be seen in the March NowCast month/month inflation rate.

INFLATION, MONTH-OVER-MONTH PERCENT CHANGE

Code: Select all

Month	        CPI	    Core CPI	
March 2025	0.01	      0.26	

Link to Cleveland Fed NowCast
https://www.clevelandfed.org/indicators ... nowcasting

BobK

Edited for double word and awkward phrasing.

[bombcar]

You’re basically right in this - most financial news is worse than worthless, which is why you ignore it.

Comparison of annualized 2024 to annualized 2025 may be useful, but can’t be done until 2025 is over.

[Thesaints]

The simple truth is that the inflation data, while deterministically measured and cslculated, is open to misinterpretation.

The CPI number in particular is a triple average: average cost of average consumption for the average family.
If Macy’s only sold average size men suits, half the population would find them too small and the other half would find them too large. And both halves would write on BH that Macy’s is stocking the wrong suit size !

[prioritarian]

I'm not saying it has changed.
What people take from the data depends on understanding how the data are calculated.

An awful lot of knowledgeable people have raised questions about CPI and as a result several independent price statistics have been developed (e.g. the billion price project and state street "price stat"). These independent measures of overall changes in prices have essentially confirmed that CPI does a decent job of measuring inflation. Ironically, one of the reasons that CPI tends to slightly underestimate inflation is because it was manipulated inorder to lower social security COLA adjustments (by those who tend to question the validity of government statistics and the need for a safety net). For this and other reasons the Federal Reserve uses PCE and core PCE as its preferred measures of inflation. The media's fixation on CPI is akin to their fixation on the DOW (e.g. useful for headlines but not that useful for those interested in better measuring price trends).

And yet, for all of its flaws, CPI does a pretty good job measuring inflation. Just look at the Massachusetts Institute of Technology’s Billion Price Project (BPP), which monitors daily price fluctuations of ~15 million items sold by +1000 online retailers in more than 70 countries. It runs fairly parallel to CPI, as seen in the three charts (above).

https://ritholtz.com/2022/05/flawed-but-useful/

[jeffyscott]

Using the latest figure of 0.2%, seasonally adjusted, tells the opposite story that the OP's misunderstanding led to.

So projecting that 0.2% over a year puts it at around 2.5%. It'd be better to find a figure with an additional significant figure before projecting, though. (Maybe toddthebod's 2.6% figure is based on having a source with more significant figures?). 2.5% (or 2.6%) is just about at the Fed target, since PCE is usually a bit lower than CPI.

Ignoring seasonality will have you fretting over high inflation every late winter into spring followed by worrying about deflation every fall to early winter.

[nisiprius]

Surely month-to-month measures of inflation, just like month-to-month measures of stock market returns, are mostly noise.

[toddthebod]

With the inflation figures released today, the general tone of headlines is that inflation "cooled" to 2.8% in February from 3.0% in January. These numbers are correct, but misleading.

Here are the Consumer Price Index (CPI-U) figures that illustrate the problem:

January 2024.......308.417
February 2024.....310.326
December 2024...315.605
January 2025.......317.671
February 2025.....319.082

The 2.8% for February 2025 comes from comparing it to February 2024: 319.082/310.326 =1.0282 = 2.82%
The 3.0% for January 2025 comes from comparing it to January 2025: 317.671/308.417 =1.03 = 3.0%

The problem is that in comparing January and February 2025 to different values, it does not reflect the very high inflation in February 2024: 310.326/308.417 x 12 =12.0743. The annualized rate for February 2024 was 7.43%, so if February 2025 simply was less than that very high rate, inflation for this past month appears to be "cooling."

For the first two months of 2025, inflation is 319.082/315.605 x 6 =6.0661, an annualized rate of 6.6%.

Are you being intentionally misleading?

Even by your own standards, inflation is cooling. Annualized January inflation (i.e., January to February) was lower than annualized December inflation.

319.082 / 317.671 = 1.0044
317.671 / 315.605 = 1.0065

So whether you look at monthly data OR year-over-year data, it is accurate to say "inflation is cooling."

[toddthebod]

With the inflation figures released today, the general tone of headlines is that inflation "cooled" to 2.8% in February from 3.0% in January. These numbers are correct, but misleading.

Here are the Consumer Price Index (CPI-U) figures that illustrate the problem:

January 2024.......308.417
February 2024.....310.326
December 2024...315.605
January 2025.......317.671
February 2025.....319.082

The 2.8% for February 2025 comes from comparing it to February 2024: 319.082/310.326 =1.0282 = 2.82%
The 3.0% for January 2025 comes from comparing it to January 2025: 317.671/308.417 =1.03 = 3.0%

The problem is that in comparing January and February 2025 to different values, it does not reflect the very high inflation in February 2024: 310.326/308.417 x 12 =12.0743. The annualized rate for February 2024 was 7.43%, so if February 2025 simply was less than that very high rate, inflation for this past month appears to be "cooling."

For the first two months of 2025, inflation is 319.082/315.605 x 6 =6.0661, an annualized rate of 6.6%.

I agree with you. This makes sense to me because it is giving me a sense for how prices are changing month over month.

I dont care about all these statistically/technically correct ways to calculate (seasonality, base effects, etc.), what matters is how are prices change as one experiences it in one's daily life.

When fruit is more expensive in January than it is in July because it is imported from South America, but it is cheaper than last January, and last July was cheaper than the previous July, do you call that inflation?

[bobcat2]

Surely month-to-month measures of inflation, just like month-to-month measures of stock market returns, are mostly noise.

True. But quarterly annualized measures of inflation are meaningful in the same way annualized quarterly GDP data are meaningful. And certainly the quarterly measures of inflation are more meaningful than comparing the latest month of inflation data to the same month a year earlier. Much of that inflation, whether high or low, occurred over six months ago.

Not to mention that measuring the inflation rate over a 12 month interval is highly influenced by whether inflation was hot or cold in a month a year ago.

BobK

[bobcat2]

With the inflation figures released today, the general tone of headlines is that inflation "cooled" to 2.8% in February from 3.0% in January. These numbers are correct, but misleading. ... For the first two months of 2025, inflation is 6.0661, an annualized rate of 6.6%.

Are you being intentionally misleading?
Even by your own standards, inflation is cooling. Annualized January inflation (i.e., January to February) was lower than annualized December inflation.
319.082 / 317.671 = 1.0044
317.671 / 315.605 = 1.0065
So whether you look at monthly data OR year-over-year data, it is accurate to say "inflation is cooling."

Looking at annualized data for the first quarter, it is clear that inflation has accelerated over the rate in the second half of last year.

Here is the Cleveland Fed CPI inflation NowCast updated today for the first quarter. Keep in mind that over 70% of the NowCast quarterly inflation estimate is now known - all of January and February and some of March. This report gets to the core of the problem the OP is concerned about IMO.

QUARTERLY ANNUALIZED PERCENT CHANGE

Code: Select all

Quarter	     CPI  Core CPI	   Updated
2025:Q1	    3.86    3.74	     03/12

The NowCast estimate for March inflation is lower than the actual Jan-Feb inflation rate and reduces the annual inflation rate for the quarter from the rate in the first two months. This can be seen in the March NowCast month/month inflation rate.

INFLATION, MONTH-OVER-MONTH PERCENT CHANGE

Code: Select all

Month	        CPI	    Core CPI	
March 2025	0.01	      0.26	

These inflation data are seasonally adjusted.
Link to Cleveland Fed NowCast
https://www.clevelandfed.org/indicators ... nowcasting

BobK

[toddthebod]

Are you being intentionally misleading?
Even by your own standards, inflation is cooling. Annualized January inflation (i.e., January to February) was lower than annualized December inflation.
319.082 / 317.671 = 1.0044
317.671 / 315.605 = 1.0065
So whether you look at monthly data OR year-over-year data, it is accurate to say "inflation is cooling."

Looking at annualized data for the first quarter, it is clear that inflation has accelerated over the rate in the second half of last year.

Here is the Cleveland Fed CPI inflation NowCast updated today for the first quarter. Keep in mind that over 70% of the NowCast quarterly inflation estimate is now known - all of January and February and some of March. This report gets to the core of the problem the OP is concerned about IMO.

QUARTERLY ANNUALIZED PERCENT CHANGE

Code: Select all

Quarter	     CPI  Core CPI	   Updated
2025:Q1	    3.86    3.74	     03/12

The NowCast estimate for March inflation is lower than the actual Jan-Feb inflation rate and reduces the annual inflation rate for the quarter from the rate in the first two months. This can be seen in the March NowCast month/month inflation rate.

INFLATION, MONTH-OVER-MONTH PERCENT CHANGE

Code: Select all

Month	        CPI	    Core CPI	
March 2025	0.01	      0.26	

Link to Cleveland Fed NowCast
https://www.clevelandfed.org/indicators ... nowcasting

BobK

There is a reasonable discussion to be had about how to accurately capture inflation and the changes in inflation. Unfortunately OP has muddied the waters with misleading figures, either intentionally or not. Your choice of quarterly figures is also potentially misleading. Why is 2 months or 3 months better than 1 month or 6 months or 12 months (all of which reflect lower inflation)?

[nisiprius]

Here are monthly inflation figures, calculated at the percentage increase in CPI, annualized by the formula (1 + monthly)^12 - 1.

I don't think it tells us much of interest.



According to my eyeball, it was level from 2000-2021, elevated from 2021-2023, and back to the previous level since then. And if there's a trend since 2023 I don't see it.

If someone else want to do polynomial regression or curve fits or whatever, I'll leave it up to them. But please use some appropriate methodology for stating significance levels.

I have opinions on where I think things are going to go, but I just don't think it's shown up in the CPI price data yet. Plus I've been wrong about everything lately, so what the heck.

[Tycoon]

The numbers are like a Rorchach test for part time economists.

[toddthebod]

Here are monthly inflation figures, calculated at the percentage increase in CPI, annualized by the formula (1 + monthly)^12 - 1.

I don't think it tells us much of interest.



According to my eyeball, it was level from 2000-2021, elevated from 2021-2023, and back to the previous level since then. And if there's a trend since 2023 I don't see it.

If someone else want to do polynomial regression or curve fits or whatever, I'll leave it up to them. But please use some appropriate methodology for stating significance levels.

I have opinions on where I think things are going to go, but I just don't think it's shown up in the CPI price data yet.

I see one thing of interest! Disproving OP's argument in his first post, the last dot is lower than the second to last dot.

[stlrick]

Looking at annualized data for the first quarter, it is clear that inflation has accelerated over the rate in the second half of last year.

Here is the Cleveland Fed CPI inflation NowCast updated today for the first quarter. Keep in mind that over 70% of the NowCast quarterly inflation estimate is now known - all of January and February and some of March. This report gets to the core of the problem the OP is concerned about IMO.

QUARTERLY ANNUALIZED PERCENT CHANGE

Code: Select all

Quarter	     CPI  Core CPI	   Updated
2025:Q1	    3.86    3.74	     03/12

The NowCast estimate for March inflation is lower than the actual Jan-Feb inflation rate and reduces the annual inflation rate for the quarter from the rate in the first two months. This can be seen in the March NowCast month/month inflation rate.

INFLATION, MONTH-OVER-MONTH PERCENT CHANGE

Code: Select all

Month	        CPI	    Core CPI	
March 2025	0.01	      0.26	

Link to Cleveland Fed NowCast
https://www.clevelandfed.org/indicators ... nowcasting

BobK

There is a reasonable discussion to be had about how to accurately capture inflation and the changes in inflation. Unfortunately OP has muddied the waters with misleading figures, either intentionally or not. Your choice of quarterly figures is also potentially misleading. Why is 2 months or 3 months better than 1 month or 6 months or 12 months (all of which reflect lower inflation)?

Toddthebod: The suggestion that I have muddied the waters with misleading figures, perhaps intentionally - made by you in three posts now - is what is misleading. I report figures (you have used them) and what I calculated from them. Everything was up front for anyone to agree or disagree with their own calculations. [Unnecessary comments removed by admin LadyGeek]

[toddthebod]

There is a reasonable discussion to be had about how to accurately capture inflation and the changes in inflation. Unfortunately OP has muddied the waters with misleading figures, either intentionally or not. Your choice of quarterly figures is also potentially misleading. Why is 2 months or 3 months better than 1 month or 6 months or 12 months (all of which reflect lower inflation)?

Toddthebod: The suggestion that I have muddied the waters with misleading figures, perhaps intentionally - made by you in three posts now - is what is misleading. I report figures (you have used them) and what I calculated from them. Everything was up front for anyone to agree or disagree with their own calculations. [Unnecessary comments removed by admin LadyGeek]

Then why did you choose 2 months of data for your final calculation?

[prioritarian]

Are you being intentionally misleading?
Even by your own standards, inflation is cooling. Annualized January inflation (i.e., January to February) was lower than annualized December inflation.
319.082 / 317.671 = 1.0044
317.671 / 315.605 = 1.0065
So whether you look at monthly data OR year-over-year data, it is accurate to say "inflation is cooling."

Looking at annualized data for the first quarter, it is clear that inflation has accelerated over the rate in the second half of last year.

Here is the Cleveland Fed CPI inflation NowCast updated today for the first quarter. Keep in mind that over 70% of the NowCast quarterly inflation estimate is now known - all of January and February and some of March. This report gets to the core of the problem the OP is concerned about IMO.

QUARTERLY ANNUALIZED PERCENT CHANGE

Code: Select all

Quarter	     CPI  Core CPI	   Updated
2025:Q1	    3.86    3.74	     03/12

The NowCast estimate for March inflation is lower than the actual Jan-Feb inflation rate and reduces the annual inflation rate for the quarter from the rate in the first two months. This can be seen in the March NowCast month/month inflation rate.

INFLATION, MONTH-OVER-MONTH PERCENT CHANGE

Code: Select all

Month	        CPI	    Core CPI	
March 2025	0.01	      0.26	

These inflation data are seasonally adjusted.
Link to Cleveland Fed NowCast
https://www.clevelandfed.org/indicators ... nowcasting

BobK

CPI includes volatile components, such as, food and energy so with only two months of cherry-picked data it's easy to tell whatever story one wants to tell.

[bobcat2]

Why is 2 months or 3 months better than 1 month or 6 months or 12 months (all of which reflect lower inflation)?

Quarterly annualized data are better because they are current while also eliminating much of the noise in monthly data.

Going forward I am much less concerned with what happened with inflation 6-12 months ago than what is happening with inflation now.

We report GDP on a quarterly annualized basis because that tells us what is happening in the economy recently. We could report this quarterly data on an annual basis, but typically that is not the headline, because much of that happened over 6 months ago. It makes more sense to report inflation the same way. Certainly Fed officials in setting monetary policy are looking at the recent inflation data and not inflation that occurred in the spring and summer of 2024.

BobK

[toddthebod]

Why is 2 months or 3 months better than 1 month or 6 months or 12 months (all of which reflect lower inflation)?

Quarterly annualized data are better because they are current while also eliminating much of the noise in monthly data.

Going forward I am much less concerned with what happened with inflation 6-12 months ago than what is happening with inflation now.

We report GDP on a quarterly annualized basis because that tells us what is happening in the economy recently. We could report this quarterly data on an annual basis, but typically that is not the headline, because much of that happened over 6 months ago. It makes more sense to report inflation the same way. Certainly Fed officials in setting monetary policy are looking at the recent inflation data and not inflation that occurred in the spring and summer of 2024.

BobK

That's a fair point. Unfortunately, I don't know enough about the seasonality of inflation to conclusively argue, but I can easily imagine certain products that are more expensive in the winter than they are in the summer, e.g., crops that have to be imported from South America or electricity in places with high solar adoption, so I am not convinced that quarterly is the ideal.

Edit: Can't agree with quarterly. Here are the average annualized monthly changes by month for the last 10 years:

Code: Select all

Month	Average annualized change from prior month
1	6.37%
2	5.73%
3	5.90%
4	4.39%
5	4.92%
6	5.22%
7	1.52%
8	1.60%
9	2.25%
10	2.20%
11	-0.78%
12	-0.87%

[investuntilimrich]

The CPI has changed several times over the years. I tend to apply the 1980 version as it makes it easier to compare apples to apples. If you use that algorithm, inflation is about double the current algorithm's numbers, so you can roughly take the recent numbers and double them to arrive at that number. I have not extensively compared prices to find out, but the 1980 algorithm "feels" roughly inline with my purchases over the past few years. The most important number is your personal rate of inflation based on the goods and services you actually use.

[prioritarian]

Quarterly annualized data are better because they are current while also eliminating much of the noise in monthly data.

CPI has a massive seasonal component due to annual changes in fees/prices/COLAs resetting at the beginning of the year.

[bobcat2]

Cleveland Fed CPI inflation NowCast QUARTERLY ANNUALIZED PERCENT CHANGE

Code: Select all

Quarter	     CPI  Core CPI	   Updated
2025:Q1	    3.86    3.74	     03/12

CPI includes volatile components, such as, food and energy so with only two months of cherry-picked data it's easy to tell whatever story one wants to tell.

I, and the Cleveland Fed, have addressed that concern by including the Core CPI numbers which exclude food and energy.
The Core CPI numbers excluding the volatile food and energy sectors are nearly as high as the CPI.
CPI 3.86%
Core CPi 3.74%

BobK

[jeffyscott]

I agree with you. This makes sense to me because it is giving me a sense for how prices are changing month over month.

I dont care about all these statistically/technically correct ways to calculate (seasonality, base effects, etc.), what matters is how are prices change as one experiences it in one's daily life.

When fruit is more expensive in January than it is in July because it is imported from South America, but it is cheaper than last January, and last July was cheaper than the previous July, do you call that inflation?

To add to that, here's an old post that may be of interest regarding seasonality. There, one thing you will see is that for the period that I looked at, average unadjusted inflation was about 1.94% in the first half of each year and -1.02% in the second half. So something like alternating between 4% inflation and 2% deflation on average during each year.

viewtopic.php?p=6743782#p6743782

There's also much more in that entire discussion. But, WRT this discussion the main point is short term unadjusted inflation figures are of no value for anything (except maybe determining the adjusted principal for TIPS).

[toddthebod]

CPI includes volatile components, such as, food and energy so with only two months of cherry-picked data it's easy to tell whatever story one wants to tell.

I, and the Cleveland Fed, have addressed that concern by including the Core CPI numbers which exclude food and energy.
The Core CPI numbers excluding the volatile food and energy sectors are nearly as high as the CPI.
CPI 3.86%
Core CPi 3.74%

BobK

Doesn't solve the seasonality problem:

Code: Select all

1	5.66%
2	5.96%
3	3.90%
4	3.36%
5	3.03%
6	3.38%
7	1.97%
8	2.83%
9	2.68%
10	3.14%
11	0.85%
12	0.49%

(10 year average of annualized Non-seasonally-adjusted CPI less food and energy by month)

[prioritarian]

The CPI has changed several times over the years. I tend to apply the 1980 version as it makes it easier to compare apples to apples. If you use that algorithm, inflation is about double the current algorithm's numbers, so you can roughly take the recent numbers and double them to arrive at that number. I have not extensively compared prices to find out, but the 1980 algorithm "feels" roughly inline with my purchases over the past few years. The most important number is your personal rate of inflation based on the goods and services you actually use.

Please explain why independent measures of actual consumer prices roughly match CPI.

https://thebillionpricesproject.com/datasets/



https://pubs.aeaweb.org/doi/pdfplus/10. ... p.30.2.151

Note: the billion price project is now the State Street Price Stat (subscription required)

[jeffyscott]

Toddthebod: The suggestion that I have muddied the waters with misleading figures, perhaps intentionally - made by you in three posts now - is what is misleading. I report figures (you have used them) and what I calculated from them. Everything was up front for anyone to agree or disagree with their own calculations. This kind of arrogant, insulting crap is completely unnecessary. Please understand that I won't be bothering to follow this thread any longer, so don't bother replying for my benefit.

Then why did you choose 2 months of data for your final calculation?

I think the OP is probably just unaware of how distorted a picture you can get if you ignore seasonality. They've been given some leads to assist in learning about this...

[FactualFran]

There is a reasonable discussion to be had about how to accurately capture inflation and the changes in inflation. Unfortunately OP has muddied the waters with misleading figures, either intentionally or not. Your choice of quarterly figures is also potentially misleading. Why is 2 months or 3 months better than 1 month or 6 months or 12 months (all of which reflect lower inflation)?

Averages over 3 months are good enough for the Cost of Living Adjustment for Social Security. On the other hand, other automatic inflation adjustments use averages over 12 months. Then there is the inflation adjustment to Series I Savings Bonds that use values for two months that are six months apart.

[#Cruncher]

... there is seasonal and idiosyncratic variance in the monthly inflation rate. Annual inflation over the previous 12 months smooths that out, but at the expense of not always reflecting what was happening most recently.

Correct. There is a trade-off in choosing the time period over which to view changes in the CPI index. Month-to-month can include "noise" as nisiprius says (here). Year-over-year smooths out the noise but at the expense of recency. Maybe quarterly change is a good compromise as bobcat2 suggests (here).

But for any time period chosen other than annual, I submit that the seasonally adjusted index is a better choice for detecting trends than the non-seasonally adjusted one. The FRED graph below shows that seasonally adjusted (the solid blue line) is smoother than the non-seasonally adjusted (dotted green line). Note how the two lines repeat their relationship over each of the two years of the graph.



Click the graph to open an interactive version.

[LadyGeek]

This thread has run its course and is locked (general economic discussion). As noted in the site owner's post in the locked thread Re: U.S. stocks in free fall:

A reminder from the forum policies:

Investing - Theory, News & General

If it's investment related and it doesn't fall into the above category it goes here.

General economic discussions do not fall under the category of "investment related". Nor do they fall under the allowed topics in any of our other forums. You will have to find another site or outlet if you wish to take part in such discussions.

Just in case the previous was unclear, we specifically addressed this issue in the following policy:

Non-actionable (Trolling) Topics

If readers can't do anything with the content of a topic other than argue about it, it does not belong here. Examples include:

• US or world economic, political, tax, health care and climate policies
• conspiracy theories of any type
• discussions of the crimes, shortcomings or stupidity of other people, whether they be political figures, celebrities, CEOs, Fed chairmen, subprime mortgage borrowers, lottery winners, federal "bailout" recipients, poor people, rich people, etc. Of course, you are welcome to talk about the stupid financial things you have done.