Prices are going up, but by how much? Is the government squashing the truth? We unpack the methodology behind the government reported Consumer Price Index (CPI) and explore sources of error including incentives for squashing the reported numbers. We explore why there are surprisingly few alternative inflation measures and analyze why some of the more popular ones have their own challenges.
.png)
Hey friends -
I previously wrote about inflation in June when chatter was just starting to pick up that inflation may exceed expectations. With inflation now at an annualized 6.8%, a rate not seen since 1982, it seems an appropriate time to revisit the topic.
In this week's letter:
Total read time: 10 minutes, 53 seconds.
Keep in mind that there are three main inflation measures. Whereas consumer price inflation concerns itself with goods and services we buy to consume, asset price inflation concerns the prices of investable assets and monetary inflation concerns the amount of money in the system.
Different measures of inflation will tell different stories. As I concluded in June:
Where does all of this leave us? We have the Consumer Price Index indicating ~2.5% inflation since 1990, a Big Mac inflation indicating 3.2%, and financial asset prices that have been increasing even faster in absolute terms, but not always in relative terms depending on your income. We also have monetary inflation that is increasing faster than the Consumer Price Index, and one would think all of that additional money in the system has to go somewhere.
The answer is that the "real" inflation rate is dependent on your situation.
I encourage you to read the previous letter for the full picture. This week's letter is focused on consumer price inflation.
Consumer price inflation is usually measured as the change in the price of the Consumer Price Index, abbreviated CPI. Formally, the CPI is:
a measure of the average change over time in the prices paid by urban consumers for a market basket of consumer goods and services
Informally, you can think of it as the changing price of the "stuff" we consume in our day-to-day lives. As that "stuff" changes, like the shifting average diet of the American family, the CPI is updated accordingly. It's the inflation number you see in headlines.
The US Bureau of Labor Statistics (BLS) is responsible for both the methodology behind the CPI - what goes in the basket, how prices as measured, and how the CPI is updated - as well as the actual CPI measurement. BLS is a government organization whose commissioner is appointed to a four-year term by the President, which naturally gives rise to endless conspiracy theories on how the CPI is manipulated.
More on conspiracy theories later. First, we need to look at how the CPI is calculated.
The CPI effectively attempts to construct the average American - what they buy and how much they buy it for. While we say "average" American, what we actually mean is the average urban American. Urban includes ~93% of the population.
Step one is to determine the basket of goods and services, which the BLS determines surveys and diaries. 24,000 consumers are interviewed four times a year on their spending habits as part of the Consumer Expenditure Survey. An additional 12,000 consumers are asked to maintain a two-week diary on everything they purchase.
By combining the two data sources, the BLS determines both the basket and the weights in the baskets. It's held steady for two years. If Americans are drinking more beer than wine, beer will get a higher weighting. The basket includes just about everything, from food to gasoline to internet bills and haircuts. The end result is enormous - around 80,000 items!
Step two is to determine the average price. Here we have to separate shelter and goods & services. Shelter is the actual service provided by a home rather than the value of the home itself and the relevant cost for CPI. Prices are gathered again by survey, in this case from 8,000 consumers who are surveyed twice a year. For homes that are owned rather than rented, the BLS calculates the "owner's equivalent rent" - the rent that the owner would have to pay if they were renting their homes.
Calculations for goods and services are done differently - with inputs directly from stores. The process is more complex than it may appear on the service. If the good for consideration is whole milk, there may be multiple brands and multiple sizes. From the survey data, the BLS calculates the probability of any given brand/size of whole milk being purchased. From yet another annual survey of 14,500 consumers, BLS identifies outlets where goods and services are purchased. BLS then selects a single milk carton from a single outlet through random sampling. That specific whole milk carton from that specific outlet is then repriced monthly or bimonthly for four years at which point a new whole milk sample from a new outlet is chosen.
You can imagine that such a process, executed across all varieties of milk yields a wide sample of milks. This is repeated across TVs, cars, and everything else in the basket to create a truly astonishing cross-section of what the average American purchases.
All of these calculations roll back up to form the CPI. It's a rollup of interrelated samples - the sample products, chosen and weighted by survey data, priced from outlets chosen by yet another survey.
Changes in the CPI should match changes in consumer spending, but there are a couple of potential error sources. One potential significant source of error is consumer substitutions. Dungeness lump crab meat tripled in price to $60 per pound this winter. Consumers, unless they were particularly insistent like my brother on his birthday, likely consumed less crab and likely increased their consumption of a cheaper substitute like shrimp. That won't get reflected in the basket that only changes every two years.
The Chained CPI updates the basket on a rolling basis. It compares what consumers purchased in the prior period to what they actually purchased this period, rather than assuming the basket is fixed for two years. The difference reduces the published inflation rate as consumers replace high-cost items with cheaper ones.
A second potential source of error is the eligible basket of goods and services. The CPI only measures those items that we consumers purchase directly, not those that are purchased indirectly on our behalf through services like health insurance. The Personal Consumption Expenditures (PCE) is an alternative measurement assembled by the Bureau of Economic Analysis, part of the Department of Commerce, that considers both types of expenditures. As may be expected, healthcare plays a much larger role in the PCE than it does in the CPI.
Interestingly, the PCE measures inflation at a lower rate than CPI. That's likely because the PCE is also chained, meaning it takes into account consumers changing preferences over time. Usually, that means substituting expensive items for cheaper ones.
A third potential source of error is the volatile prices of food and energy. Core Inflation is the CPI excluding food and energy prices. While preferred by economists for giving a better indication of "true price changes," it excludes over 20% of the average American's expenditures in any period.
Even with all of these considerations, some remain convinced that the entire thing is government manipulated or that there are more effective ways of measuring inflation.
To the conspiracy theorists - I sympathize. Attempts to make the recent inflation spike seem not as real as it is have been laughable at best. From a White House briefing in September the National Economic Council Director:
About half of the overall increase in grocery prices can be attributed to a significant increase in prices in three products: in beef, in pork, and in poultry. And in beef and in pork, we’ve seen double-digit increases in prices over the last couple of months. If you take out those three categories, we’ve actually seen price increases that are more in line with historical norms.
Right. If you exclude all of the items where prices are going up, the price increases are lower. Similar to how WeWork was profitable, but only if you measured it based on their community-adjusted EBITDA. Such efforts don't instill confidence.
John Williams of ShadowStats has taken his objections much further. He went and assembled an entirely new inflation index.
ShadowStats purports that inflation has been higher than CPI. Much higher.
This chart has been making the rounds with increasing frequency as CPI numbers have increased. It's a popular fallback for gold and bitcoin maximalists. Like most good conspiracies, there's likely an element of truth buried in the data but it's almost impossible to know because ShadowStats doesn't formally publish its methodology.
Carl Sagan coined the aphorism that extraordinary claims require extraordinary evidence. It applies here. Persistent inflation way in excess of CPI is an extraordinary claim. Were ShadowStats correct, it would mean that we’ve been in a recession for the past 20 years (inflation much greater than nominal GDP).
Despite the odds presented by the lack of a published methodology, Ed Dolan dug into the ShadowStats calculations. His 2015 analysis hasn't been getting as much airtime, probably because it convincingly dismantles ShadowStats. In short - the extraordinary evidence isn’t there.
Dolan's work is first-rate and I highly recommend reading it. I'll highlight two key elements here.
Dolan got ahold of a 1982 grocery advertisement and compared the expected 2015 prices of those groceries using ShadowStats and CPI. The results weren't pretty.
The far right two columns indicate the error between the actual price of the good and the predicted price using the inflation index. A “1.00” score would be a perfect predictor. A “0.50” would mean the index predicted a 50% lower price than what the good actually cost and a “1.50” would mean a 50% overstatement.
The CPI understates inflation by 9% over the 23 years as measured by the items sampled by Dolan. ShadowStats overstates it by 292%. Even for the item that ShadowStats predicts the best - chuck roast - inflation is overstated by 174%. That's almost 5 times the largest CPI error.
Despite the lack of a published methodology, Dolan also dug into how the index was calculated using past commentary from ShadowStats. BLS, from time to time, updates the methodology they use to calculate CPI. They publish an alternative CPI measure that recalculates CPI as if those changes had been incorporated since 1980. Most methodological changes do tend to reduce the reported inflation, albeit by small amounts.
ShadowStats appears to take issue with those recalculations and reverses them when calculating its own inflation index. In so doing, it double counts the revisions thereby increasing the discrepancy between CPI and ShadowStats. Dolan estimates this accounts for a full 5.1% of the total 7% reporting gap between the two indices.
Either piece of the analysis should raise eyebrows. Together, they're damning.
Despite the problems with ShadowStats, it is worth considering the motivations behind artificially depressing CPI. Many large government transfers (read: payments) including social security and treasury inflation-protected securities (TIPS) are CPI inflation-indexed. Depressing CPI would materially reduce payments.
That's a big incentive. As Charlie Munger said, “show me the incentives and I’ll show you the outcome.” Methodological changes over time have almost universally depressed reported CPI numbers. PCE, the other government-published consumer inflation index, is subject to continual revision for reasons such as changing the estimates of consumer spending patterns.
Both inflation measures involve wildly complex methodologies. While necessary, such complexity provides ample opportunity for shenanigans that would be difficult for the layperson to detect.
The data collection necessary calculate inflation is also an expensive endeavor. That is a real and meaningful barrier to entry for competing inflation indices that aren’t funded by taxpayer dollars.
It's worth considering other measures.
My personal favorite - but by no means the most useful - is the Big Mac index, which I highlighted in a previous letter. Constructed by the Economist, it tracks the price of a Big Mac in countries around the world. As of the end of 2020, the Big Mac index indicated prices had increased 2.5 times since 1990, 25% more than as measured by CPI.
The Billion Prices Project gathered daily pricing data from online retailers to concoct an alternative inflation index. According to JP Koning, the data more-or-less reinforced CPI published inflation rates when it was publicly available. That's all but impossible to determine now. Billion Prices Project sold out to State Street and no longer makes data publicly available.
The American Institute for Economic Research publishes yet another index, the Everyday Price Index (EPI). Whereas it uses the same data and a similar methodology to the CPI, it limits the basket of goods and services to only those that meet two criteria:
The result is that the EPI only tracks ~37% of total household expenditures. Major differences between it and the CPI include shelter costs and big-ticket items like cars.
For this past November, official CPI inflation was 6.8%. The EPI estimated it at 9.3%. That's a major difference from a reputable source with a well-published methodology.
I leave finding the true inflation rate to you.
Next week we'll turn our attention from inflation calculations to analyzing current inflation from a historical perspective. Hopefully, it'll help us avoid takes like this:
jack⚡️ @jack
Hyperinflation is going to change everything. It’s happening.
October 22nd 2021
18,623 Retweets77,026 Likes
Yes, we have inflation. But not hyperinflation. Hyperinflation is defined as 40% price increases per month. We're not even in the ballpark of 40% per year.
What we have is hyped inflation.
A holiday classic that's become one of my favorites, courtesy of Alton Brown.
12 Large Eggs
1 pound Sugar
1 teaspoon Freshly Grated Nutmeg
1 pint Half-and-Half
1 pint Whole Milk
1 pint Heavy Cream
1 cup Jamaican Rum
1 cup Cognac
1 cup Bourbon
1/4 teaspoon Kosher Salt
Separate the yolks. Beat the yolks with the sugar and nutmeg in a large mixing bowl until the mixture lightens in color and falls off the whisk in a solid ribbon. Combine the dairy, booze, and salt in a second bowl and slowly beat into the egg mixture. Consume immediately or refrigerate in a clean, air-tight glass jar for as long as you'd like before consuming. For added fun - just before serving, whip the egg whites until they hold stiff peaks and dollop onto the eggnog.
Don't knock it until you try it. This is an eggnog that will convert the eggnog nonbelievers and bring joy to the enthusiasts. There's no need to age it the first time, although as someone who stocks a batch with an age that can be measured in units greater than months, I can assure you that aging it can be done safely. While the idea of whole raw egg or egg yolk in a drink may seem foreign today, there is an entire family of drinks popular in colonial America - Flips - for which it is the hallmark. Eggnog is a close sister of the flip and builds on the textural and caloric egg base with milk or cream. If the recipe is too much, scale it back while keeping the ratios the same. A fair warning - if you're serving it to a group, what you make will disappear faster than you expect.
Cheers,
Jared
