“Of all the economic statistics produced by the U.S. federal government, none has a direct impact on the lives of everyday Americans quite like the Consumer Price Index (CPI)." - BLS
The CPI directly impacts many government programs, such as Social Security benefits, and millions of contracts such as business agreements, government obligations, leases, court orders... all adjusting settlements to the purchasing power of the dollar. And yet ... the CPI remains one of the most complex and elusive of economic measures.
In this article we discuss the importance of understanding inflation, provide a historical perspective, and then lay out our view on where inflation may go in the future. We do this in four sections:
Introduction - An overview of what inflation is and its measures.
Past - The history of inflation measures going back to before 1900.
Present - A look at recent trends in inflation.
Future - Where inflation may go in the future.
Few questions in the world of finance are more important today than the nature and causes of inflation. Low inflation means that fiat currencies are retaining their purchasing power. That is a good thing, but a belief that inflation will continue to stay lower for longer has contributed to lower interest rates than at any time in recorded history. Low interest rates are justified if money received in the future is nearly as valuable as money today. This is the current market expectation, a view increasingly shared by central banks.
Government debt levels across the industrialized world have never been so high, a fact made possible by negative or near zero interest rates. Central banks can effectively control interest rates on government debt. For example, during WWII the Federal Reserve helped the US government pay for the war effort by buying Treasury bills at 0.375%, effectively putting a ceiling on the rate of interest. Inflation surged to an annual rate of 8.7% as government spending on good and services expanded relative to economic output. The inflation shock was expected. The classic recipe for inflation is too much money chasing too few goods and services.
COVID19 has created war like deficit spending, leading investors like Jeffrey Gundlach seeing risks of inflation right around the corner. However, other notable investors like Stanley Druckenmiller believe we have "cracked the credit bubble" and sees risks of deflation. Time will tell but understanding what inflation is and its history can help provide some insight into the risks and potential future scenarios.
1.1. What is inflation?
Inflation is defined as a general increase in prices and fall in the purchasing value of money. The concept is simple, but the practice of quantifying inflation is complex and leaves plenty of room for interpretation. Which prices are used? How are those prices weighted? What about substitution effects between similar goods and services? These questions and more have made estimating and comparing inflation over time challenging.
The oldest measure of inflation in existence is the Consumer Price Index (CPI) which was first constructed in the early 19th century. How CPI was measured changed over the years with improvements in data and methods. Many measures of inflation exist today, each with their own strengths and weaknesses. A review of these measures can help to illustrate the tradeoffs and complexity required when measuring price movements.
1.2. CPI vs PCE
The USA primarily uses two primary measures of inflation: CPI and the Personal Consumption Expenditures price index (PCE). CPI is produced by the Bureau of Labor Statistics whereas PCE is produced by the BEA. Four key factors differentiate CPI and PCE:
Formula Effect: PCE allows for substitution effects. For example, if the price of chicken goes up relative to beef then PCE estimated how much consumers shifted spending toward beef. As a result, CPI is typically higher than PCE as a result of ignoring consumer’s ability to change short term spending patterns in responses to changing prices.
Weight Effect: CPI and PCE both attempt to measure price movements in a basket of goods and services, but the weights on these baskets ae determined by different data sources. CPI is based on consumer expenditure surveys whereas PCE is derived from business surveys such as the Conesus Bureau’s retail trade survey.
Scope Effect: CPI is only designed to measure price movements for on out-of-pocket expenditures for urban households. In contrast, PCE is designed to measure price movements for all households as well as nonprofit institutions serving households. The broader scope of PCE includes some key expenses provided for by insurance companies and governments such as employer provided health insurance and medical care paid for by government programs like Medicare and Medicaid.
Other effects: There are numerous other detailed differences in how CPI and PCE are calculated such as how to capture seasonality and measure prices.
Economists also distinguish between Core and Headline inflation. Core inflation ignores the price of energy and food whereas headline inflation makes no exemptions. Which measure is better depends on what question we are trying to answer. If we care about the past then we should focus on headline inflation. But if we want to forecast the future than recent core inflation is going to be more predictive.
PCE was created to overcome shortcomings of the CPI, so it is not surprising that since the year 2000 the FOMC began using the PCE to frame its inflation forecasts. An Economic Letter from the Federal Reserve Bank of San Francisco describes the differences between the CPI and the PCE (which is denoted as PCEPI), and why the Fed prefers the PCE:
“The PCEPI and CPI share many of the same features. For example, the PCEPI, like the CPI, is designed to track the prices of goods and services consumed by households, and it includes much of the same data. However, the PCEPI differs from the CPI on many dimensions. The FOMC cited three of these as reasons for switching its focus from the CPI to the PCEPI (Board of Governors 2000). First, the PCEPI’s formula adjusts to changing consumption patterns, while the CPI is based on a basket of goods and services that is largely fixed. Second, the PCEPI is revised over time, allowing for inflation to be tracked as a more consistent series. Third, the PCEPI's larger scope of goods and services provides a more comprehensive picture of the nation's consumer spending than the CPI.”
For this reason, PCE should arguably receive more attention than CPI. This is especially true today because of how entitlements are treated. Healthcare is the largest component of PCE. In contrast, housing is the largest component for CPI. The reason is that most health care related expenses are not out-of-pocket and thus largely ignored by CPI. As the population ages this key difference will become more material.
2.1. Pre 1900
The first attempt to index retail prices occurred same year as the Bureau of Labor Statistics (BLS), in 1884, in an effort to, “collect information upon the subject of labor, its relation to capital, the hours of labor, and the earnings of laboring men and women, and the means of promoting their material, social, intellectual, and moral prosperity.” In their first attempt, the BLS gathered data on spending habits from 8,544 families. Retail prices were collected on “215 commodities, including 67 food items, in 70 localities.”
Much of the economy around 1900 was built around agriculture and so to was the first “price index”. But that didn’t last long. The industrial revolution, “an increase of 13 categories, including … eight new cloth and clothing items; and the creation of the new category of anthracite and bituminous coal and gas for domestic use.” Even with these new categories there was concern that even the more expanded retail price index could not accurately represent the ever-increasing costs of goods and services in industrial centers throughout the United States.
It wasn’t until 1901 that the BLS attempted a measure of “cost-of-living”. A big reason for the interest was debate on how the McKinley Tariff would impact the wages, prices, and hours of work in the iron and steel, coal, textile, and glass industries. The next cost of living survey occurred between 1917-1919 but was focused only on households with “moderate income” and excludes non-white non-english speaking households. The scope of family expenditures included, “food, clothing, rent, fuel and light, furniture, and “miscellaneous goods”. Upon examination, congress concluded this list to be insufficient because, “the budget did not include savings and insurance or recreational or educational items that ought to be included in a “proper American standard of living.” At the time, vacations and college education were still pretty rare, but that was changing.
In 1929 the BLS conducted a study that found their 1919 basket of goods to be outdated. Technological innovations such as mass production were rapidly changing the purchasing patterns of urbanizing Americans. Under pressure to justify Federal worker wages decreases during the Great Depression, FDR refocused efforts to improve the BLS measures for cost-of-living. Numerous problems with measure were found such as overstating of the importance of food in the index due to its weight in the household basket being calculated back in 1913.
Major changes to the measure were implemented in 1935 such as increasing the survey’s scope of households and better integration of price data across the major categories which remained unchanged at six including food, clothing, rent, fuel and light, house furnishing goods, and miscellaneous goods. A high degree of judgment was still imposed on what constituted a “working-class family” such as making between $500-$2000 a year. But at least the survey included some non-white families and allowed for women to be the primary breadwinner.
The basket of household goods wasn’t revised from the 1919 until 1940 when the BLS introduced the Cost of Living of Wage Earners. This new index was based on a 1934–1936 expenditure study which found that the older item structure failed to reflect new spending habits driven by technological progress which made many upper-income goods more affordable. Weights for the food and clothing decreased while home furnishings and “miscellaneous items” increased. The “fuel and light” index was revised to include “fuel, electricity, and ice” due to the displacement of coal by electricity. A transportation sub-component, which had previously consisted only of streetcar fare, now included several means of transportation, with the majority weight on automobiles and automobile-related goods and services (fuel, taxes, insurance, etc.). This is rather staggering given that car ownership in the United States was already nearly 40% by 1920.
WWII highlighted several challenges with calculating cost-of-living such as the disappearance of goods, dramatic changes in quality, and other factors that caused the “plane of living” to fluctuate. Shortages and disappearing goods occurred throughout the war. BLS personal started attempting to take in to account change in the quality of products. This introduced another area of subjectivity. For example, if a cheaper version of a product didn’t exist anymore because of rationing, but a more expensive version was available, BLS personal were asked to immediately mark up the price for the rationed good to the price of its more expensive substitute. But no adjustment was made for the higher quality.
Inflation surged during WWII as government deficit spending accelerated beyond the countries capacity to produce the goods needed to fight the war. During this time the Federal Reserve instituted yield curve targeting. Bounds were explicitly set along the yield curve to ensure a rising yield curve and low borrowing costs for the Treasury. This allowed insurance companies and other investors to earn some yield (around 2% nominal) on Treasury Bonds while the Federal Reserve kept bond buying mostly to the short end of the curve. Targeting near zero short term rates made buying further out on the curve unnecessary because markets perceived the low rate environment to be persistent. But the real value of bonds was eroded by inflation during the war.
The War had big impacts on spending habits that were expensive to measure at a time of budget cuts so the BLS resorted to guess work. Table 1 below shows a list of actions taken by the BLS in 1942 to re-weight the index such as completely removing new car sales. Difficulties measuring cost-of-living were quantified by a labor union study which estimated true cost increases from the war to be 43.5% from January 1941 to December 1943 instead of the BLS estimate of 23.5%.
Criticisms of the BLS cost-of-living measure during WWII included a lack of scope an inability to capture total “living costs” affecting individual’s well-being such as product quality, ease of acquiring products, and uncertainty about what products would be available. The index wasn’t accounting for the many more families eating away from home and completely ignored income taxes and bond subscriptions. Only 70% of expenditures for captured for the rather narrow scope of households making between $1,250 - $2,000 in (1935 survey).
US dollars experienced a run up in inflation during WWII of around 10-15%. This was in large part because of large deficit spending. Deficits from WWII were only partially paid for with printed money. For example, only 22% of WWII funding was financed through money printing. Twice this amount was financed through higher taxes with the remaining financed through private borrowing like war bonds.
In contrast the German Reichsbank relied much more on printing money. Germany did issue a long-term war loans through the war to the general public to pay off its short-term debt. But the Reichsbank ultimately ended up printing money to buy much of the debt. As a result...German currency in circulation rose 599% during the war. In contrast, Britain saw an increase of just 91%.
2.3. Post WWII
After WWII, the BLS gradually received more funding and attempted to rebrand their cost-of-living measure under the new more recognizable name, “Consumer’s Price Index for Moderate Income Families in Large Cities.” The scope of families was expanded include those making less than $10,000 a year but continued to exclude domestic workers and single individuals living alone.
In 1953 the CPI was expanded to include 9 major categories compared to six in 1935.
“Personal Care” remained unchanged.
“Food” retained its name, but increasingly included options for eating out.
“Clothing” became “Apparel” to capture a wider array of less than utilitarian fashion trends took off throughout the middle class.
“Transportation” was added in 1940…mostly because of automobiles.
“Medical Care” was a new addition in 1953. A growing portion of household medical expenses had started to move into the formal economy as a result of better professional medical training and services.
“Reading” was another new addition, coinciding with a rise in general education as more families moved out of the agricultural industry and into areas requiring more training.
“Recreation” was yet another new category. The five day work week had already been accepted by many businesses like Ford in 1926 but it wasn’t until after WWII that middle income households started to spending meaningful amounts of money purely for fun.
“Housing” was a new category in 1953 that enveloped the 1935 categories of “house furnishing goods” and “rent”, but was expanded to include costs of homeownership. This should be surprising as homeownership had already surpassed 60% by 1950.
“Other goods and services” replaced “miscellaneous goods”. This catch-all category increasingly seems to account for new expenses that had previously existed. In this case it was adjusted to account for the increasing share of services…a trend that continues to this day.
By the late 1950s it was clear that the CPI required a fundamental change in methodology. Until then the BLS adhered to a “constant-goods” framework meaning that the basket of goods was expected to remain constant or change very slowly. However, it was becoming clear that this was no longer the case. In response to the Stigler Committee review the CPI program, a “constant-utility” approach was adopted which attempted to more frequently revise expenditure weights and quickly account for new products and services being purchased.
In 1964 the CPI expanded to include single workers living alone and all households regardless of income so long as at least 50% of the income came from wages. The geographic sample was also expanded to include smaller urban population centers. Still, it wasn’t until April 1977, that a new broader index capturing “all urban households” would increase population coverage from under 45 percent to approximately 80 percent of the total non-institutional U.S. population.
By the 1980s a significant amount of the government’s budget was tied to CPI. Escalating inflation was putting pressure on the BLS to deal with known limitations such as a lack of quality adjustments and “Housing” measure that was attributing rises in home prices as a rising cost instead of an increase in wealth. A big change occurred in 1981 when a “rental equivalence” approach was used for homeowners which approximated cost of owning a home with foregone rental income.
A long list of biases in CPI were identified during the 1980s and 1990s such as substitution bias, outlet substitution bias, quality bias, new-goods bias, bias in the treatment of durable goods…etc, with the net effect being an overestimation of inflation. For example, “substitution bias” means that households would respond to rising prices in a particular product like red delicious apples by buying other kind of apples. Personal computers were also getting more powerful at a rapid rate. Televisions been improving as well, but yearly improvements in the utility of the computer were astonishing compared to TVs.
By 2002 a wide range of CPI measures were available (see Table 2), many aided by the ability of computers to process much larger quantities of data such as the Chained Consumer Price Index which aggregates 8,018 elementary indexes (211 elementary item categories × 38 areas).
2.4. Modern CPI
“Food and Beverages” (unchanged)
“Housing” uses owners' equivalent rent methodology but is otherwise largely unchanged.
“Transportation” now includes motor vehicle insurance but is otherwise largely unchanged.
“Recreation” is largely unchanged but ignores all the new free forms of entertainment like Facebook.
“Medical Care” has grown drastically as a percent of total household expenditures due to a dizzying array of new drugs and services as well as growing health related problems like heart disease, cancer, diabetes, mental illness.
“Education and Communication” has expanded from the 1953 category “Reading” to include college tuition, telecommunication services, smartphones, and computer software like Facebook’s messenger.
“Other Goods and Services” swallowed the 1953 category “Personal Care” (dropping the number of categories from 9 to 8) along with products like tobacco and funeral expenses.
Inflation, defined as “a general increase in prices and fall in the purchasing value of money”. But this general rise is often quite varied across components. Households buy many types of goods and services, and the weights and prices of these have changed considerably over time. Since 1950 we can see several trends.
Without services inflation, PCE would be falling. Over half of the current PCE basket falls under the header of “Services” which includes “Housing” and “Medical Care”. These drive much of the overall increase in price level.
Housing component of CPI has continued to rise in part because of top income earners biding up the price of higher end homes.
Households are spending more on Healthcare, but there are two big reasons why First, people are living longer and more people are getting preventable diseases driven (usually) by poor diet like diabetes and heart disease. While the amount of money spent on healthcare for the average household has been rising the depth of services included is also rising. For example, braces were once considered a luxury, but are now increasingly covered by health insurance. Free services like WebMD are also capturing market share from traditional service providers.
Goods are getting cheaper. This includes nondurables after removing Food including cosmetics, cleaning products, office supplies, packaging and containers, paper and paper products, personal products, rubber and plastics and durable goods like mobile phones (after accounting for quality improvements), computer equipment, industrial machinery, planes, trains, and automobiles. Inflation in goods is easier to measure because the nature of most of these goods hasn’t changed much over time.
Households are spending less on Apparel and Transportation. The cost of clothes are falling. Prices for cars have been rising, but fewer people need cars today give the rise in urbanization, mass transportation, and the ability to work from home.
Households are spending more on “Communication” (i.e. wireless and landline phone services) because of the internet, smartphones, and related free services. More and more of our lives are being lived digitally. Many services that we receive for free can be upgraded for a fee. WIX, Linkedin, YouTube, and hundreds more ... entice more spending in the area of "communication" which increasingly seems worth it because attention is such a scarce commodity.
Households are spending more on “Recreation” and “Education”. A big part of this has been the rise in the cost of education, specifically post high school. College debt levels have doubled since the Financial Crisis. Many including Peter Theil believe that the education system as a whole is in the middle of a giant bubble. Technological disruptions that are poised to pop this bubble are growing rapidly including companies like 2U, Coursera, Khan Academy and YouTube. Recreation spending is driven primarily by upper income households.
What does all this mean for the future of inflation?
Inflation, defined as “a general increase in prices and fall in the purchasing value of money”, requires calculating changes in prices over time for an “average household” basket of goods and services. This calculation may no longer be practical for two reasons.
The “average household” concept is less representative in today’s industrialized world which is characterized by increasing wealth inequality.
Technology is producing valuable services that cost almost nothing to produce like search engines, free online learning tools, and international communication that are free.
Tyler Cowen wrote a book titled, “Average is Over” which articulated how technology is the driving force behind income inequality. Software and instantaneous communication can augment the value of a single person. The classic example is Mark Zuckerberg who built the foundation for one of the world’s largest companies by himself. It was impossible for a single individual to create so much value back in 1950. This is why, Tyler argues, that average is over…meaning that the middle class will continue to shrink and be replaced by two kinds of people: those that are compliments to computers, and those that are substitutes.
Ray Dalio of Bridgewater agrees stating that, “To understand what’s going on in ‘the economy,’ it is a serious mistake to look at average statistics”. This is the first line in his article titled, “The Two Economies: The Top 40% and the Bottom 60%”. Ray shares a variety of statistics in this article that illustrate what we already know, that the picture of households on top is very different from those at the bottom.
Economists should recognize that that there is no longer a "Average Household" …there are those that are benefiting from automation, robots, instant communication, AI, and the broader software revolution…and those that are moving into the gig economy characterized by service industries like Uber drivers and a growing percent of contract and part-time workers.
Given these facts it’s difficult to justify an “Average Household” basket of goods and services. The BEA calculates these averages, but the average is somewhat meaningless.
The biggest take away from recent trends in inflation suggest that inflation is driven to a large degree by technology and lifestyle changes. Historical patterns in inflation do not need to repeat in a future that is so much more heavily influenced by factors like zero cost to scale digital technologies.
E-Commerce is drastically reducing costs in ways not captured by CPI. Research is so easy now that we have access to ratings and comments on many products, restaurants, and even service providers like drivers and doctors. This is improving competition and leading to better quality products…a factor that is difficult to capture in CPI. We no longer need to physically drive to brick & mortar locations which reduces transportation costs and saves time.
Top-notch schools like MIT, Harvard, and Yale offer an abundance of free course materials. Dozens of websites provide competitive educational services such as Coursera, Academic Earth, Khan Academy Duolingo. If This Then That (IFTTT) is a free timesaving tool that can be incorporated in many ways—the limit is up to you.
Much of what we value today is free...and its very hard to fully appreciate this fact. Music is essentially free. Classic movies are available on YouTube as well. If you’re more into documentaries, you can find many for free on sites like Documentary Heaven. Don’t want to buy Microsoft Office? You can use Google Docs. The IRS has a handy site that has a bunch of different free tax software options to choose from. Many of them free. Dropbox, the classic example, offers 2 GB for free, Microsoft’s OneDrive offers 5 GB, Google Drive offers 15 GB, Box offers 10 GB, and Copy offers 15 GB. CNet has a great comparison grid that offers useful comparisons of various cloud storage services. The list goes on and on...
In short, our conclusion on the future of inflation is that it is not as important today as it once was. So much of our "basket" is a function of individual choices today.
When the CPI was first conceived about half the working age population farmed crops. The "prices" that mattered most were, as a result, crop prices. These prices mattered a lot to everyone because even if you were not a farmer then you were closely connected to them through supporting industries and the broader economy.
Today...the sensitivity we have to prices very much depends on were you live, how much money you have, how old you are, and what you like to do. Perhaps you want to live in the city and spend your extra money traveling to Europe. Perhaps you are older and living on a fixed income and enjoy watching Netflix with your grandkids. The amount of flexibility in how we live and spend money is enormous compared to the past and that makes changes in prices less about what's happening to the value of money and more about the expression of choice and the creation of new technologies.