Can We Trust CPI Inflation Data?

By in
Can We Trust CPI Inflation Data?

How the US government measures inflation is a perennially controversial topic. Typically, the debate centers on claims that the reported numbers are artificially low. Many people simply don’t believe inflation is just 1.9%, the current Consumer Price Index headline reading. The New York Fed’s regular consumer surveys show this well: as of March, respondents expect inflation to be 2.8% over the next year, almost 50% higher than current levels.

The simplest explanation for this disconnect is rooted in behavioral economics. We place more value on the cash we have in hand than an item or service we need/want to buy. That makes everything seem more expensive than it should be.

But there is also the matter of how the Bureau of Labor Statistics actually calculates the CPI: it is by design a clunky, one-size-fits-all measurement. The CPI is a basket of what a typical household purchases, but of course everyone is different. Here are the current major weights for reference:

  • Shelter (owned or rented): 33.3%
  • Food and Energy (mostly gasoline and electricity): 20.6%
  • Purchased commodities (mostly clothes, furniture, drugs, cars): 19.6%
  • Medical Services: 7.0%
  • Transportation services (car repairs/insurance/airfares): 6.0%
  • Education/Communication (including Internet access): 6.0%
  • Recreation (including cable/satellite TV): 3.9%
  • Other services: 3.6%

You can see the problem here: if you live in a large city you might spend 50% of your household budget on shelter, but none on gasoline. And if you have children, the CPI’s allocation of 3.0% for their education (part of the 6.0% listed above) will seem laughable. Your inflation mileage will vary depending on your personal situation.

On top of that, there is the problem of “hedonic adjustment”, where the BLS tries to adjust for the changing quality of goods and services in the CPI basket. This is where data nerds really have a problem with the CPI, so here’s how it works:

#1: The BLS adjusts the following CPI components for changes in quality over time:

  • Owners equivalent rent (imputed homeowner inflation): 24.0% of current CPI
  • Rent of primary residence: 7.9%
  • Men’s, women’s and children’s apparel and footwear: 3.1%
  • Landlines, Internet service and cable/satellite TV: 2.9% of CPI
  • Televisions, major appliances, video/photo equipment: 0.2%
  • Telephone hardware (smartphones): 0.07%

#2: If the last 2 numbers look low, now you understand why hedonic adjustment is so problematic. Televisions, for example, have gotten a lot better in the last 5 years but are they 40% “cheaper” as a result? That’s what the CPI weights imply:

  • In March 2014, the weighting for televisions in the CPI was 0.159%
  • Now, it is 0.096%
  • The net change: a 40% reduction in the CPI weight for TVs
  • More consistent with what most consumers see: TVs are basically the same price as 5 years ago, but they do a lot more (streaming Netflix, for example).

#3: On the flip side, let’s look at the cost of shelter, which the BLS includes in its list of components that go through hedonic adjustment but also vary in weight based on market prices through time:

  • In March 2014, Rent of Primary Residence was 7.0% of the CPI basket. Now, it is 7.9%, or a 13% greater weight.
  • In March 2014, Owner’s Equivalent Rent was 23.8% of the CPI. Now, it is 24.0%, or a 1% greater weight.
  • The takeaway: in contrast to TVs, rising household expense for housing as a percent of income makes sense.

The bottom line on CPI: it may be an imperfect measure but it gets more things right than wrong. The rapid pace of technological advance has played havoc with personal electronics, but the weights here pale in significance compared to housing, energy, food, and other staples expenses. On those counts, the CPI does a reasonably good job.

Now, if you still are not convinced, check out something called the “Billion Prices Project”, a product that scrapes daily online prices and calculates its own US inflation rate. The data goes back to 2009, and more often than not their price index tracks CPI quite closely. You can see that data here:


BLS Hedonic Adjustments:

Current and prior CPI reports with weightings: