Why Inflation Still Looms Large for US Voters

If we measured inflation as the BLS did in the 1970s, the recent bout of inflation would have been even higher than the worst of the 1970s! It really is as bad now as it was then.

The main difference is that the old measure counts the price and interest rate you have to pay to buy a new house as the cost of the house, while the new measure is based on what it costs to rent a house. You can see that the new and old methods are the same in the initial run-up of inflation starting in 2021, but then the synthetic old measure shoots up when the Fed raised interest rates, while rents did not rise so quickly.

The change in measurement makes it hard to compare over time. It would be nice to recompute the old data with new and better measurements. But when you can’t do that, computing new data with the old measurement is a nice way to compare over time on an even basis. The University of California at Berkeley’s Christina Romer has investigated changes in employment volatility using this method: Unemployment seems less volatile after WWII than it was before 1930, which many economists have attributed to the wonders of Keynesian stabilization policy. But when you construct modern unemployment data using old methods, they look the same, Romer demonstrates.

Bolhuis, Cramer, Schulz, and Summers generalize the idea by adding interest costs to buy cars and other goods to their calculations of the cost of living. You can see in the chart below that it’s almost the same story.

What’s the right way to compute inflation?

The new way is closer to right if the question is: What is the change in the cost of living right now for the average person? We tend to jump to answers without stating the question. Stating the question is a good idea.

Most people live in older houses with fixed mortgages, so higher prices and mortgage rates for new houses don’t affect them. People who rent are not affected right now by higher house prices. While higher interest rates are a cost to borrowers, and higher house prices a cost to buyers, higher interest rates are a boon to savers and higher house prices a boon to downsizers. Those wash, on balance.

But to the average American, the idea that inflation is falling and they are better off because they could rent a house just like theirs for less money is a bit of a head-scratcher, to put it mildly. It’s right, economically. The “homeowner” comprises a landlord plus a renter. The rent that the renter half pays is the cost of living. Lower rental income and higher interest or purchase costs for the landlord half are not part of the cost of living, just as all business profits and capital gains and losses are not. Still, explain that to the average voter.

There are a couple of central problems here, with some unresolved economics. The Consumer Price Index is essentially a static concept. It’s the cost of buying a typical basket of consumption goods. If you think apples, bananas, and strawberries, that makes sense. But if you think cars, houses, and stocks, all of a sudden it doesn’t make that much sense anymore. The world isn’t static. The future matters.

The right question might be: How much does it cost me today to buy my lifetime consumption? If rents are low, but house prices and interest costs are high, it’s likely that rents will rise in the future. (Really. Chicago Booth’s Eugene F. Fama and Dartmouth’s Kenneth R. French have provided recent evidence.) Buying a house locks in the right to live there forever. The cost of a lifetime of housing did go up, though today’s rents did not.

Similarly, if stocks go up, that typically does not mean expected future dividends rise, so the cost of saving for retirement has risen. There is a lot of talk of “asset price inflation.” Economists (including me) usually sneer about that being a relative price and not the price level. Don’t use the word inflation for car inflation, stock inflation, house inflation, and so forth. Those are individual and usually relative prices. Inflation means the average level of all prices, and a decline in the value of money. Period. But “asset price inflation” isn’t necessarily wrong, it’s just an answer to a different question. If you want to know the cost of providing for a lifetime of consumption, higher asset prices and lower real interest rates mean that cost has risen.

The CPI asks a different question, the cost of this year’s average (across people) consumption. The Lifetime Consumption Cost Index would be a fun thing to calculate.

In the same vein, we often look at different measures of inflation to try to forecast what properly measured inflation will be in the future. The core-versus-headline inflation argument comes down to this question. Core inflation ignores food and energy. “Well,” says the average person, “it’s awfully nice that prices excluding food and energy aren’t rising so fast, but I have to eat and buy gas.”

Is it dishonest to report core? Not if the goal is to forecast what inflation will be in the future. Food and energy prices are volatile. Volatile is thought to mean predictable: a rise in price today can reliably forecast a decline in that price in the future. If food and energy prices really are predictable in this way, core is more useful for forecasting future inflation than headline inflation is.

Another example: the BLS looks at the average rent people are currently paying. But rent is sticky. Landlords don’t raise the rent on existing houses and apartments nearly as quickly as they raise the rent on new leases. Thus, new rents are, I think, a good forecast of where average rents will be in the future.

I wish this idea were taken to its logical conclusion. The question is: What is headline inflation going to be next year? We should examine carefully and systematically how the various components such as core actually do in making such a forecast. My sense is that we may be able to reliably forecast the headline CPI by looking at the various components, but such forecasts are going to be less precise than most pundits think. Similarly, higher interest rates and house prices might well make sense as ingredients in forecasting next year’s rents and hence next year’s inflation.

Inflation dynamics

In the graphs, the standard (rental) measure of inflation turns around while the version that incorporates interest rates keeps going up. This makes a big difference in how we measure inflation dynamics, the response of inflation to interest rates.

Inflation came seemingly from nowhere to most analysts. For some of us, it was perfectly obvious: drop $5 trillion from helicopters, and inflation breaks out. Some economists chalk it up to simple supply and demand, but to me it is central that people do not expect fiscal surpluses to pay back that $5 trillion anytime soon. I note speculatively that inflation really broke out in February 2021, which is plausibly when it became really clear that large deficits would continue.

I think the easing of inflation was also perfectly predictable. A one-time fiscal shock gives a one-time price-level rise. Inflation eases when that is done. The Fed helps, and higher interest rates temporarily lower inflation, but it’s not central. (This is the explanation given by the fiscal theory of the price level.)

The puzzle for standard analysis is that inflation eased just as the Fed started raising rates, long before rates exceeded inflation, and with no recession. Adieu, Phillips curve.

Most estimates say inflation goes up gently for a year or two after a rate rise before falling gently, maybe. Interest rates lower inflation with “long and variable lags.” In this context, inflation easing one month after the Fed gently started raising rates is nearly miraculous. Talk shifts to “expectations”: somehow this time a few basis points of short rate showed everyone just how tough the Fed would be, though past rate rises took years to have any effect.

But this difference may simply come down to how we measure inflation. The data on which the standard view developed used the old measure of inflation, in which higher interest rates almost mechanically raise inflation for a while. The more recent event reflects the new measure. Perhaps inflation, measured as it is now, always declined quickly after interest-rate rises. Such a finding would also be wonderful for modern theory, which predicts an immediate effect and not “long and variable lags.”

To give a sense of how including interest rates in the CPI would affect our view of inflation dynamics, I did a rough computation of the interest-rate-adjusted CPI by adding 0.07 times the growth of the 30-year mortgage rate to the index. I picked 0.07 so that the peak would be about the same as that identified by Bolhuis, Cramer, Schulz, and Summers. You can see the results in the chart below.

Yes, I added the growth rate of the mortgage interest rate, not the level. We’re measuring inflation, the rate at which prices rise, and the level of the mortgage interest rate is the relevant price. The level of the mortgage rate goes up and stays up. Cumulative inflation, the rise in the price level, is about double under the old measure. Average people are annoyed by the price level, not the inflation rate.

The vertical line marks “liftoff,” when the Fed starts raising rates. You see the CPI turn around almost immediately. But since long-term rates follow the federal funds rate, the growth in mortgage rates peaks a good deal later.

The interest-rate-adjusted measure of the CPI captures something quite common in the zeitgeist: interest rates are a cost, and raising interest rates means more inflation in the short run.

Thus, maybe this observation resolves the “price puzzle” posed by historic estimates showing inflation rising a while after interest-rate rises. Maybe that was all spurious, and inflation, measured by the current method, really does turn around just as interest rates rise. Getting models to generate a delay has been devilishly hard, and to this day, most modern models (rational expectations, new Keynesian, forward looking) say that inflation jumps down the minute interest rates rise. Maybe the models are right after all. (Whew!)

The bottom line is much of measured inflation includes things such as imputed home rents, government services, healthcare, and so on. That’s the right thing to do, but don’t expect huge precision. The CPI measures the cost of the average consumption basket, but the relative price changes (such as home price relative to rent) affect different people differently, and a lot. It doesn’t measure lifetime consumption. It doesn’t really measure the value of the dollar, which is what the Fed should be focusing on.

There is a lot of interesting economics to be done in thinking about how to measure inflation. Just make sure to state the question before you jump to the answer.

John H. Cochrane is a senior fellow of the Hoover Institution at Stanford University and was previously a professor of finance at Chicago Booth. This essay is adapted from a post on his blog, The Grumpy Economist.