The Mismeasurement of Money and Its Consequences: A Quick Note on Inflation

This forms part of the background research for a longer post on the macroeconomics of free banking and competitive note issuance. I decided to break out some of the preliminary work to keep the whole thing at a manageable length.

Inflation is a sustained increase in the overall price level. Sounds simple, but how do we know that overall prices are going up? After all, some prices clearly increase regularly, but others exhibit dramatic declines both in relative and absolute terms: the price of TVs fell by 20% in just one year alone, between 2018 and 2019. Other prices are famously static, even over extremely long periods of time. Nor is simple averaging of prices sufficent to discern the overall price level because consumption is highly weighted towards certain types of goods: if hotel accommodation prices go up a lot, but food prices fall slightly, consumer purchasing power is likely to increase even if average prices have gone up across these two domains. Clearly we need some kind of overall index weighted by actual consumer spending to measure inflation, which is exactly what popular indices such as the consumer price index or retail price index aim to do.

Unfortunately, as simple as this sounds, our methods of measuring inflation are necessarily imperfect and filled with bias. In general, inflation estimates are upwardly biased i.e inflation is typically overestimated. The degree of the bias changes over time as statistical agencies update their methods, and varies considerably by country (e.g it is widely thought that Japanese inflation is considerably more overestimated than inflation in the U.S, where estimation methods are more sophisticated) but long-term powerful trends in the wider economy put a lot of upward pressure on the size of the bias.

The types of bias can be roughly counted as follows:

  1. Quality bias. The quality of goods and services generally get better over time in ways not always reflected in the price. If you buy a TV for X in one year, and then the next year get to buy a TV 10% better for the same price, your purchasing power has clearly increased even though prices are static. Inflation and deflation don’t have to show up in prices: they can also show up in quality losses and (much more typically) gains. Official statistics struggle to capture the extent of changes in quality, although the size of the bias has shrunk since the work of the Boskin Commission. It is worth noting that for many services (as opposed to good) there is no attempt to measure quality improvements at all - and this includes in some massive sectors such as healthcare!

  2. Substitution bias: the tendency of consumers to switch over time from higher cost to lower cost items while keeping utility constant. This can happen in a few different ways: consumers can switch between brands in response to price rises (e.g from Toyotas to Skodas), but also between different categories (eg from cars to cargo bikes, an increasingly common trend in densely populated countries where fuel prices are high).

  3. Outlet bias: new lower-cost retail outlets open, selling the same items as more expensive outlets but at lower prices.

Prior to the Boskin Commission, US CPI estimates assumed no substitution at all. Combined with the other types of bias, the the Commission estimated an long-term upward bias of about 1.1 percentage points in US CPI estimates. This is a very sizable bias! Assuming the Fed hits its inflation target of 2% for the year, true inflation will only be running at 0.9%. Subsequent reforms to CPI measurement eliminated some forms of the bias and lowered others, but subsequent research puts the upward bias still at around 0.85 percentage points, conservatively estimated..

Since the Commission reported, however, the economy has changed a great deal. Product turnover has increased, as products take an average just 20 months to disappear from the market compared 25-30 months to in the 1990s. Moreover, an explosion of free goods consumed in massive quantities (Google, Whatsapp, FB, Wikipedia) complicates inflation measurement, since none of these show up in conventional consumption metrics, which in turn has large implications for how we think about the maintenance of purchasing power over time. As investment in intangible assets (R&D, IP, firm-specific knowledge) grows rapidly as a share of GDP, the price of increasingly valuable things at the core of the economy becomes harder to put a precise value on and (it is very likely) increasingly easy to overestimate. The issue is discussed more fully here. and here.

Without getting too much into the weeds, though, the key takeaway is that any central bank that successfully hits a measured inflation rate of 2% is very likely hitting a true inflation rate of under 1% (especially for those countries with less sophisticated statistical agencies than the U.S), and it remains quite plausible that the years after the GFC but prior to the covid-19 pandemic were years of deflation (and, consequently, higher GDP and productivity growth than conventionally thought). In any case, “price stability” as modern central bankers today understand it is not actually an inflation rate of 2%: the desired true inflation rate is not far above zero given these persistent measurement issues (as explicitly mentioned as far back as 1992 by the Governor of the Bank of England here.).

As we develop our discussion of the macroeconomics of free banking and competitive note issuance - especially with non-fiat money as the underlying - the importance of this issue will be more apparent. The gap between historic experiences of near-zero inflation/deflation under commodity money and free banking vs the desired goals of central banks today may be much narrower than it appears.