Debt and Growth Revisited (Reinhart, Rogoff)

High-debt episodes in the sample

The episodes that attract our interest are those where debt levels were historically high. As convenient as it is to focus exclusively on a particular country or a single episode for a single country (like the US around World War II, where the data is readily available, or an interesting ongoing case, like Japan), the basis for an empirical regularity is multiple observations. Because our data span 44 countries with many going back to the 1800s or (at least the beginning of the 19th century), our analysis is based on all the episodes of high (above 90%) debt for the post World War II period; for the pre-war sample it covers all those that are encompassed by the availability of data. Table 1 (from Reinhart and Rogoff 2010a) describes the coverage and the basic statistics for the various debt levels for the advanced economies.⁴

It is common knowledge that the US emerged after World War II with a very high debt level. But this also held for Australia, Canada, and most markedly the UK, where public/debt GDP peaked at near 240% in 1948. These cases from the aftermath of World War II are joined in our sample by a number of peacetime high-debt episodes:

• the 1920s and 1980s to the present in Belgium,
• the 1920s in France,
• Greece in the 1920s,
• 1930s and 1990s to the present,
• Ireland in the 1980s,
• Italy in the 1990s,
• Spain at the turn of the last century,
• the UK in the interwar period and prior to the 1860s and, of course,
• Japan in the past decade.

As will be discussed, episodes where debt is above 90% are themselves rare and, as shown in Table 1, a number of countries have never had debt entries above 90%.

Debt thresholds and nonlinearities: the 90% benchmark

Thresholds and non-linearities play a key role in understanding the relationship between debt and growth that should not be ignored in casual re-interpretations.

(i) Thresholds. Those who have done data work know that mapping vague concepts like “high debt” or “overvalued exchange rates” into workable definitions requires arbitrary judgments about where to draw lines; there is no other way to interpret the facts and inform the discussion. In the case of debt, we worked with four data “buckets”: 0-30%, 30-60%, 60-90%, and over 90%. The last one turned out to be the critical one for detecting a difference in growth performance, so we single it out for discussion here.