Macroeconomic Crises Since 1870

The estimated probability of disaster [a decline in national income or consumption of more than 10 percent in a year] is around 3.5 percent per year, with an average size of 22 percent and an average duration of 3.5 years. Typically, GDP and consumption fall concurrently, though consumption tends to fall proportionately more in wartime.

An economic crisis may be defined as a situation in which a nation's per capita GDP or consumption suffers a fall of at least 10 percent over a short period. For example, in the United States from 1929 to 1933, per capita GDP fell by 29 percent, while per capita consumer spending fell by 21 percent. In France, per capita GDP fell by 41 percent from 1939 to 1944, while per capita consumer spending fell by 58 percent from 1938 to 1945. Because large macroeconomic disasters are rare, determining their probability and the distribution of disaster sizes requires long time series of national-accounts data for many countries. In Macroeconomic Crises Since 1870 (NBER Working Paper No. 13940), co-authors Robert Barro and Jose Ursua expand the scope and reliability of the Maddison data and examine patterns in the long-term GDP and consumer-spending data. Among other things, the researchers study the interplay between macroeconomic variables and rates of return on various financial assets.

The research focuses on cases with full annual time series from before 1914 and often back to 1870. Using these data, the study identifies 152 GDP crises for 36 countries and 95 consumption crises for 24 countries. The principal disaster events worldwide were World War II, World War I, the Great Depression, post-World War II crises in Latin America and Asia, and possibly the Great Influenza Epidemic of 1918-20. The estimated probability of disaster is around 3.5 percent per year, with an average size of 22 percent and an average duration of 3.5 years. Typically, GDP and consumption fall concurrently, though consumption tends to fall proportionately more in wartime. For example, during both world wars, GDP increased in the United Kingdom while consumer expenditure fell sharply - the difference chiefly being attributable to the growth in military spending.

Long-term data on rates of return for stocks, bills, and bonds come from Global Financial Data. Simple models calibrated with the new data on macroeconomic disasters turn out to be consistent with the average equity premium of around 7 percent and the average real bill rate of around 1 percent shown in the financial-returns data. This match between theory and data requires a reasonable coefficient of relative risk aversion of around 3.5. This result is robust to several variations in specification and sample, except for limiting the sample to non-war crises, a selection that eliminates most of the largest declines in personal consumer expenditure and GDP.

Barro and Ursua plan a statistical analysis that uses all the time-series data and includes estimation of long-term effects of crises on levels and growth rates of personal consumer expenditure and GDP. They will also allow for time-varying disaster probability, an extension needed to account for the high volatility of stock returns. And, they will study the bond-bill premium (empirically around 1 percent). In addition, they will expand the 24-country sample for personal consumer expenditure and the 36-country sample for GDP. Candidates for this expansion are Malaysia and Singapore (with gaps around World War II); Russia back to the pre-World War I period; Turkey (where data for the Ottoman Empire have to be added through World War I); and Ireland (where data are missing prior to independence). The researchers are also reassessing the pre-1929 data for the United States, including the period of missing information during the Civil War years.

The authors hope to go further in measuring the division of personal consumer expenditure between durables and non-durables, and possibly government consumption as well. They also plan to construct time series for personal consumer expenditure and GDP per capita at regional levels, such as the OECD, Western Europe, Latin America, or even the "world." Such study is valuable when countries are integrated through financial and other markets. Finally, Barro and Ursua are working on a different approach to measuring time-varying disaster probabilities, this one using U.S. data since the early 1980s on prices of stock-index options to gauge changing market perceptions of the likelihood of adverse shocks.

-- Matt Nesvisky