Andrew Atkeson, and Patrick J. Kehoe, 2004.
Are deflation and depression empirically linked? No, concludes a broad historical study of inflation and real output growth rates. Deflation and depression do seem to have been linked during the 1930s. But in the rest of the data for 17 countries and more than 100 years, there is virtually no evidence of such a link.
According to standard economic theory, deflation is the necessary consequence of optimal monetary policy. In 1969, Milton Friedman argued that under the optimal policy, the nominal interest rate should be zero and the price level should fall steadily at the real rate of interest. Since then, Friedman’s argument has been confirmed in a formal setting. (See, for example, V. V. Chari, Lawrence Christiano, and Patrick Kehoe 1996 and Harold Cole and Narayana Kocherlakota 1998.)
Most policymakers, however, are extremely reluctant to implement any policy that would lead to deflation. This reluctance seems to stem from the experience of the Great Depression, in which deflation and depression appear to have been tightly linked. (See, for example, Ben Bernanke and Kevin Carey 1996.) That experience has led to theories in which deflation leads to depression. The quantitative ability of those theories to account for the Great Depression is now being debated. (See, for example, Cole and Lee Ohanian 2001.)
Here we examine the empirical relationship between deflation and depression in a broad historical context, including but not limited to the Great Depression. We use a panel data set on inflation and real output growth for 17 countries and more than 100 years. To focus on medium-term fluctuations, we break the time series on inflation and real output growth for each country into five-year episodes, and for each episode, we compute the average annual inflation rate and the average annual real output growth rate. For any episode, we define a deflation as a negative average inflation rate and a depression as a negative average real output growth rate. Throughout, we restrict attention to moderate inflations, those with average annual inflation below 20 percent.
Our main finding is that the only episode in which we find evidence of a link between deflation and depression is the Great Depression (1929—34). We find virtually no evidence of such a link in any other period. Here we have made no attempt to distinguish anticipated from unanticipated deflations, while theory, of course, makes a sharp distinction. Optimal monetary policy in a broad class of models with nominal rigidities dictates engineering small anticipated deflations and avoiding unanticipated deflations altogether. To the extent that the deflation in the Great Depression is thought of as unanticipated, as in most existing theories, this episode is not relevant for evaluating the costs of anticipated deflation. Our finding thus suggests that policymakers’ fear of anticipated policy-induced deflation that would result from following, say, the Friedman rule is greatly overblown.
We note here that since WWII, one country has come close to having both a depression and a deflation: Japan in the late 1990s. Is Japan’s recent slowdown from a historically high average growth primarily due to its very low inflation rates? We doubt it. Since 1960, Japan’s average growth rates have basically fallen monotonically, and since 1970, its average inflation rates have too. Attributing this 40-year slowdown to monetary forces is a stretch. More reasonable, we think, is that much of the slowdown is the natural pattern for a country that was far behind the world leaders and had begun to catch up.
I. The Data
All of our series end in 2000. Here are the countries (and the years their data start): Argentina (1885), Australia (1862), Brazil (1861), Canada (1870), Chile (1908), Denmark (1871), France (1820), Germany (1830), Italy (1867), Japan (1885), the Netherlands (1900), Norway (1865), Portugal (1833), Spain (1849), Sweden (1861), the United Kingdom (1870), and the United States (1820). For all countries except Australia and Denmark, the data up to 1980 are taken from Arthur Rolnick and Warren Weber (1997). The data for Australia and Denmark up to 1980 are taken from David Backus and Kehoe (1992). The data for 1980 on are from the International Financial Statistics of the International Monetary Fund. (The pre-1980 data are missing some observations.)
We choose the five-year episodes to start and end with years ending in 9 or 4. This way, we include the Great Depression of 1929—33 in a single episode covered by the five-year average of 1929—34 and the depression of 1921—22 in the five-year average of 1919—24.
II. The Findings
A. The Great Depression Episode
We start with the Great Depression episode, 1929—34. The data for this episode do seem to show a link between deflation and depression–but not an overwhelmingly tight link.
In Figure 1, we plot average inflation and output growth for the 16 countries for which we have data for this period. (Chile’s output data are missing.) In 1929—34, all 16 countries had deflation, 8 had deflation and depression, and the other 8 had deflation but no depression. For all 16 countries, a regression of output growth on a constant and the inflation rate has a slope coefficient of .40 (.28). (Here, and throughout, we record the standard error of a regression coefficient in parentheses.) The slope coefficient implies that, on average, a one percentage point lower inflation rate is associated with a drop in growth of .4 of a percentage point, say, from 3.40 to 3.00 percent. (We summarize all our regressions in Table 1.)
In the Great Depression episode, then, all countries had deflation, and half of them had depression. In a regression sense, there seems to be a link here. Researchers disagree on how strong that link is. Some, like Barry Eichengreen and Jeffrey Sachs (1985) and Bernanke and Carey (1996), argue that deflation and depression during the Great Depression were closely linked; Cole, Ohanian, and Leung (2003) argue to the contrary.
B. Outside the Great Depression
While the debate about the Great Depression episode is ongoing, our interest lies mainly in looking for a robust relationship in a broader historical context. If we find none, the Great Depression may have been a special experience with little to offer policymakers considering a deflationary policy today. And that is what we find.
In Figure 2, we plot average inflation and output growth rates for the 17 countries for all five-year periods except 1929—34. Here we see only 8 episodes with both deflation and depression. There are 65 episodes of deflation without depression and 21 of depression without deflation. Thus, 65 of 73 deflation episodes had no depression, and 8 of 29 depression episodes had no deflation. What is striking is that nearly 90% of the episodes with deflation did not have depression. In a broad historical context, beyond the Great Depression, the notion that deflation and depression are linked virtually disappears.
Note that most of the episodes in the data set that have deflation and no depression occurred under a gold standard; does that somehow make them irrelevant for shedding light on deflations under a fiat system? No. If we interpret such standards as a rule with some commitment, as many (like Bordo and Finn Kydland 1995) do, then these episodes seem at least as relevant for thinking about the effects of following a Friedman rule as are the episodes in the post-WWII data, when policy was more discretionary. And if we think that the world economy has changed so much since WWII that all the data before it sheds no light on what might happen in an economy today, then there is not much to discuss, since in the post-WWII period, there are no episodes of deflation. Moreover, as has been commonly noted, inflation is actually negatively related to output growth in the post-WWII data.
A more compelling objection is that the data from periods of world wars are just not relevant for other periods. We thus investigate the data excluding all war-related episodes. We find that war does seem to play a role in generating depression. Of the 21 depression episodes without deflation, 10 were related to a world war: 4 during WWI, 3 during WWII, and 3 right after WWII. (We view Japan’s dismal growth in 1949—54 as war-related.) But that does not change our result about deflation and depression. Based on all of the data outside of the Great Depression, a regression of output growth on inflation has a slope coefficient of .04 (.03). Excluding all the data from the war-related episodes (1914—19, 1939—44, 1944—49) and the 1949—54 episode for Japan gives a slope coefficient of .10 (.03). These data thus show little or no relationship between deflation and depression. This finding is consistent with that of Michael Bordo, John Landon-Lane, and Angela Redish (2002).
C. The Overall Relation
In Figure 3, we display all of the data. Here the regression line has a slope coefficient of .08 (.03). Thus, a one percentage point drop in inflation is associated with a drop in the average real growth rate of just .08 of a percentage point, say, from 3.08 to 3.00 percent.
The relation between deflation and depression does differ pre- and post-WWII. The slope coefficients are .11 (.04) in the pre-WWII data and −.03 (.04) in the post-WWII data. The postwar result is consistent with the results of many studies, which typically associate higher inflation and lower growth.
In the postwar data, Japan’s experience in the 1990s is the closest that any major economy has come to having both a deflation and a depression. Economists debate whether this experience is evidence of a link between these conditions. To put Japan’s recent experience in context, consider Figure 4, where we plot its annual rates of output growth and inflation as well as their five-year averages since 1960.
Figure 4 essentially shows a 40-year decline in the output growth rate (4A) and a 30-year decline in the inflation rate (4B). We think standard theories – either neoclassical or new Keynesian – would have a hard time blaming Japan’s secular growth slowdown on its secular decline in inflation. But that slowdown would naturally arise in many growth models in which a country grew rapidly in the early postwar period because it had been below its steady-state growth path; as it caught up to this path, its growth would naturally slow.
Has Japan’s growth slowed too much? Not relative to countries like Italy and France. At 1.41, Japan’s growth in the 1990s was dismal compared to the U.S growth of 3.20, but not compared to the growth of Italy (1.61) or France (1.84).
III. Concluding Remarks
The data suggest that deflation is not closely related to depression. A broad historical look finds many more periods of deflation with reasonable growth than with depression and many more periods of depression with inflation than with deflation. Overall, the data show virtually no link between deflation and depression.
This study simply characterizes the relation in the raw data between deflation and output growth, with no attempt to control for anything, like the type of monetary regime or the extent to which the deflation was anticipated. Perhaps a link between deflation and depression could be teased out of the data with a well-motivated set of controls. Our contribution here is to note that without such controls, the data show no obvious relationship. The bar has thus been raised for those who claim that deflation and depression are closely linked.