Are We at the Start of a Recession?

On Thursday morning, April 28, the Bureau of Economic Analysis (BEA) released its “advance” estimate for the change in real GDP during the first quarter of 2022. As shown in the first line of the following table, somewhat surprisingly, the estimate showed that real GDP had declined by 1.4 percent during the first quarter. The Federal Reserve Bank of Atlanta’s “GDP Now” forecast had indicated that real GDP would increase by 0.4 percent in the first quarter. Earlier in April, the Wall Street Journal’s panel of academic, business, and financial economists had forecast an increase of 1.2 percent. (A subscription may be required to access the forecast data from the Wall Street Journal’s panel.)

Do the data on real GDP from the first quarter of 2022 mean that U.S. economy may already be in recession? Not necessarily, for several reasons:

First, as we note in the Apply the Concept, “Trying to Hit a Moving Target: Making Policy with ‘Real-Time’ Data,” in Macroeconomics, Chapter 15, Section 15.3 (Economics, Chapter 25, Section 25.3): “The GDP data the BEA provides are frequently revised, and the revisions can be large enough that the actual state of the economy can be different for what it at first appears to be.”

Second, even though business writers often define a recession as being at least two consecutive quarters of declining real GDP, the National Bureau of Economic Research has a broader definition: “A recession is a significant decline in activity across the economy, lasting more than a few months, visible in industrial production, employment, real income, and wholesale-retail trade.” Particularly given the volatile movements in real GDP during and after the pandemic, it’s possible that even if real GDP declines during the second quarter of 2022, the NBER might not decide to label the period as being a recession.

Third, and most importantly, there are indications in the underlying data that the U.S. economy performed better during the first quarter of 2022 than the estimate of declining real GDP would indicate. In a blog post in January discussing the BEA’s advance estimate of real GDP during the fourth quarter of 2021, we noted that the majority of the 6.9 percent increase in real GDP that quarter was attributable to inventory accumulation. The earlier table indicates that the same was true during the first quarter of 2022: 60 percent of the decline in real GDP during the quarter was the result of a 0.84 decline in inventory investment.

We don’t know whether the decline in inventories indicates that firms had trouble meeting demand for goods from current inventories or whether they decided to reverse some of the increases in inventories from the previous quarter. With supply chain disruptions continuing as China grapples with another wave of Covid-19, firms may be having difficulty gauging how easily they can replace goods sold from their current inventories. Note the corresponding point that the decline in sales of domestic product (line 2 in the table) was smaller than the decline in real GDP.

The table below shows changes in the components of real GDP. Note the very large decline exports and in purchases of goods and services by the federal government. (Recall from Macroeconomics, Chapter 16, Section 16.1, the distinction between government purchases of goods and services and total government expenditures, which include transfer payments.) The decline in federal defense spending was particularly large. It seems likely from media reports that the escalation of Russia’s invasion of Ukraine will lead Congress and President Biden to increase defense spending.

Notice also that increases in the non-government components of aggregate demand remained fairly strong: personal consumption expenditures increased 2.7 percent, gross private domestic investment increased 2.3 percent, and imports surged by 17.7 percent. These data indicate that private demand in the U.S. economy remains strong.

So, should we conclude that the economy will shrug off the decline in real GDP during the first quarter and expand during the remainder of the year? Unfortunately, there are still clouds on the horizon. First, there are the difficult to predict effects of continuing supply chain problems and of the war in Ukraine. Second, the Federal Reserve has begun tightening monetary policy. Whether Fed Chair Jerome Powell will be able to bring about a soft landing, slowing inflation significantly while not causing a large jump in unemployment, remains the great unknown of economic policy. Finally, if high inflation rates persist, households and firms may respond in ways that are difficult to predict and, may, in particular decide to reduce their spending from the current strong levels.

In short, the macroeconomic forecast is cloudy!

Source: The BEA’s web site can be found here.

New! – 4/07/22 Podcast – Authors Glenn Hubbard & Tony O’Brien revisit the role of inflation in today’s economy & likely Fed responses in trying to manage it.

Authors Glenn Hubbard and Tony O’Brien reconsider the role of inflation in today’s economy. They discuss the Fed’s possible responses by considering responses to similar inflation threats in previous generations – notably the Fed’s response led by Paul Volcker that directly led to the early 1980’s recession. The markets are reflecting stark differences in our collective expectations about what will happen next. Listen to find out more about the Fed’s likely next steps.

The Remarkable Movement in Inventory Investment in the New GDP Numbers

The Bureau of Economic Analysis (BEA) released its “advance estimate” of real GDP for the fourth quarter of 2021 on January 27, 2022. (The BEA’s advance estimate is its first, or preliminary, estimate of real GDP for the period.) At an annual rate, real GDP grew by 6.9 percent in the fourth quarter, which was a rate well above what most economists had forecast.  It’s always worth bearing in mind that the advance estimate will be revised several times in future BEA reports, but at this point the growth rate is the highest since the second quarter of 2000. 

The following table shows the interesting fact that final sales of goods and services (line 2) grew only about 2 percent, higher than in the third quarter of 2021, but well below the growth in sales during the previous four quarters. In fact, more than 70 percent of the growth in real GDP during the quarter took the form of increases in inventories (line 3).

Is the fact that economic growth during the quarter mainly took the form of businesses accumulating inventories bad news for the economy? Most likely not. It is true that we often sees firms accumulate inventories at the beginning of a recession. This outcome occurs when firms are too optimistic about sales and end up adding goods to inventory that they had expected to be able to sell. In other words, actual investment expenditures turn out to be greater than plannedinvestment expenditures; the difference between planned and actual investment being equal to the value of unintended inventory accumulation. (We discuss the relationship among planned investment, actual investment, and unintended inventory accumulation in Economics, Chapter 22, Section 22.1 and in Macroeconomics, Chapter 12, Section 12.1.)

It’s possible that some of the inventories firms accumulated during the fourth quarter of 2021 were the result of sales being below the level that the firms had forecast. During the quarter, the Omicron variant of the Covid virus was spreading in several parts of the United States and some consumers cut back their purchases, partly because they were more reluctant to enter stores. It seems likely, though, that the majority of the inventory accumulation was voluntary—and therefore part of planned investment—as firms attempted to rebuild inventories they had drawn down earlier in the year. Some firms also may have decided to hold more inventories than they typically had prior to the pandemic because they wanted to avoid missing sales in case Omicron resulted in further disruptions to their supply chains. 

Source:  The BEA’s website can be found at this link.

We Will Never See Anything Like It Again: Movements in Real GDP during the Covid-19 Recession

There are a number of ways in which the Covid-19 pandemic was unlike anything the United States has experienced since the 1918 influenza pandemic. Most striking from an economic perspective were the extraordinary swings in real GDP. The following figure shows quarterly changes in real GDP seasonally adjusted and calculated at an annual rate. There were three recessions during this period (shown by the shaded areas).

The first of these recessions occurred during 2001 and was similar to most recessions in the United States since 1950 in being short and relatively mild. Real GDP declined by 1.5 percent during the third quarter of 2001. The recession of 2007–2009 was the most severe to that point since the Great Depression of the 1930s. The worst periods of the 2007–2009 were the fourth quarter of 2008, when real GDP declined by 8.5 percent—the largest decline to that point during any quarter since 1960—and the first quarter of 2009, when real GDP declined by 4.6 percent. 

Turning to the 2020 recession, during the first quarter of 2020, only at the end of which did Covid-19 begin to seriously affect the U.S. economy, real GDP declined by 5.1 percent. Then in the second quarter a collapse in production occurred unlike anything previously experienced in the United States over such a short period: Real GDP declined by 31.2 percent. But that collapse was followed in the next quarter by an extraordinary recovery in production when real GDP increased by 33.8 percent—by far the largest increase in a single quarter in U.S. history.

The following figure shows the changes in the components of real GDP during the second and third quarters of 2020. In the second quarter of 2020, consumption spending declined by about the same percentage as GDP, but investment spending declined by more, as many residential and commercial construction projects were closed. Exports declined by nearly 60 percent and imports declined by nearly as much as many ports were temporarily closed. In the third quarter of 2020, many state and local governments relaxed their restrictions on business operations and the components of spending bounced back, although they remained below their levels of late 2019 until mid-2021. 

Even when compared with the Great Depression of the 1930s, the movements in real GDP during the Covid-19 pandemic stand out for the size of the fluctuations. The official U.S. Bureau of Economic Analysis data on real GDP are available only annually for the 1930s. The following figure shows the changes in these annual data for the years 1929 to 1939. As severe as the Great Depression was, in 1932, the worst year of the downturn, real GDP declined by less than 13 percent—or only about a third as much as real GDP declined during the worst of the 2020 recession.

We have to hope that we will never again experience a pandemic as severe as the Covid-19 pandemic or fluctuations in the economy as severe as those of 2020.

Source: U.S. Bureau of Economic Analysis. Note: Because the BEA doesn’t provide an estimate of real GDP in 1928, our value for the change in real GDP during 1929 is the percentage change in real GDP per capita from 1928 to 1929 using the data on real GDP per capita compiled by Robert J. Barro and José F. Ursúa. LINK

NEW! – 02/19/21 Podcast – Authors Glenn Hubbard & Tony O’Brien discuss early thoughts on the Biden Administration’s economic plan.

Authors Glenn Hubbard and Tony O’Brien discuss early thoughts on the Biden Administration’s economic plan. They consider criticisms of the most recent stimulus packages price tag of $1.9B that it may spur inflation in future quarters. They offer thoughts on how this may become the primary legislative initiative of Biden’s first term as it crowds out other potential policy initiatives. Questions are asked about what bounce we may see for the economy and comparisons are made to the Post World War II era. Please listen and share with students!

The following editorials are mentioned in the podcast:

Glenn Hubbard’s Washington Post Editorial with Alan Blinder

Olivier Blanchard’s comments on the Stimulus in a Peterson Institute for International Economics post

Larry Summer’s WaPo editorial about the risks of the stimulus:

Just search Hubbard O’Brien Economics on Apple iTunes or any other Podcast provider and subscribe!

Please listen & share!

Does Automation Lead to Permanent Job Losses?

This post on the Federal Reserve Bank of St. Louis’s Page One blog discusses how the belief that automation can lead to permanent job losses is an example of the “lump of labor” fallacy. Click HERE to read the article.

The post refers to the circular-flow diagram, which we discuss in Chapter 2 and in Chapter 18 in the textbook. We discuss the effects of automation and robots on the labor market in Chapter 16.

COVID-19 Update: Lessons from the Influenza Pandemic of 1918-1919 for the Coronavirus Pandemic of 2020


Economics – Chapter 7, The Economics of Health Care; Chapter 18, GDP:  Measuring Total Production and Income; Micro – Chapter 7, The Economics of Health Care; Macro Chapter 5, The Economics of Health Care; Chapter 8, GDP:  Measuring Total Production and Income; Essentials – Chapter 5, The Economics of Health Care; Chapter 12, GDP:  Measuring Total Production and Income

Lessons from the Influenza Pandemic of 1918-1919 for the Coronavirus Pandemic of 2020

The coronavirus pandemic of 2020 was by far the most serious epidemic to affect the United States since the influenza pandemic of 1918-1919, sometimes called the Spanish Flu.  Does the 1918-1919 influenza pandemic provide clues that help us predict how the coronavirus pandemic might affect the U.S. economy?   In this discussion, we:

  • Summarize scholarly and popular articles that address this question.
  • Provide links to the full articles.
  • Draw some tentative conclusions.


What Effect Did the 1918-1919 Influenza Pandemic Have on Death Rates and Real GDP?

            A National Bureau of Economic Research (NBER) Working Paper by Robert Barro of Harvard University, José Ursúa of Dodge & Cox, a mutual fund firm, and Joanna Weng of EverLife, an online food firm, estimate that the 1918-1919 pandemic killed about 39 million or about 2.0 percent of the world’s population.  An equivalent percentage of the world’s population today would be 150 million people.  The pandemic killed about 550,000 people in the United States or about 0.5 percent of the population.  If the death rate in the United States from the coronavirus were also 0.5 percent, the result would be 1.65 million deaths. 

            Barro, Ursúa, and Weng also estimate that the influenza pandemic reduced real GDP per capita in the typical country by 6 percent and real private consumption per capita by 8 percent. The comparable estimates for the United States are a decline in real GDP per capita of 1.5 percent and of real private consumption per capita by 2.0 percent.  They conclude that, “At this point, the probability that COVID-19 reaches anything close to the Great Influenza Pandemic seems remote, given advances in public-health care and measures that are being taken to mitigate propagation.”  The following table summarizes the Barro, Ursúa, and Weng estimates.

The What Effect Did Air Pollution Have on theWhat Effect Did Social Distancing Have on the 1918-1919 Influenza Pandemic?

A working paper by Sergio Correia, of the Federal Reserve Board, Stephan Luck of the Federal Reserve Bank of New York, and Emil Verner of the MIT School of Management examines the benefits and costs of non-pharmaceutical interventions (NPIs)—such as social distancing policies—during the 1918 pandemic.  They find that cities that implemented NPIs early and maintained them for longer experienced both lower mortality and higher economic growth, as measured by increases in manufacturing employment between 1914 and 1919. The cities of Seattle, Portland, Oakland, Los Angeles, and Omaha particularly stand out in this respect. Cities such as Pittsburgh, Philadelphia, and Boston that were slow to implement NPIs, or kept them in place for shorter periods, experienced both higher mortality rates and slower economic growth. The authors note: “This suggests that NPIs play a role in attenuating mortality, but without reducing economic activity. If anything, cities with longer NPIs grow faster in the medium term.”

            Correia, Luck, and Verner find substantial positive economic effect from early and prolonged implementation of NPIs: “Reacting 10 days earlier to the arrival of the pandemic in a given city increases manufacturing employment by around 5% in the post period. Likewise, implementing NPIs for an additional 50 days increases manufacturing employment by 6.5% after the pandemic.” They suggest that early implementation of NPIs may “flatten the curve,” keeping hospitals from being overwhelmed, and reducing mortality. They note that aggressive early use of NPIs appear to have successfully reduced both mortality rates and economic losses in Taiwan and Singapore.

What Effect Did Air Pollution Have on the 1918-1919 Influenza Pandemic?

Karen Clay and Edson Severnini of Carnegie Mellon University and Joshua Lewis of the Université de Montréal find that U.S. cities with worse air pollution—largely as the result of local utilities using more coal to generate electric power—suffered significantly higher death rates: “Cities that used more coal experienced tens of thousands of excess deaths in 1918 relative to cities that used less coal with similar pre-pandemic socioeconomic conditions and baseline health.”

Sources:  Robert J. Barro, José F. Ursúa, and Joanna Weng, “The Coronavirus and the Great Influenza Pandemic: Lessons from the “Spanish Flu” for the Coronavirus’s Potential Effects on Mortality and Economic Activity,” National Bureau of Economic Research, March 2020—The paper can be found here (the NBER is providing free access to working papers related to the coronavirus epidemic):; Sergio Correia, Stephan Luck, and Emil Verner, “Pandemics Depress the Economy, Public Health Interventions Do Not: Evidence from the 1918 Flu,” March 26, 2020—the paper can be found here: ; and Karen Clay, Joshua Lewis, and Edson Severnini, “Pollution, Infectious Disease, and Mortality: Evidence from the 1918 Spanish Influenza Pandemic,” Journal of Economic History, Vol. 78, No. 4, December 2018, pp. 1179-1209—The NBER working paper version can be found here.


Did People Who Were in Utero during the 1918-1919 Influenza Pandemic Suffer Negative Health Effects?

In an influential academic article published in 2008, Douglas Almond of Columbia University argued that people who were in utero during the influenza pandemic “displayed reduced educational attainment, increased rates of physical disability, lower income, lower socioeconomic status, and higher transfer payments compared with other birth cohorts.”  However, research by Ryan Brown of the University of Colorado, Denver and Duncan Thomas of Duke University provides evidence that the women who became pregnant during the pandemic were likely to be from lower socio-economic status than were women who became pregnant during earlier or later years. This result is attributable to the pandemic occurring during 1918 when many men of higher-than-average socio-economic status had been drafted to fight in World War I.   After correcting for the socio-economic status of the parents of people who were in utero during the pandemic, Brown and Thomas find that “there is little evidence that individuals born in 1919 have worse socio-economic outcomes in adulthood relative to surrounding birth cohorts.”

            Brian Beach of the College of William and Mary, Joseph P. Ferrie of Northwestern University, and Martin H. Saavedra of Oberlin College study this issue using individual data from the population censuses of 1920 and 1930 linked to World War II enlistment records. Their sample is large enough to contain many brothers, which allows them to completely control for the effects of socio-economic factors. They conclude that in utero exposure reduced high school graduation rates by about 2 percentage points but had no effect on adult height, weight, or body mass index (BMI).

Sources: Douglas Almond, “Is the 1918 Influenza Pandemic Over? Long- Term Effects of In Utero Influenza Exposure in the Post-1940 U.S. Population,” Journal of Political Economy, Vol. 114, No. 4, August 2006, pp. 672-712; Ryan Brown and Duncan Thomas, “On the Long Term Effects of the 1918 U.S. Influenza Pandemic,” June 2018 (; and Brian Beach, Joseph P. Ferrie, and Martin H. Saavedra, “Fetal Shock or Selection? The 1918 Influenza Pandemic and Human Capital Development,” National Bureau of Economic Research, Working Paper 24725, June 2018 (


An article on notes that compared with the 1918-1919 influenza pandemic, the coronavirus pandemic may turn out to be more contagious but with a lower death rate (although the death rate from the coronavirus appeared higher when the article was first published).

            In an opinion column in the New York Times, John Barry, a professor of public health at Tulane University and the author of the best-known history of the 1918-1919 pandemic, notes that an analysis of differences in the policies enacted among U.S. cities during the influenza pandemic indicates that when social distancing happened “before a virus spreads throughout the community, [it] did flatten the curve”—that is, it avoided a spike in deaths that would overwhelm hospitals.  

            An article on has some interesting graphs showing death rates from the influenza pandemic across many U.S. cities. Some cities experienced two peaks during 1918, while others did not. The article, which summarizes earlier epidemiological research, concludes that “death rates were around 50 percent lower in cities that implemented preventative measures early on, versus those that did so late or not at all.” The cities that kept these measures in place longest were the ones that did not experience a second spike in death rates.

            Although the article attributes the relatively low death rate in New York City during the influenza pandemic to the early implementation of quarantine and other public health measures, this op-ed in the New York Times by historian Mike Wallace indicates that the city did not close its public schools or most of its theaters, although it did levy fines to enforce a prohibition on public spitting—a common habit among men at the time.

            This essay in the Wall Street Journal by a medical doctor and adjunct professor at the Duke Global Health Institute reviews the course of the influenza pandemic in the United States and attempts to draw some lessons for the current coronavirus pandemic. The essay discusses the often-mentioned mistake by the Philadelphia city government of allowing a crowd of 200,000 to attend a parade to sell Liberty Loans: “within 72 hours, every bed in the city’s 31 hospitals was filled.” The doctor  notes that “cities which implemented isolation policies (such as quarantining houses where influenza was present) and ‘social distancing’ measures (such as closing down schools, theaters and churches) saw death rates 50% lower than those that did not.”

Tentative Conclusions

There are limits to drawing parallels between the 1918-1919 influenza pandemic and the 2020 coronavirus pandemic for at least three key reasons:

  1. Epidemiological profiles  The viruses are different and have different epidemiological profiles.  The coronavirus appears to be more contagious than the 1918 influenza virus but has a lower death rate.  The influenza virus killed more people in the prime age groups, particularly people aged 25 to 34. The influenza virus also had a higher death rate among children. In both pandemics, men were more likely to die from the coronavirus than women.
  2. Medical knowledge, drugs, and equipment   The state of medical knowledge is greater today than it was in 1918-1919, which may be helping to reduce the death rate. There are many more intensive care units in hospitals—such units were rare in hospitals in 1918.  Antibiotics had not yet been discovered in 1918, so people died of secondary infections, particularly pneumonia, who have been saved in the current pandemic. There are also antiviral drugs available today that were unknown in 1918, although at this writing (early April 2020) it was unclear whether any current antiviral drug will be effective against the coronavirus. No medical equipment similar to current-day respirators were available in 1918, which made it difficult for people suffering from severely reduced lung capacity to survive.
  3. Knowledge of policies during 1918  Knowledge of the course of pandemics is greater now than in 1918, partly because the breadth and scope of the influenza pandemic made it the subject of close study during the decades since.  Three lessons in particular have affected the response to the coronavirus pandemic. First, errors committed by local government officials in 1918, notably the mayor of Philadelphia allowing the Liberty Loan parade to take place despite warnings from local health officials, have been widely publicized. As a result these errors have largely been avoided in 2020. For instance, most cities canceled their St. Patrick’s Day parades and U.S. sports leagues shut down promptly in mid-March.  Second, we know that those cities that enacted policies of quarantines and social distancing in 1918-1919 had the lowest death rates.   This knowledge contributed to government officials and the general public supporting similar policies in 2020. Finally, we know that the 1918-1919 pandemic occurred in three waves—with the second wave during the fall of 1918 being the worst in the United States.  In 2020, the public and government officials are aware that the initial coronavirus wave in the late winter-early spring of 2020 would likely be followed by additional waves in the absence of an effective vaccine or the development (or repurposing) of therapeutic drugs.

Despite the differences between the 1918-1919 and 2020 pandemics, we can offer several tentative observations:

1. The decline in real GDP from the coronavirus pandemic is likely to be greater than the decline in real GDP from the influenza pandemic.  As noted earlier, Barro, Ursúa, and Weng found a surprisingly small decline in real GDP as a result of the 1918-1919 pandemic. Determining the macroeconomic effects of the pandemic is difficult because it began during the last year of World War I and because it preceded the short, but sharp, recession that lasted from January 1920 to July 1921. Most economic historians don’t believe that the pandemic was a significant cause of the 1920-1921 recession.

 This period was also before the U.S. Bureau of Economic Analysis began collecting GDP data. Several economists have estimated changes in GDP during these years, but their estimates differ significantly.  Nathan Balke of Southern Methodist University and Robert Gordon of Northwestern University estimate that real GDP declined by 2.9 percent from 1918 to 1919. Christina Romer of the University of California, Berkeley estimates that real GDP increased by 1.1 percent from 1918 to 1919.  Robert Barro and José Ursúa estimate that real GDP per capita declined by 3.4 percent from 1918 to 1919 (note that the 1.5 percent decline quoted earlier is their estimate of how much of the total 3.4 percent decline was due to the pandemic).

The decline in U.S. real GDP during the second quarter of 2020 is likely to be substantial—perhaps as high as 20 percent on an annualized basis. Still unknown is whether the U.S. economy will experience a V-shaped recession—a sharp decline in real GDP followed by a sharp rebound—or what has been called a “Nike swoosh-shaped recession”—a sharp decline followed by a slower recovery.  If the United States experiences a swoosh-shaped recession, the decline in real GDP is likely to significantly exceed the decline during 1918-1919.

As we discussed earlier, during 2020 the social distancing policies recommended by the federal government and implemented by many states and cities far exceeded anything implemented in 1918.  In 1918, even New York City, which historians generally praise for its vigorous public health response, allowed its schools, restaurants, and most theaters to remain open.  In comparing 2020 with 1918, we can say that in fighting the coronavirus, the United States was willing in 2020 to accept large declines in production and employment in order to reduce projected death rates.  In 1918, perhaps inadvertently, the United States   experienced a higher death rate in part because economic life was not disrupted to the extent that it was in 2020.

2. The long-range health effects of the coronavirus pandemic are not known.  As we noted earlier, Douglas Almond’s research has been frequently cited for its finding that people who were in utero in 1918 suffered substantial negative health and socioeconomic effects as adults. As discussed, recent unpublished research has called Almond’s results into question. It remains unclear whether the influenza pandemic had lasting effects either in the way Almond’s study suggests or perhaps because some people who recovered from the flu suffered reduced lung capacity or were more prone to developing pneumonia or other respiratory diseases later in life. Here’s the key point:  The coronavirus is not a type of influenza, so the long-run health effects from the influenza epidemic—if there were any—may not be relevant in evaluating coronavirus.

In 2020, there was some discussion in the news media that patients with coronavirus who needed to use ventilators might subsequently suffer from reduced lung capacity. Modern ventilators were not available in 1918, so that episode doesn’t provide evidence for the consequences of their widespread use.

Sources: For 1918-1919 real GDP estimates: Balke and Gordon: Nathan S. Balke and Robert J. Gordon, “The Estimation of Prewar Gross National Product: Methodology and New Evidence,” Journal of Political Economy, Vol. 97, No. 1, February 1989, pp. 38–92; Christina Romer: Christina D. Romer, “The Prewar Business Cycle Reconsidered: New Estimates of Gross National Product, 1869–1908,” Journal of Political Economy, Vol. 97, No. 1, February 1989, pp. 1–37; Barro and Ursúa: Excel file posted under “Data Sets” on Barro’s homepage,