The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
Trade Deficits Aren’t as Bad as You Think BY GEORGE ALESSANDRIA A lthough the amount of U.S. imports and exports has varied greatly over time, in recent years, the U.S. has been running trade deficits. Some people react to such trade deficits with doom and gloom; others cite them as evidence that foreign governments are not playing fair in U.S. markets; still others argue that deficits demonstrate that we are living beyond our means. In this article, George Alessandria offers an alternative view: Trade deficits have benefits. They shift worldwide production to its most productive locations, and they allow individuals to smooth out their consumption over the business cycle. We live in a global world. Americans drive automobiles produced in Germany and drink Italian wine. Europeans watch movies of Jedi Knights battling the Dark Side on televisions produced in Mexico. This was not always the case. For instance, the value of U.S. imports of goods and services has grown from 5.1 percent of gross domestic product (GDP) in 1969 to 15.2 percent George Alessandria is a senior economist in the Research Department of the Philadelphia Fed. This article is available free of charge at www. philadelphiafed.org/econ/br/index.html. www.philadelphiafed.org of GDP in 2004. Likewise, the value of U.S. exports of goods and services has grown from 5.3 percent of GDP in 1969 to 10.0 percent of GDP in 2004. The amount of U.S imports and exports has also varied quite a lot over time. At times, the U.S. has run trade surpluses, with exports exceeding imports, and at other times, it has run trade deficits, with imports exceeding exports. Recently, though, the U.S. has imported a lot more goods and services from abroad than it has exported to the rest of the world. In 2004, this resulted in the U.S. running a trade deficit of 5.2 percent of GDP. Through the third quarter of 2005, the trade deficit has averaged 5.7 percent of GDP. Some people react to the trade deficit with doom and gloom. They argue that the trade deficit is evidence that American firms are unproductive and can’t compete with foreign firms. Others point to it as clear evidence that foreign governments are not playing fair in U.S. markets. Still others argue that it demonstrates that we are living beyond our means. But there is an alternative view. In this view, these unbalanced trade flows have two benefits: They shift worldwide production to its most productive location, and they allow individuals to smooth out their consumption over the business cycle. According to this view, the trade balance declines, or moves into deficit, when a country’s firms or government is investing in physical capital to take advantage of productive opportunities. These investments expand the infrastructure, build capacity to access natural resources, and take advantage of new technologies. This increase in investment is financed in part by borrowing in international financial markets. By borrowing internationally, a country can invest more without cutting current consumption. When it repays this borrowing in the future, the trade balance increases or goes into surplus. In this respect, a trade deficit may be a sign of a growing and robust economy. Moreover, by increasing a country’s productive capacity, these unbalanced trade flows are vital to sustaining the economy’s expansion into the future. This view is consistent with some properties of the trade balance in the U.S. and other countries. MEASURING INTERNATIONAL TRANSACTIONS Before discussing the reasons that a country runs a trade deficit or surplus, it’s useful to review the different Business Review Q1 2007 1 measures of a country’s international transactions. These are recorded in the balance of payment accounts (Table 1). The two main components of the balance of payments are the current account and the capital and financial account. The current account records the value of currently produced goods and services, both imported and exported, as well as the international payment of interest, dividends, wages, and transfers. The capital and financial account records transactions in real and financial assets.1 The easiest way to understand the components of the balance of payments is to think of a monthly credit card statement. One part of the statement reports the difference between new charges and payments. This difference corresponds to the current account. The second part of the statement shows the change in the balance on the account. This measures the amount of new borrowing from the credit card company and corresponds to the capital and financial account. By definition, any unpaid portion of the bill adds one-for-one to the balance. Similarly, a current account deficit generates a capital and financial account surplus of equal magnitude. When a country is spending more than it earns, it is also selling assets to foreigners. The left half of Table 1 summarizes the different components of the U.S.’s $668 billion current account deficit in 2004. From this we see that 1 In the balance of payments accounts, the purchase and sale of assets by central banks, such as the Federal Reserve in the U.S., are often measured separately in the official settlements balance. To simplify the presentation, we have included these transactions in the capital and financial account. In 2004, net purchases by foreign central banks equaled $392 billion, or 59 percent, of the capital and financial account. For more information on the official settlements balance, see the Survey of Current Business, Bureau of Economic Analysis, July 2005. 2 Q1 2007 Business Review TABLE 1 U.S. Balance of Payments, 2004* (Billions of Dollars) Current Account Net Exports Net Income Receipts Net Unilateral Transfers Current Account Balance Capital and Financial Account -617.5 30.4 -80.9 -668 Capital Account Financial Account Statistical Discrepancy Capital and Financial Account Balance -1.6 584.6 85.1 668 *Data are from the Bureau of Economic Analysis’ Balance of Payments Accounts. Details may not add to totals because of rounding. For more details, see the July issue of the Bureau of Economic Analysis’ Survey of Current Business. the trade balance, which is the difference between the value of exports and the value of imports, was the largest determinant of the current account deficit. But there are two additional, smaller components: net unilateral transfers and net income from abroad. Net unilateral transfers measure the value of gifts, foreign aid, and nonmilitary grants. Net foreign income measures the difference of income payments to American capital and workers employed overseas and income payments to foreign capital and workers employed here.2 For the U.S., net 2 A growing and serious concern about measuring the current account is how we treat capital gains and losses on cross-border asset holdings. Economists Pierre-Olivier Gourinchas and Helene Rey construct a measure of the current account with this adjustment and show that current account fluctuations are substantially smaller. In fact, recently, those periods in which the U.S. has run large trade deficits also tended to be those periods in which American asset holdings overseas made large capital gains relative to foreign assets in the U.S. foreign income mostly depends on the difference in capital income — that is, the difference between interest and profit payments to Americans on overseas investments and interest and profit payments to foreigners from investments in the U.S. To finance its current account deficit, the U.S. ran a capital and financial account surplus of $668 billion. Foreign purchases of U.S. assets exceeded U.S. purchases of foreign assets by $668 billion. These foreign purchases of American assets funneled foreign savings toward the U.S. Thus, a current account deficit represents periods when foreign savings are flowing into a country. This brings us to another way of measuring the current account: as the difference between a country’s savings and investment. Savings is the difference between what a country produces, measured as GDP, and what is consumed privately and by the gov- www.philadelphiafed.org ernment.3 When investment exceeds savings, a country finances this gap by borrowing from abroad. Since 1929, the current account and the trade balance have been nearly identical. The average difference is 0.02 percent of GDP. There have been some large differences of up to 1 percent of GDP, but these have generally been short-lived. This may not continue to be the case. If the U.S. continues to run large current account deficits and to borrow from the rest of the world, the stock of foreign assets in the U.S. will grow relative to the stock of U.S. assets overseas. The payments on this debt can lead to deficits in the future, just as a high credit card balance today means more interest payments in the future. For now, though, we will consider the current account and the trade balance interchangeably, partly because, as we have seen, historically they have not differed by much. individuals:4 When these individuals collectively spend more than they earn, they finance the difference by either selling assets or borrowing. However, I might go to my neighbor (indirectly through a bank or credit card) to borrow the amount by which my purchases exceed my income. When a country’s purchases exceed its income, it pays for the difference by borrowing from its trading partners. Thus, a her income is low again, and she lives off the income from her savings. This borrowing and lending over her lifetime reflects intertemporal trade. She has traded part of her income stream when she is working for some payments when she is young and some payments when she is old. This intertemporal trade can involve long periods of borrowing and long periods of saving.5 This borrowing and lending International financial markets allow countries to borrow and lend over time through the purchase and sale of financial assets. INTRODUCING INTERTEMPORAL TRADE Just as an increase in the balance on a credit card bill involves new borrowing from the credit card company, when foreigners buy U.S. assets, Americans are borrowing from the rest of the world. This international borrowing and lending is based on the concept of intertemporal trade. The notion of intertemporal trade is based on the idea that people’s purchases and income may not always match up over time. When this occurs, people use financial markets to borrow and save to make up the difference between what they buy and what they earn. Countries are just a collection of country can have a trade deficit either because it is borrowing or because it has made some loans in the past for which it is currently being repaid. A useful way to think about intertemporal trade is to consider the life cycle of a typical doctor. When she is young, she does not have many skills. Rather than work at a low-wage job, she goes to college and then on to medical school, followed by an internship and residency. Before starting to work, she has little to no income, so she must borrow to pay for school and her living expenses. While in school, she is investing in accumulating skills. These skills raise the wage she can command once she is working. In this case, she borrows when she is young and invests in education. Once out of school, she can repay these loans and start accumulating savings for retirement. Through financial markets she lends her savings to finance other people’s investments. Once she has retired, is efficient, since it allows a person to enter a profession, such as medicine, that makes the best use of her abilities. International financial markets allow countries to borrow and lend over time through the purchase and sale of financial assets. Just as the doctor benefits from intertemporal trade, international financial markets generate similar benefits. Let’s consider two important reasons why countries borrow and lend over time. International Production Shifting. The basis of the idea of international production shifting is the notion that you want to make hay while the sun shines. That is, when good productive opportunities present themselves, people can take advantage of them by investing and working more. Over time, the productive opportunities in a country change. New opportunities present themselves and old ones close. Some industries make technological advances, while others 3 4 5 Strictly speaking, when our doctor borrows to finance her education and expenditures, she is selling a financial asset with a claim against her future income. Lenders carry these assets as a credit on their balance sheets. For those familiar with national income and product accounts, this is the familiar relationship: Trade Balance =Savings-Investment, where Savings = GDP-Private ConsumptionGovernment Consumption. www.philadelphiafed.org Countries are composed of individuals, firms, and governments. However, individuals own firms and governments are made up of people. So, for simplicity, we view countries as a collection of individuals. Business Review Q1 2007 3 become obsolete. Some of these opportunities are small, and others are large. To take advantage of these opportunities, firms need to hire workers and invest in new equipment, structures, and software. Norway provides a clear example of one of these productive opportunities. In the 1960s, rich petroleum deposits were discovered in the North Sea. Norway was one of the major beneficiaries of this discovery. Getting to these valuable oil and gas deposits required large and repeated investments in infrastructure, such as off-shore oil platforms, transport pipelines, ships, and helicopters. Norway also needed to develop a knowledge of exploration and extraction to precisely locate and exploit these resources. At the time of these discoveries, Norway lacked the equipment and expertise to take advantage of the opportunity. To do so, it borrowed from the rest of the world. Because of the time involved in building infrastructure, oil production did not start in earnest until the mid1970s. Although the oil revenue would eventually pay for them, the investments had to be paid for in advance. Norway financed these investments by borrowing from abroad (Figure 1). From the figure, we can see Norwegian investment grew substantially from 1969 to 1977, financed in part by a series of almost continual trade deficits from 1969 to 1977. Once the oil came online, Norway began running persistent trade surpluses, which were used to repay its original borrowing and to save for the day when the petroleum reserves are exhausted. We can see that, since 1978, Norway has annually run trade surpluses that average 6 percent of GDP. There have been some fluctuations in the size of these trade surpluses because of changes in the price of oil and the Norwegian business cycle. (See The Terms of Trade and A Theory 4 Q1 2007 Business Review FIGURE 1 Norwegian Investment and Trade Balance Percent of GDP 40.0 Investment 30.0 20.0 10.0 0.0 Trade Balance -10.0 1968 1973 1978 1983 1988 1993 1998 2003 *Data are from Statistics Norway. of International Business Cycles, for a further discussion of these two forms of trade-balance fluctuations.) The Norway story is an example of a large productive opportunity, but there are also smaller changes in productivity that may be important over the business cycle. For instance, in the 1990s, the information technology and telecommunication sectors in the U.S. developed many new technologies. These productive opportunities affect both the private and public sectors. For instance, in Norway, the state had sovereignty over the exploration and production of sub-sea natural resources, and much of the development was done within state-owned enterprises. To take advantage of productive opportunities, firms and governments need to invest in machines and infrastructure. This can be done by borrowing capital from the rest of the world. Foreign investors are happy to make these loans, even if it means less investment in the investors’ own countries, because the capital is more productive overseas and thus earns a higher return.6 This increase in investment increases the productive capacity of an economy in subsequent periods and keeps the economy going strong into the future. Smoothing Consumption. Another important idea for understanding the dynamics of the current account is consumption smoothing: the notion that people would prefer a relatively stable consumption pattern to a variable one. 6 Some international lending is done by foreign governments. In the case of the U.S., recently these foreign investments have tended to be in relatively low-interest bearing, highly liquid assets. Arguably, the liquidity these investments provide is highly valued by foreign governments and compensates for the relatively low returns. www.philadelphiafed.org The Terms of Trade T here is another important determinant that these variables tend to move in opposite direcof the trade balance: the terms of trade. tions. In particular, notice that the large increases in This is the price of imports relative to the oil prices in 1973 and 1979 were associated with large price of exports. decreases in the trade balance. More recently, the risOver time, the terms of trade may ing price of oil has contributed to the worsening trade vary because the cost of producing imports or exports balance.* changes or the demand for these goods changes. Quite If we return to the case of Norway, which is a often, we see that when the terms of trade worsen, so large exporter of oil, we see that changes in the price that imports become more expensive, the trade balance of oil affect its trade balance in the exact opposite way. declines. This often occurs because, despite the relatively From Figure 1 in the text, we can see that Norway’s high price of imports, we do not cut back much on our trade balance has increased substantially along with purchase of these imports. If we hold quantities roughly the increase in oil prices since 1998. Similarly, the big constant, and the terms of trade increase, the trade baldrop in Norway’s trade balance in 1985 coincided with ance will decrease. This has been an important source of a drop in the price of oil. fluctuations in the trade balance over time. More generally, the terms of trade can matter for Oil is one good that the U.S imports a lot of, and the other goods, such as certain industrial supplies, agridemand for oil is fairly slow to respond to price changes. cultural products, and capital equipment, for which This slow response occurs in part because oil is an imdemand is relatively insensitive to changes in price in portant input into production in industries such as transthe short run. portation and energy and there are few substitutes for oil. These industries have made large investments in air* David Backus and Mario Crucini have shown that the market planes, trucks, and power plants whose energy efficiency for oil can help to explain some of the behavior of the U.S. trade balance in the 1970s and 1980s. is largely fixed. Therefore, just as it is costly for the owner of a gasFIGURE guzzling SUV to sell that car and buy a smaller, more U.S. Oil Trade Balance and Oil Terms of Trade energy-efficient car, it is difOil trade balance as share of GDP ficult for an industry to change Oil terms of trade (percent) its use of oil in the short run. 0.0 600 Thus, an increase in the price Oil Trade Balance of oil tends to raise the value -0.5 500 of imports almost one-for-one and lowers the trade balance -1.0 400 by the same amount in the short run. In the long run, -1.5 300 after firms and individuals -2.0 invest in new, energy-efficient 200 technologies, the demand -2.5 for oil declines, so imports Oil Terms of Trade decline and the trade balance 100 -3.0 increases. The figure bears this out. -3.5 0 1967-I 1970-I 1973-I 1976-I 1979-I 1982-I 1985-I 1988-I 1991-I 1994-I 1997-I 2000-I 2003-I It shows the trade balance in petroleum and the price of petroleum imports deflated by * Data are from the Bureau of Economic Analysis. the price of exports. Notice www.philadelphiafed.org Business Review Q1 2007 5 A Theory of International Business Cycles E conomists David Backus, Patrick Kehoe, and Nobel Prize recipient Finn Kydland have shown that an international realbusiness-cycle model can account for the properties of business cycles in the G7 countries.* This is a model that includes both consumption smoothing and production shifting. In their view, the efficiency with which countries use capital and labor varies over time. These changes in productivity are generally not synchronized across countries, so that productivity may differ internationally. When there are productivity differences across countries, it makes sense to reduce investment in those countries where productivity is relatively low while increasing investment in the country where productivity is relatively high. Initially, this requires the trade balance of the high productivity country to decline. This effectively shifts production to the more productive location. The larger the differences in international productivity, the greater the incentives to shift production toward the more productive countries and the larger the trade deficit. Because investment raises a country’s stock of capital, these capital flows tend to raise future output and lead to sustained increases in output. Foreign investors are happy to make these loans because they can get a better return by lending to firms in the country with more productive opportunities. Notice that by borrowing from abroad, the more productive country does not have to sacrifice consumption to invest in these opportunities, allowing it to keep its consumption smoother. Economist Martin Boileau has shown that the effect of production shifting is particularly important, since a large part of trade consists of capital and durable goods, such as industrial machines, aircraft, and automobiles. Thus, periods when investment is high are also periods in which imports will tend to be high. Moreover, if investment is low in the rest of the world, a country will tend to run a trade deficit. * For a primer on the real-business-cycle view of the macroeconomy, see the article by Satyajit Chatterjee. tion in income to achieve a smooth pattern of consumption. Now imagine we restrict households to borrowing and lending from households in the same country. To smooth out consumption, we need to find someone from the same country willing to lend $25,000 in the first year and be repaid in the second year. Financial markets do this for us. They channel savings from those households with temporarily high incomes to those households with temporarily low incomes. This lets us smooth out the household-specific fluctuations in individual income. But what happens when everyone in the same country experiences the same shock to their income, as in a recession? For instance, suppose average income in a country is $50,000 in year 1 and $100,000 in year 2. If we restrict borrowing and lending with foreign countries, consumption will vary along with income. If we allow international borrowing and lending, consumption smoothing will lead to a $25,000 current account deficit in year 1 and a $25,000 current account surplus in year 2. So countries can use international financial markets to smooth out countrywide fluctuations in income, such as those that occur over the business cycle.7 With these ideas in mind, let’s take a look at the U.S. current account over time. 7 A simple example should make this clear. Suppose you could choose between consuming $50,000 this year and $100,000 the following year or consuming $75,000 each year for the next two years. Most people would prefer the second plan, that is, a smooth pattern of consumption. Now, suppose your income varies, as in the first plan. These types of in- 6 Q1 2007 Business Review come variations tend to occur because some workers receive bonuses and others may temporarily lose their jobs. If households can’t save or borrow, their consumption will follow their income and will vary over time. Suppose one can borrow at a zero interest rate. Then by borrowing $25,000 in the first year and repaying it in the second year, a household can even out this varia- Similarly, fluctuations in government expenditures can be smoothed out by borrowing internationally. When government expenditures exceed tax revenue, the resulting government fiscal deficit is financed by borrowing. Whether this borrowing results in a current account deficit depends on the private savings response of a country’s citizens. It is often claimed that fiscal deficits go hand-in-hand with trade deficits. For the U.S., there are certainly periods with these twin deficits, but there are also periods of government surpluses and trade deficits. See the article by Michele Cavallo for a summary of the links between fiscal and current account deficits. www.philadelphiafed.org A LONG-TERM VIEW OF THE U.S. CURRENT ACCOUNT It’s not possible to describe in detail all the ups and downs of the current account, so let’s focus on some particular periods and events that are important in U.S. history (Figure 2).8 First, let’s consider the secondhalf of the 19th century. In this period, the U.S. was still a relatively small economy that was poised for major economic expansion. The country experienced substantial immigration, and there was a great migration westward. The American railroad network was built, and municipalities invested in infrastructure such as ports, roads, and municipal sewage.9 During this period, the U.S. ran current account deficits each year from 1862 to 1876 and 1882 to 1896. Over these two periods, the average annual current account deficit was 1.5 percent of GDP. Investors in London invested heavily in these enterprises, since the returns to these projects exceeded those to be found in England.10 These trade deficits helped finance the American economic expansion and were followed by a long period of current account surpluses. Second, let’s consider the periods around the two world wars, during which the U.S. ran large and persistent current account surpluses. From 1915 to 1921, the U.S. annually ran current account surpluses, on average, of 4.1 percent of GDP. These loans financed both the war effort of its allies as well 8 The U.S. GDP data from 1860 to 1869 are only an approximation, assuming a 2 percent annual growth rate. 9 From the conclusion of the American Civil War, the American railroad system expanded from 35,021 miles in 1865 to 74,096 miles in 1875 and 128,320 miles in 1885. (Statistical Abstract of the United States: Bicentennial edition, 1975) 10 See the book by Kevin O’Rourke and Jeffrey Williamson, p. 211. www.philadelphiafed.org FIGURE 2 U.S. Current Account Percent of GDP 8.0 6.0 4.0 2.0 0.0 -2.0 -4.0 -6.0 1860 1880 1900 1920 1940 1960 1980 2000 *The U.S. current account is constructed from multiple sources. The period from 1929 is based on data from the Bureau of Economic Analysis. The data from 1869 to 1929 are from the study by Maurice Obstfeld and Matthew Jones. The current account data from 1860 to 1869 are also from Obstfeld and Jones. The U.S. data from 1860 to 1869 are only an approximation. as their subsequent postwar reconstruction. The dynamics of the U.S. current account around World War II are similar to those in the period around World War I. In the buildup to the second world war and before the U.S. entered the war, from 1938 to 1941, the U.S. ran annual current account surpluses of 1.3 percent of GDP. Much of this lending financed the United Kingdom’s war effort. From the U.S. perspective, this was a very good investment. Once the U.S. entered the war, it financed its war effort in part by borrowing from its trading partners. Thus, from 1942 to 1945, the U.S. ran small current account deficits. Following World War II, the U.S. ran some very large trade surpluses from 1946 to 1949. A large amount of both lending and foreign aid was directed toward Europe and Japan to help them rebuild. Given the lack of productive capital in place in these countries and their relatively highly skilled work forces, the goods from the U.S. were effectively used to build up the productive capacity of these countries. These surpluses were very important for rebuilding the European nations and Japan following WWII. Finally, a careful eye may notice that the behavior of the current account since 1980 appears to have a lot in common with the period from 1860 to 1914. In both periods, there are large, sustained swings in the current account. In contrast, in the interwar and postwar periods, fluctuations tended to be small and tended toward balanced trade. These differences across eras are a sign of the uneven progress toward the current world of unrestricted capital flows across borders. International financial flows were much greater in the period before World War I because there were very few restrictions on them. Following WWI, a number of restrictions were Business Review Q1 2007 7 placed on the mobility of international capital, and they were further increased during the Great Depression (1929 to 1939). The postwar financial system maintained these restrictions, which were only gradually loosened in the 1970s. Thus, while today’s current account deficits are quite large, the comparison with the postwar period, when capital flows were partially restricted, exaggerates their magnitude. COMMON FEATURES OF RECENT TRADE DYNAMICS ACROSS COUNTRIES Over long periods of American history, we’ve seen that production shifting and consumption smoothing have mattered for the trade balance. Now, we want to see if the same is true over the business cycle and for other countries. We can do this by studying how the trade balance and other key measures of economic activity vary over time for a group of industrialized countries. First, we can look at some properties of the trade balance, output, consumption, and investment for the G7 countries11 in the period 1980 to 2002 (Table 2).12 From the table we see that certain features of the business cycle 11 The Group of 7 is a coalition of the major industrial nations: Canada, France, Germany, Italy, Japan, the United Kingdom, and the United States. 12 Nobel laureate Robert Lucas has argued that business cycles can be thought of as deviations from a trend around which variables tend to move together. Thus, we want to focus on the medium-term fluctuations in economic activity. These are the fluctuations that last from a year and a half to eight years. We don’t think of very short-run changes in the economic environment, such as those due to really bad weather, as being part of the business cycle. We also don’t think of the really long-term changes in the economy, such as those arising from increased female participation in the labor force, as being part of the business cycle. These are more related to the trend component of the economy. All of the statistics reported in Table 2 are based on these medium-term fluctuations. 8 Q1 2007 Business Review are quite similar across countries.13 From the first two columns, we see that fluctuations in consumption are generally smaller than fluctuations in output, while fluctuations in investment are much larger than fluctuations in output. The second common feature is that both consumption and investment are highly correlated with output. What this means is that when output is growing fast, as in an economic expansion, both investment A CONTRARIAN VIEW OF THE TRADE DEFICIT The view developed here is that the trade balance reflects the optimal response of individuals, firms, investors, and governments to changes in productive opportunities and needs throughout the world. However, an alternative view argues that trade deficits may result from individuals borrowing to spend beyond their means. For instance, individuals may The trade balance reflects the optimal response of individuals, firms, investors, and governments to changes in productive opportunities and needs throughout the world. and consumption are also growing. Since investment is more volatile than output, investment grows much faster than output. From our earlier accounting, this implies that the trade balance should be declining. In fact, from the fifth column we see that trade balances are negatively correlated with output, so that during economic expansions a country’s trade balance tends to decline. If we put these facts together, a common picture of business cycles emerges. When countries are expanding, they tend to be investing quite a bit. Some of the extra production not consumed is invested, but a lot of the resources for investment come from outside the country, so the country runs a trade deficit. Borrowing abroad to increase investment contributes to future increases in GDP without requiring cuts in current consumption. 13 Economists David Backus and Patrick Kehoe find similar properties of the data for a broader group of countries over different periods. not fully take into account the size of their future expenditures, such as those from government-sponsored oldage and medical benefit programs, and not save enough today. Proponents of this “overspending” view argue that closing the current U.S. trade deficit will require some policy actions to increase savings in the U.S. Absent these policy changes, researchers expect that closing the trade deficit may involve some dramatic events. For instance, economists Maurice Obstfeld and Kenneth Rogoff argue that restoring trade balance will require a large depreciation of the U.S. dollar. Similarly, economists Nouriel Roubini and Brad Setser have argued that financing the international debt incurred following these persistent trade deficits will require an increase in interest rates that will discourage investment and economic growth. The properties of the trade balance, evident over the last almost century and a half in the U.S. as well as over the business cycle among industrialized countries, provide ample www.philadelphiafed.org TABLE 2 Business Cycle Statistics* Standard deviation relative to GDP Consumption Investment Correlation with GDP Consumption Investment Trade Balance Canada 0.80 2.84 0.88 0.70 -0.15 France 0.92 3.14 0.74 0.89 -0.43 Germany 0.88 2.32 0.66 0.78 -0.16 Italy 1.32 3.28 0.66 0.76 -0.37 Japan 0.67 2.54 0.64 0.91 -0.48 United Kingdom 1.17 3.34 0.86 0.74 -0.52 United States 0.75 2.75 0.85 0.94 -0.52 Mean - G7 0.93 2.89 0.75 0.82 -0.38 * Consumption, investment, GDP, and trade data are from the OECD's Quarterly National Accounts data set, from 1980:Q1 to 2002:Q2. The Hodrick-Prescott filter was used to remove the long-term trends in each data series. evidence of substantial production shifting and consumption smoothing and cast doubt on this overspending view. SUMMARY The current U.S. trade deficit appears unusually large when compared with that in the postwar period. But in the postwar period, the mobility of capital was fairly limited. In com- www.philadelphiafed.org parison to an earlier era of fairly free mobility of international capital, the current U.S. trade deficits don’t look so unusual. Trade deficits tend to be a sign of good things to come. Countries tend to run trade deficits when they are borrowing to finance productive investment opportunities. This is a way to shift world production toward more productive locations. This inter- national borrowing and lending has played a prominent role in some of the most significant events in U.S. history — from the western expansion after the Civil War to the financing of the two world wars. Over the business cycle, we also see that trade deficits are often associated with strong and continued economic growth and are a sign of good things to come. BR Business Review Q1 2007 9 REFERENCES Backus, David, and Mario Crucini. “Oil Prices and the Terms of Trade,” Journal of International Economics 50 (2000), pp. 185-213. Backus, David, and Patrick Kehoe. “International Evidence on the Historical Properties of Business Cycles,” American Economic Review, 82, 4 (1992), pp. 864-88. Backus, David, Patrick Kehoe, and Finn Kydland. “International Business Cycles: Theory and Evidence,” in T. Cooley (ed.), Frontiers of Business Cycle Research. Princeton: Princeton University Press (1995). Boileau, Martin. “Trade in Capital Goods and the Volatility of Net Exports and the Terms of Trade,” Journal of International Economics (1999), pp. 347-65. Cavallo, Michele. “Understanding the Twin Deficits: New Approaches, New Results,” Federal Reserve Bank of San Francisco Economic Letter, Number 2005-16 (July 22, 2005). 10 Q1 2007 Business Review 2007 Business Review Chatterjee, Satyajit. “Real Business Cycles: A Legacy of Countercyclical Policies?” Federal Reserve Bank of Philadelphia Business Review (January/ February 1999). Gourinchas, Pierre-Olivier, and Helene Rey. “International Financial Adjustment,” National Bureau of Economic Research Working Paper 11155 (2005). Hodrick, Robert J., and Edward C. Prescott. “Postwar U.S. Business Cycles: An Empirical Investigation,” Journal of Money, Credit, and Banking 29:1 (1997), pp. 1-16. Leduc, Sylvain. “Globalization Is Weaker Than You Think,” Federal Reserve Bank of Philadelphia Business Review (Second Quarter 2005). Obstfeld, Maurice, and Kenneth S. Rogoff. “Global Current Account Imbalances and Exchange Rate Adjustments,” Brookings Papers on Economic Activity, 1 (2005), pp. 67146. O’Rourke, Kevin H., and Jeffrey G. Williamson. Globalization and History: The Evolution of a Nineteenth-Century Atlantic Economy. Cambridge, MA: MIT Press (1999). Roubini, Nouriel, and Brad Setser. “The U.S. as a Net Debtor: The Sustainability of the U.S. External Balance,” mimeo, Stern School of Business, NYU (September 2004). Sill, Keith. “The Gains from International Risk-Sharing,” Federal Reserve Bank of Philadelphia Business Review (Third Quarter 2001). Obstfeld, Maurice, and Matthew T. Jones. “Saving, Investment, and Gold: A Reassessment of Historical Current Account Data” in G. Calvo, R. Dornbusch, and M. Obstfeld (eds.), Money, Capital Mobility, and Trade: Essays in Honor of Robert Mundell, Cambridge, MA: MIT Press (2001). www.philadelphiafed.org The Great Moderation in Economic Volatility: A View from the States BY GERALD A. CARLINO S ince the middle of the 1980s, economic growth in the U.S. has become much more stable than it was in the preceding three decades. And the magnitude of the decline is substantial. What accounts for the decline in volatility, and why is the decline important for policymakers? In this article, Jerry Carlino discusses these questions and makes the case that using state-level data, rather than just national data, offers a much larger testing ground for analyzing the decline in economic volatility. Since the middle of the 1980s, growth of the U.S. economy appears to have become much more stable than it was in the preceding three decades. The magnitude of the decline in volatility is substantial: For the nation, growth of output has been one-half and growth of employment two-thirds less volatile than they were in the 1960s and 1970s. An aspect of the change in volatility that has been largely unexplored is its manifestation at the sub-national level. Recently, economists have started to look at the Jerry Carlino is a senior economic advisor and economist in the Research Department of the Philadelphia Fed. This article is available free of charge at www. philadelphiafed.org/econ/br/index.html. www.philadelphiafed.org volatility of employment growth at the state level. Studies have found that while all states shared in the decline, declines were more dramatic in some states than in others. What accounts for the decline in volatility for the nation and its states? The most common explanations for the increased stability and lower volatility of the national economy include structural change in the form of better inventory control practices, improved monetary policy since the late 1970s and early 1980s, and good luck in the form of smaller shocks hitting the economy.1 But when accounting for the various sources of the increased 1 Economists use the term shocks to refer to unanticipated changes in economic variables. Examples include unanticipated changes in monetary and fiscal policy, extreme environmental conditions, and events that alter the world price of energy. economic stability, the national studies pay only modest attention to other types of structural changes that may have helped to lower volatility in general, such as deregulation of the banking industry, increased globalization, fewer unionized workers, and a variety of demographic changes not considered in the national studies. For example, banking deregulation in the 1970s and 1980s may have contributed to lower volatility by allowing consumers and firms to smooth spending over time. Importantly, financial deregulation occurred at about the same time that monetary policy is believed to have improved. The national studies’ failure to take deregulation into account may have led to an overstatement of monetary policy’s role in the great moderation. Why is the decline in volatility important to policymakers? Reduced volatility of employment leads to less economic uncertainty confronting firms and households. Understanding the forces that govern the volatility of employment growth at the subnational level is important to both national and local policymakers, since volatility at the state and national levels are closely related. At the national level, researchers have one observation (the nation) to gain insight into these forces. The advantage of using state data is that such data offer a much larger testing ground for conducting the analysis. TAKING STOCK OF THE GREAT MODERATION Growth of the U.S. economy appears to have become much more stable since the middle of the 1980s Business Review Q1 2007 11 relative to the preceding three decades. A graph of the growth rate of employment in the U.S. depicts this increased stability. From the mid-1950s to the early 1980s, quarterly employment growth largely fluctuated in a range of around 2.0 percent to -1.5 percent. Since the mid-1980s, however, employment growth has hovered in a much narrower range: from less than 1 percent to about -0.5 percent (Figure 1). The volatility of employment growth can also be measured using the standard deviation, which shows how much employment growth moves up and down around its average value.2 By this measure, average volatility of U.S. employment growth fell from a bit under 1.0 percent during the early 1960s to about 0.3 percent in 2005 (Figure 2). More specifically, volatility fell precipitously during the 1960s: from a high of 0.96 during the second quarter of 1962 to 0.31 during the fourth quarter of 1969. Beginning in the 1970s, employment growth volatility reversed its previously declining trend and nearly tripled. This rise in volatility coincides with the generally poor economic conditions of the 1970s, during which time the economy experienced rising inflation and slow growth. From the early 1980s on, however, volatility generally declined as economic performance improved relative to the 1970s. This is an important period in that most studies have tried to account for increased economic stability since the early 1980s. Despite the general decline since the mid-1980s, volatility temporarily increased during the 1990-91 recession and the 2001 recession. Volatility fell FIGURE 1 Quarterly Employment Growth Percent 2.5 2 1.5 1 0.5 0 -0.5 -1 -1.5 -2 1952 1957 1963 1969 1975 1980 1986 1992 1998 2003 FIGURE 2 Standard Deviation of Total Employment Growth Volatility Percentage Points 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 2 The data used in this article are quarterly from 1961:1-2005:2. The data were seasonally adjusted before computing our volatility measure. Volatility is measured as the standard deviation of employment growth over the previous 20 quarters. 12 Q1 2007 Business Review 0.1 0 1961 1966 1972 1978 1984 1989 1995 2001 www.philadelphiafed.org dramatically during the expansion in the 1990s, as it has during the current expansion. There is a debate among economists about whether the decline in volatility is best represented as a sudden one-time “break” around 1984 as opposed to a more moderate longrun decline in volatility over several decades. Casual inspection of Figure 2 suggests that employment growth volatility fell sharply in the mid-1980s. The figure shows that volatility of employment growth fell from an average of around 0.7 percent in the mid-1980s to an average of about 0.3 percent in 2005. Using a variety of statistical methods, economists find evidence that a one-time drop, or break, in volatility seems to have occurred around 1984. Following this convention, we will look at the change in employment growth volatility at the state level between two periods: 1956 to 1983 and 1984 to 2002.3 Table 1 shows that while all states shared in the decline, employment growth volatility declined much more dramatically in some states than in others. The state with the largest post3 See the article by Keith Sill for a discussion of the two views: a one-time break in volatility vs. a long-run gradual decline. While the various measures used to analyze volatility have differed from study to study, all studies find that volatility has declined since the mid-1980s. In this article we assume that 1984 represents the break date for each state, too. Michael Owyang, Jeremy Piger, and Howard Wall report finding differences in both the break date and the magnitude of the reduction in volatility across individual states. However, they did not examine whether the break date they found for any given state is significantly different from the break dates found for other states. TABLE 1 Change in Employment Growth Volatility by State* State West Virginia Michigan Ohio Indiana Pennsylvania Alabama Kentucky Wisconsin Arkansas North Dakota Washington Minnesota Oregon Kansas Idaho Iowa Tennessee United States Montana Florida Illinois Nevada Utah Mississippi Arizona Percent Decrease in Employment Growth Volatility: 1956-1983 to 1984-2002 75.9 63.6 57.8 57.1 56.9 53.8 53.7 52.5 52.1 51.9 50.2 47.4 47.3 46.0 46.0 45.3 44.6 43.9 43.2 42.9 42.7 42.2 41.3 40.7 39.3 State New Mexico Delaware Maryland Missouri South Dakota North Carolina South Carolina Louisiana California Wyoming Colorado Nebraska Massachusetts Rhode Island Vermont Connecticut Georgia Oklahoma Texas Virginia Maine New Jersey New Hampshire New York Percent Decrease in Employment Growth Volatility: 1956-1983 to 1984-2002 37.9 37.6 37.1 36.6 35.9 33.4 29.6 29.5 28.5 27.7 25.7 25.5 25.0 24.6 24.5 24.4 23.3 23.2 20.7 16.9 16.3 15.3 10.2 8.8 * Excluding Alaska and Hawaii www.philadelphiafed.org Business Review Q1 2007 13 war decline in employment growth volatility is West Virginia, which saw a drop of almost 76 percent. The state with the smallest decline is New York, at about 9 percent, compared with a decline of about 44 percent nationally. Looking at the three states in the Third Federal Reserve District, we find that Pennsylvania was among the top five states in terms of the decline in the state’s employment growth volatility, falling almost 60 percent. The decline in employment growth volatility in New Jersey (about 15.3 percent) was well below the national average; in Delaware (about 38 percent), it was somewhat below the national average. In general, similar declines in the volatility of total employment growth occurred at about the same time in most major sectors. Figure 3 shows employment growth volatility by sector for the nation for our two periods: 1956 to 1983 and 1984 to 2002.4 With the exception of the finance, insurance, and real estate (FIRE) sector, the figure shows more stable employment growth by sector in the later period than in the earlier one. Table 2 shows the decline in volatility by state for two important sectors: manufacturing and services.5 The 4 Because of recent changes in the way industries are assigned to broad sectors, we do not have a consistent series by sector that extends back sufficiently through time. Thus, our analysis at the sectoral level ends in 2002. Since industrial reclassification did not affect aggregate employment, the analysis using aggregate data extends through 2005. 5 Services include personal services, business services, educational services, and social and other services. Services provided by finance, insurance, and real estate industries are included in the FIRE sector. While manufacturing’s share of national total employment has gone down over time, services’ share has increased. Taken together, manufacturing and services have accounted for roughly 40 percent of total national nonfarm employment since the 1950s. The remaining 60 percent of national nonfarm employment is accounted for by trade; government; transportation, communication, and public utilities; mining; and construction. 14 Q1 2007 Business Review FIGURE 3 Standard Deviation of Employment Growth Volatility by Industry Percent 5 4.5 4 3.5 3 2.5 2 1.5 1 0.5 0 Mining Construction Manufacturing Transportation & Utilities 1956-1983 state with the largest postwar decline in manufacturing employment growth volatility is Michigan, which experienced a 66 percent drop, while South Dakota saw a 17 percent decline, the smallest among all states. Manufacturing employment growth volatility fell 56 percent in the nation. The change in employment growth volatility for services is not given for some states because of insufficient data in the earlier period. Since the table shows the decline in volatility between the earlier and the later period, a negative number for a state indicates that employment growth volatility increased in that state. While employment growth volatility in the services sector decreased a modest 2.5 percent for the nation between the earlier and later period, there was substantial variation across states. The state with the largest decline in employment growth volatility in services Trade FIRE Services Government 1984-2002 was Kentucky, which experienced a drop of 64 percent. On the other hand, in Mississippi, employment growth volatility in services increased almost 29 percent. Still, the vast majority of states for which data are available experienced declining volatility in their service sectors, as well as in other broad sectors. SEEKING SOURCES OF THE GREAT MODERATION Economists have offered a number of possible explanations for the decline in the U.S. economy’s volatility. These can be grouped under three broad headings: better policy, good luck, and structural change. Better Policy. Economists have noted that improved monetary policy — the greater emphasis the Fed placed on controlling inflation in the VolckerGreenspan years — might have dampened the effects of economic fluctua- www.philadelphiafed.org TABLE 2 Percent Decrease in Volatility: 1956-1983 to 1984-2002* State United States Alabama Arizona Arkansas California Colorado Connecticut Delaware Florida Georgia Idaho Illinois Indiana Iowa Kansas Kentucky Louisiana Maine Maryland Massachusetts Michigan Minnesota Mississippi Missouri Montana Nebraska Nevada New Hampshire New Jersey New Mexico New York North Carolina North Dakota Ohio Oklahoma Oregon Pennsylvania Rhode Island South Carolina South Dakota Tennessee Texas Utah Vermont Virginia Washington Wisconsin West Virginia Wyoming Manufacturing Employment 55.7 56.5 45.0 56.6 44.3 46.3 50.9 23.2 48.4 46.2 56.8 55.7 63.1 48.8 63.0 56.9 36.5 37.8 55.6 36.0 66.3 45.5 42.7 45.5 54.5 36.7 52.7 32.8 45.9 35.1 50.2 43.0 53.8 64.3 33.0 46.9 63.1 48.9 43.4 17.4 50.9 39.1 45.4 54.1 43.5 32.7 56.9 56.0 39.1 Services Employment 2.5 5.4 10.3 24.6 20.1 14.8 1.3 N/A 36.6 24.2 39.5 -3.0 40.6 31.1 24.2 63.8 25.2 N/A N/A N/A N/A 18.6 -28.5 -5.8 25.5 25.0 47.4 32.6 -5.0 39.6 -2.7 18.7 6.2 17.6 18.5 22.2 20.5 N/A 18.9 30.1 11.8 4.3 N/A 31.8 -12.9 26.3 41.2 39.2 62.3 tions, leading to a more stable economy. According to Olivier Blanchard and John Simon, the volatility of output and the volatility of inflation have tended to display a strong positive correlation. Low and stable inflation makes economic planning easier and improves the functioning of markets. Stable inflation may contribute to more stability in the growth of output and employment. In the pre-Volcker era, monetary policy was characterized as “accommodative” in that policymakers did not respond strongly enough to keep inflationary pressures under control. The conduct of monetary policy appears to have changed significantly beginning in the Volcker era in an effort to bring high and rising inflation pressures under control.6 A recent study by James Stock and Mark Watson shows that the increased stability of output and employment since the mid-1980s is partly due to monetary policymakers’ greater emphasis on inflation and their success at controlling it. In studying the various sources of the moderation in output volatility since the mid1980s, Stock and Watson find that better monetary policy since the early 1980s accounts for about 20 percent of the decline in volatility.7 Still, Stock and Watson find that half the decline in volatility is unaccounted for and they attributed it to sheer luck. Good Luck. The word “shock” represents economists’ shorthand for a factor or force that causes an unex- 6 Paul Volcker served as Chairman of the Federal Reserve from 1979 to 1987. Alan Greenspan, who succeeded Volcker, served as Chairman from 1987 to 2006. 7 Sylvain Leduc and Keith Sill also assessed the importance of monetary policy for the decline in U.S. output volatility that has occurred since the mid-1980s. They find that improved monetary policy accounted for about 10 percent of the decline in real output volatility, half the size found by Stock and Watson. * Excluding Alaska and Hawaii www.philadelphiafed.org Business Review Q1 2007 15 pected change in an economic variable, such as employment growth. Examples include weather-related events, strikes, and domestic and foreign political crises. To the extent that volatility is the result of large adverse shocks, it will decline if these unlucky events are smaller in magnitude or happen less frequently. Hurricanes and other weather-related events represent a type of shock that affects states differently. The damage done to Louisiana, Mississippi, and Alabama by Hurricane Katrina is an obvious case. As we have indicated, a substantial part of the decline in national volatility cannot be accounted for and may be due merely to good luck. Unfortunately, the good luck the economy has experienced since the mid-1980s may be temporary. If the bad luck the economy experienced prior to the mid-1980s returns, economic volatility may increase. Structural Changes. Many types of structural changes may have helped to lower the volatility of employment, such as the shift of jobs from manufacturing to services, better inventory management methods, fewer unionized workers, and banking deregulation. Redistribution of jobs. Perhaps the most intuitive explanation for the decline in employment growth volatility in the mid-1980s involves the shift of employment from the relatively more volatile goods-producing sector to the relatively less volatile services sector. According to this view, manufacturing contributed more to the decline in volatility than other sectors, both because manufacturing is a relatively high-volatility sector and because manufacturing’s share of employment has declined.8 Manufacturing’s share of total U.S. employment fell from an average of 27 percent between 1956 and 1983 to an average of just over 16 percent between 1984 and 2002. At the same time, services’ share increased from 16 Q1 2007 Business Review FIGURE 4 Ratio of Manufacturing to Services Employment Volatility Ratio 1.6 1.4 1.2 1 0.8 0.6 0.4 0.2 0 1961 1964 1967 1970 1973 1976 1979 1982 1985 1988 1991 1994 1997 2000 an average of about 17 percent in the earlier period to an average of 27 percent in the later period. In the earlier period, manufacturing employment growth was, on average, almost twice as volatile as employment growth in services. While manufacturing continues to be more volatile than services, the gap has narrowed substantially. In the later period, manufacturing employment growth was, on average, only about 50 percent more volatile than was employment growth in services. Figure 4 shows the volatility of manufacturing employment growth relative 8 Although volatility in the mining and construction sectors fell more than volatility in the manufacturing sector, the share of employment in both the mining and the construction sectors accounts for at most about 6 percent of total U.S. employment. Given their relatively small share of total employment, these two sectors contribute very little to the decline in total employment volatility, despite the relatively large declines in volatility recorded by industries in these sectors. to the volatility of services employment growth. After declining for the better part of the 1960s, relative volatility increased somewhat between the late 1960s and early 1970s, before increasing dramatically in the period 1973 to 1979. The jump in relative volatility is largely due to a jump in volatility in manufacturing. The disruption in oil supplies in the 1970s may have led to much greater volatility in manufacturing than in services. Of importance for this article is the sharp drop in relative volatility since the mid-1980s, which is consistent with the observed sharp drop in the volatility of total employment growth.9 How much does the shift of jobs from the relatively high-volatility sec9 A couple of studies have found that energy price shocks since the mid-1980s have played virtually no role in accounting for the increased stability of the national economy. See the article by Stock and Watson and the one by Leduc and Sill. www.philadelphiafed.org tors to the relatively low-volatility sectors matter in explaining the overall decline in employment growth volatility? To address this issue, we conducted an experiment in which we constructed a hypothetical series for total employment growth, holding each industry’s share of total employment fixed at its 1961 level. Since industry shares are held constant at their 1961 levels over the period 1961 to 2002, all of the variability in the hypothetical series will be due to changing volatility in the various sectors.10 Figure 5 shows hypothetical volatility juxtaposed with actual volatility. The volatility of the hypothetical series is generally above that of the actual series. Still, the largest difference between the hypothetical series and the actual series is in the 1970s and early 1980s. The difference between these two series was much narrower after 1984, suggesting that the shift of jobs away from manufacturing is not an important cause of the decline in volatility since the mid-1980s. In fact, two recent studies using state-level data on volatility and a statistical technique called regression analysis find that the redistribution of jobs toward the less volatile sectors has played only a minor role in accounting for the decline in employment volatility observed since the mid-1980s.11 Better inventory management. Some studies point to innovation in inventory management techniques (such as the explosion in information 10 To construct the hypothetical series, employment growth rates for each major industry for each year were weighted by each industry’s 1961 share of total employment. The hypothetical employment growth series was used to compute the hypothetical volatility (defined as the standard deviation of employment growth over the previous 20 quarters) shown in Figure 5. 11 See my study with Robert DeFina and Keith Sill and the study by Owyang, Piger, and Wall for evidence on the role of the shift of jobs from manufacturing to services in explaining changing employment growth volatility. www.philadelphiafed.org FIGURE 5 Hypothetical and Actual Volatility in Total Employment Growth Standard Deviation 1.2 1 Hypothetical 0.8 0.6 0.4 Actual Employment Shares 0.2 0 1961 1964 1968 1972 1976 1979 technology and just-in-time production techniques) that have allowed firms to better use inventories to smooth production and employment. For example, just-in-time inventory techniques allow producers to maintain lower stock levels and to better match production with sales. Changes in demand result in smaller swings (that is, less volatility) in production now than in past decades. Despite this theory’s appeal, the extent to which improved inventory management methods have contributed to increased stability is subject to some debate by economists and the question of its role is far from settled.12 Fewer unionized workers. Other 12 See the article by James Kahn, Margaret McConnell, and Gabriel Perez-Quiros for a discussion of the role of improved inventory management in the decline in economic volatility. The bulk of the research suggests little role for inventories in reducing volatility. See the article by Keith Sill for a good review of the relevant studies. 1983 1987 1991 1994 1998 2002 types of structural changes, such as fewer unionized workers and increased globalization, may also have been at work, and these changes may have helped to lower volatility. For example, an important structural change is the sharp drop in the number of union members over the past 40 years. In 1964, almost 30 percent of workers were union members. By 1994, the share had fallen to less than 13 percent. In fact, the decline in the share of unionized workers accelerated after 1980. Between 1964 and 1980, the share of workers covered by unions fell about 1.3 percent per year, but the share fell about 1.8 percent per year between 1980 and 2004. The acceleration in the decline of unionized workers after 1980 may have contributed to the economy’s increased stability. Why might employment volatility decrease as the number of unionized workers decreases? Since unions are generally unwilling to accept decreases Business Review Q1 2007 17 in work hours and wages, when demand falls, unionized firms can adjust only by changing employment. When demand improves, unionized firms rehire many of the same workers they laid off during bad times. These layoffs and subsequent rehires may induce greater volatility in employment growth than would have occurred if wages had borne more of the adjustment to changing demand. But the decline in the share of unionized workers occurred gradually, making it an unlikely explanation for the sharp drop in volatility we observe. Banking deregulation. An important type of structural change that began in the early 1980s was the deregulation of the banking industry. Until the 1980s, commercial banks in the U.S. faced restrictions on the interest rates they could pay depositors and charge borrowers. When market interest rates rose above the legal ceilings that banks were allowed to pay for deposits, many depositors withdrew their funds from the banking system. This led to a drop in the amount of credit that banks could extend to firms and households, thereby hurting bankdependent borrowers. Housing in the 1960s and 1970s was particularly hard hit when market interest rates rose above these interest rate ceilings. But once the ceilings were removed in the 1980s, banks and savings and loans were able to offer competitive interest rates to their depositors, thus preventing a wholesale withdrawal of deposits and allowing banks to continue to make construction and mortgage loans. In fact, economists Karen Dynan, Douglas Elmendorf, and Daniel Sichel show that there has been a substantial decline in the volatility of residential investment since the mid-1980s. Until the 1980s, banks also faced geographic limitations in that bank holding companies were not permitted 18 Q1 2007 Business Review to cross state borders. The geographic restrictions also made banks’ ability to lend more vulnerable to economic shocks that affected their own states. In the absence of a national banking system integrated across states, the allocation of funds and the resulting distribution of money and credit can be uneven. That is, it can get "stuck" in a state, depending on where and how the deposit and withdrawal activity takes place. In this case, money and credit would flow less easily from one state to another in the face of a state shock. Although banking markets tended Strahan finds that state employment volatility fell substantially after interstate banking was permitted.13 States deregulated their banking sectors at different times. In 1978, Maine was the first state to pass a law that allowed entry by bank holding companies from any state that reciprocated by allowing Maine banks to enter their banking markets. Following Maine’s lead, states deregulated in waves, with the bulk of states approving legislation to allow deregulation between 1985 and 1988. With the exception of Hawaii, all states allowed interstate banking Until the 1980s, banks also faced geographic limitations in that bank holding companies were not permitted to cross state borders. to be local in nature prior to deregulation, a bank in one state that needs money could borrow in national credit markets, such as the fed funds market (borrowing of funds overnight from other banks), through bank holding companies that issue commercial paper to raise funds, and the Eurodollar market (deposits from banks outside the U.S.). However, raising funds in these national and international markets imposed some additional costs on banks, and these costs may have limited banks’ willingness to raise funds from these sources. Today, most of these restrictions on commercial banks have been phased out. Shocks to a state can be met with inflows or outflows of funds, and thus, the adjustment to the shock is likely to be smoother. In essence, deregulation made the banking system more efficient and, in the process, allowed the financial sector to act more as a stabilizer for the real sector. A recent study by Donald Morgan, Bertrand Rime, and Philip by 1993. In their study, Morgan, Rime, and Strahan use the staggered timing in state-level action to relax interstate banking restrictions to explain some of the cross-state differences in employment growth volatility as well as the increased stability of state economies. They conclude that the increased stability following regulatory change made state economies much less sensitive to the fortunes of their own banks. The finding that interstate banking appears to have contributed to increased economic stability raises an 13 According to the theory developed in Morgan, Rime, and Strahan’s study, it’s possible for volatility to rise, fall, or remain mostly unchanged following legislation that allowed interstate banking. Deregulation’s net effect on employment growth volatility is, therefore, an empirical issue. As indicated in the study by Morgan and co-authors, the net effect is, on balance, negative, suggesting that employment volatility became more stable after interstate banking was allowed than before such deregulation. For a discussion of why deregulation’s net effect on employment growth volatility might be positive, see the article by Philip Strahan. www.philadelphiafed.org important concern with Stock and Watson’s study, which attributed 20 percent of reduced volatility since the mid-1980s to improved monetary policy. Since financial deregulation occurred at roughly the same time that monetary policy is supposed to have improved, it’s possible that the Fed did not make as substantial a contribution to increased stability as some believe; rather, banks were better able to implement monetary policy decisions following deregulation. By not controlling for financial deregulation, Stock and Watson may have overstated monetary policy’s role in lowering volatility. Similarly, while Morgan and co-authors considered banking deregulation’s contribution to volatility, they did not adequately control for the role that improved monetary policy may have played. In their study, Morgan and co-authors account for the common or average effect of monetary policy on state volatility. In an earlier Business Review article, Robert DeFina and I found that monetary policy affects economic activity in the states quite differently. It’s conceivable that changes in the conduct of www.philadelphiafed.org monetary policy may have contributed to substantial state-level deviations in the growth of employment volatility from the average effect measured by Morgan and co-authors. If the unaccounted-for differences in the impact of monetary policy are correlated with the date at which states deregulated, Morgan and co-authors’ estimates of deregulation’s effect on the volatility of employment growth may be overstated. We believe there is evidence of such bias.14 To date, no study has accounted adequately for both forces — improved monetary policy and deregulation — simultaneously. 14 States that tend to be more sensitive to monetary policy actions might have deregulated earlier than states that are less sensitive to policy actions in an attempt to smooth employment volatility in the more responsive states. This would impart a negative correlation between the differential state responses to monetary policy action and the timing of banking deregulation. Using estimates of the differential state responses to monetary policy action reported in my paper with Robert DeFina, I found a negative (-0.278) and significant correlation between differential state responses to monetary policy action and the timing of banking deregulation. CONCLUSION The question of what generates volatility in employment growth at the state level is closely related to what generates volatility at the national level. Understanding the forces that govern employment growth volatility at the sub-national level is important to both national and local policymakers. While progress has been made in identifying some of the sources of the great moderation, there appear to be other forces at work that could improve our understanding of the increased stability of local and national economies. While some studies have looked at the relative roles that the shift of jobs to services, better inventory management, better monetary policy, and financial deregulation have played in producing a more stable economy, no study has satisfactorily controlled for all of these forces simultaneously. Accounting for all of these forces together is an important next step to understanding the relative contributions these various forces may have individually played in explaining the great moderation. BR Business Review Q1 2007 19 REFERENCES Bernanke, Ben S. “The Great Moderation,” speech at the Eastern Economic Association, Washington (February 20, 2004). www.federalreserve.gov/boarddocs/ speeches/2004/20040220/default. htm#f1 Dynan, Karen, Douglas Elmendorf, and Daniel Sichel. “Can Financial Innovation Help to Explain the Reduced Volatility of Economic Activity?,” Journal of Monetary Economics, 53, 1 (January 2006), pp. 123-50. Blanchard, Olivier, and John Simon. “The Long and Large Decline in U.S. Output Volatility,” Brookings Papers on Economic Activity 1:2001 (2001), pp. 135-64. Kahn, James, Margaret McConnell, and Gabriel Perez-Quiros. “On the Causes of the Increased Stability of the U.S. Economy,” Federal Reserve Bank of New York Economic Policy Review (May 2002), pp. 183-202. Carlino, Gerald, and Robert DeFina. “Do States Respond Differently to Changes in Monetary Policy?” Federal Reserve Bank of Philadelphia Business Review (March/April 1999). Carlino, Gerald, Robert DeFina, and Keith Sill. “Postwar Period Changes in Employment Volatility: New Evidence from State/Industry Panel Data,” Working Paper 03-18, Federal Reserve Bank of Philadelphia (2003). 20 Q1 2007 Business Review Leduc, Sylvain, and Keith Sill. “Monetary Policy, Oil Shocks, and TFP: Accounting for the Decline in U.S. Volatility,” Working Paper 03-22/ R, Federal Reserve Bank of Philadelphia (2003). Morgan, Donald, Bertrand Rime, and Philip Strahan. “Bank Integration and State Business Cycles,” Quarterly Journal of Economics (November 2004), pp. 1555-84. Owyang, Michael, Jeremy Piger, and Howard Wall. “A State-Level Analysis of the Great Moderation,” unpublished manuscript, Federal Reserve Bank of St. Louis (September 14, 2005). Sill, Keith. “What Accounts for the Postwar Decline in Economic Volatility?” Federal Reserve Bank of Philadelphia Business Review (First Quarter 2004). Stock, James, and Mark Watson. “Has the Business Cycle Changed and Why?” NBER Working Paper 9127 (August 2002). Strahan, Philip E. “The Real Effects of U.S. Banking Deregulation,” Federal Reserve Bank of St Louis Review (July/ August 2003). “The Unfinished Recession: A Survey of the Word Economy,” The Economist, September 28, 2002. www.philadelphiafed.org The Macroeconomics of Oil Shocks BY KEITH SILL F or various reasons, oil-price increases may lead to significant slowdowns in economic growth. Five of the last seven U.S. recessions were preceded by significant increases in the price of oil. In this article, Keith Sill examines the effect of changes in oil prices on U.S. economic activity, focusing on how runups in the price of oil can affect output growth and inflation. He also discusses the channels by which oil-price increases might affect the economy and the historical evidence on the relationship between oil prices, economic growth, and inflation. During the first quarter of 2002, the price of crude oil averaged $19.67 per barrel. Four years later, in the first quarter of 2006, the average price of oil had risen to $63 per barrel. Indeed, the high price of oil may not be a short-lived phenomenon: Futures markets indicate that investors expect the price of oil to remain above $70 per barrel through 2008. For the postwar U.S. economy, the data show a clear tendency for oil-price spikes to precede Keith Sill is a senior economist in the Research Department of the Philadelphia Fed. This article is available free of charge at: www. philadelphiafed.org/ econ/br/index.html. www.philadelphiafed.org economic downturns. Though most of these episodes occurred at a time when oil’s share as an input into U.S. production was larger than it is today, there is still much debate about how oil prices affect the economy. How concerned should we be about the economic consequences of persistently high oil prices? Oil prices matter for the economy in several ways. Changes in oil prices directly affect transportation costs, heating bills, and the prices of goods made with petroleum products. Oilprice spikes induce greater uncertainty about the future, which may lead to firms’ and households’ delaying purchases and investments. Changes in oil prices also lead to reallocations of labor and capital between energyintensive sectors of the economy and those that are not energy-intensive. For these reasons and others, oil- price increases may lead to significant slowdowns in economic growth. In the postwar U.S. data, the correlation between oil-price spikes and economic downturns is not perfect — some oil-price increases are not followed by recessions. But five of the last seven U.S. recessions were preceded by significant increases in the price of oil. The most recent rise in the price of oil has not led (at least not yet) to an economic recession, but history nonetheless suggests that oil prices are an important element in assessing the economy’s near-term prospects. OIL PRICES From the late 1940s to the early 1970s, the price of oil was very stable, moving up only slightly.1 From the early 1970s to the early 1980s, the price of oil rose dramatically in a sequence of steps associated with the rise of OPEC and disruptions in the supply of oil from the Middle East oil-producing countries (Figure 1).2 1 From 1948 to 1972, the price of oil produced in the U.S. was influenced by the production quotas set by the Texas Railroad Commission (TRC). Each month, the TRC (and other state regulatory agencies like it) made forecasts of petroleum demand for the upcoming month and set production quotas to meet the forecasted demand. Since the quantity of oil produced was adjusted to meet forecasted demand, the price of oil remained fairly stable. However, in the face of growing world demand for oil relative to supply, and the peaking of U.S. domestic oil production in 1970, the TRC set the production quotas at 100 percent in March 1971. 2 The Organization of Petroleum Exporting Countries (OPEC) was formed in 1960 with five founding members: Iran, Iraq, Kuwait, Saudi Arabia, and Venezuela. By the end of 1971, Qatar, Indonesia, Libya, the United Arab Emirates, Algeria, and Nigeria had joined. Business Review Q1 2007 21 FIGURE 1 Nominal Price of Crude Oil Dollars / Barrel 70 60 50 40 30 20 10 Source: Jan-04 Jan-02 Jan-98 Jan-00 Jan-96 Jan-94 Jan-90 Jan-92 Jan-88 Jan-84 Jan-86 Jan-76 Jan-78 Jan-80 Jan-82 Jan-74 Jan-70 Jan-72 Jan-68 Jan-62 Jan-64 Jan-66 Jan-58 Jan-60 Jan-54 Jan-56 Jan-52 Jan-50 0 Haver Analytics. Price of West Texas Intermediate. OPEC first experienced the power it had over the price of oil during the Yom Kippur War, which started in October 1973. As a result of U.S. and European support of Israel, OPEC imposed an oil embargo on western countries. Oil production was cut by 5 million barrels a day (though about 1 million barrels a day in production was made up by other countries). The cutback amounted to about 7 percent of world production, and the price of oil increased 400 percent in six months. From 1974 to 1978 crude-oil prices were relatively stable, ranging from $12 to $14 per barrel. The next big round of oil-price increases came with the Iranian revolution and Iran-Iraq war in 1979 and 1980. World production fell 10 percent, and this resulted in the price of oil rising from $14 to $35 per barrel. However, the high price of oil was leading consumers and firms to conserve energy. Homeowners insu- 22 Q1 2007 Business Review lated their houses. Commuters bought more fuel-efficient cars. Firms bought equipment that was more energy efficient. High oil prices also led to increased exploration and production of oil from countries outside of OPEC. From 1982 to 1985 OPEC sought to stabilize the price of oil through production quotas, but conservation efforts, a global recession, and cheating on production quotas by OPEC members eventually led to a plummeting of oil prices to below $10 per barrel by 1986.3 Since the mid-1980s the frequency of oil-price changes has been much greater than in the past. OPEC continued to influence the price using pro3 Over the period 1982-86, Saudi Arabia acted as the marginal oil producer, cutting its production in an effort to keep oil prices from falling. In August 1982, the Saudis abandoned that strategy and linked their oil price to the spot market for crude. duction quotas, but it has been unable to stabilize it. In fact, OPEC’s share of world oil production has fallen from a peak of 55 percent in 1973 to about 42 percent today. U.S. imports of oil from OPEC, as a share of total petroleum imports, peaked at 70.3 percent in 1977 and have since fallen to about 43 percent.4 Today, the major suppliers of imported oil for the U.S. are Canada and Mexico, followed by Saudi Arabia and Venezuela. Oil Prices, Recessions, and Inflation. We can plot the real price of oil, that is, the price of oil adjusted for inflation, the rate of inflation as measured by the consumer price index (CPI), and U.S. recessions as defined by the National Bureau of Economic Research (Figure 2). The figure indicates that even with the substantial runup in the nominal price of oil since 1999, oil remains cheaper, in real terms, than it was during the late 1970s. A striking aspect of the postwar history of oil prices and the economy is that oil prices spike upward right around the time of recessions. A clear relationship between oil prices and inflation is harder to discern. During some episodes, such as 1973-74 and 1980, it appears that inflation rises at the same time that the price of oil rises. At other times, such as 2002 to the present, oil prices rise while inflation remains stable. The figure suggests that the relationship between oil-price increases and the real economy might be stronger than the relationship between oil prices and inflation. A characteristic of the oil-price spikes that occur around recessions is that they tend to be both large and 4 U.S. oil production peaked at 9.6 million barrels a day in 1970 and has since fallen to about 5.4 million barrels a day. Even at that rate, the U.S. remains one of the world’s largest oil producers. In fact, only Saudi Arabia and Russia produce more oil in a year than the U.S. www.philadelphiafed.org WHY MIGHT OIL-PRICE SPIKES CAUSE RECESSIONS? Is it plausible that an increase in the price of oil leads to recession when oil represents such a small (and declining) share of U.S. output? Oil consumption as a share of gross domestic output (GDP) was slightly below 4.5 percent in the early 1970s, but it has since declined steadily to a little over 2 percent in 2004 (Figure 3).5 How could a change in the price of an input that represents such a small share of the 5 Oil consumption is measured using the Energy Information Administration’s data on U.S. total crude oil and petroleum products supplied to U.S. households, firms, and government. www.philadelphiafed.org FIGURE 2 Real Oil Price & CPI Inflation Dollars 60 50 40 Real Oil Price 30 20 10 0 CPI Inflation Source: Jan-04 Jan-02 Jan-98 Jan-00 Jan-96 Jan-94 Jan-90 Jan-92 Jan-88 Jan-84 Jan-86 Jan-76 Jan-78 Jan-80 Jan-82 Jan-74 Jan-70 Jan-72 Jan-68 Jan-62 Jan-64 Jan-66 Jan-58 Jan-60 Jan-54 Jan-56 Jan-52 -10 Jan-50 abrupt. By abrupt, we mean that the price changes are sharp upward movements rather than slow and gradual upward drifts. Historically, the prerecession spikes are associated with disruptions in supply from the Middle East. These supply disruptions tend to be associated with wars that led to significant reductions in the amount of oil exports by the affected countries (see the table). The table shows that Middle East conflicts led to rather large reductions in the world supply of oil. Absent a large drop in demand, such large supply disruptions could lead to large increases in the world price of oil. The table confirms that U.S. business-cycle peaks also tended to occur close in time to the dates of the conflicts. Note, though, that the length of time between the oil-supply disruption and the business-cycle peak varies, ranging from about contemporaneously to a little over one year. What the table and Figure 2 by themselves cannot tell us is whether oilprice increases or Middle East conflicts or some other factor, such as monetary policy, led to recessions in 1957, 1973, 1980-81, and 1990. However, the data suggest the possibility of a link between oil and the macroeconomy. Haver Analytics. Real oil price is West Texas Intermediate price divided by CPI inflation. economy have such a dramatic economic effect? Oil prices affect the economy through a multitude of channels. When all of these effects are added up, oil prices could have a larger impact than what might be expected from oil’s small share in the economy. The key is that oil-price changes affect both supply and demand. Changes in oil prices affect supply because they make it more costly for firms to produce goods; they affect demand because they influence wealth and can induce uncertainty about the future. First, let’s consider this: An oilprice increase acts like a tax on firms and households. The United States imports a large fraction of the oil it uses from other countries, and the payments we make to foreign countries for oil represent an outflow of funds from the U.S. Higher payments to foreigners for oil reduce income available for spending on other goods and services. The lower demand for domestically produced goods and services might mean lower production of goods and services. The demand effect is stronger if the foreign countries to which we make payments for oil do not trade much with the U.S. That means that the dollars spent on oil do not get recycled to the U.S. economy in the form of foreign purchases of U.S. goods and services.6 A second channel by which a jump in the price of oil can reduce output growth comes from the manner in which energy and capital, such as machines, are used in production. Energy and capital are largely complements in production, which means that to 6 However, dollars may get recycled back to the U.S. economy if petroleum-producing countries purchase U.S. assets. Such purchases could drive down U.S. interest rates and boost consumption and investment. Business Review Q1 2007 23 TABLE Middle East Conflicts and Effects on Oil Supply Date World Supply Disruption Recession Date 10.1% Event Months from Disruption to Cycle Peak Aug. 1957 8 Nov. 1956 Suez Crisis Nov. 1973 Yom Kippur War 7.8% Nov. 1973 0 Nov. 1978 Iranian Revolution 8.9% Jan. 1980 13 Oct. 1980 Iran-Iraq War 7.2% July 1981 8 Aug. 1990 Persian Gulf War 8.8% July 1990 -1 Source: Hamilton (2003). The table lists the major Middle East conflicts since 1950, the amount by which the conflict reduced the world supply of oil, and the number of months to the onset of the nearest U.S. recession. run machines you need energy, and to run machines more intensively takes more energy. If energy becomes more expensive, firms may have to purchase new energy-efficient machines if they want to maintain profit margins. Firms stuck with less fuel-efficient machines see their profit margins suffer, and so they may invest less in capital and labor. Firms may also delay or change their investment plans in response to a rise in the price of oil. These various investment factors slow both demand in the economy as a whole and the economy’s rate of output growth. In addition, oil-price changes might have reallocative effects on the economy. Some sectors use energy more intensively than others. For example, the transportation sector is a heavy user of petroleum products compared with the trade sector. When the price of oil rises, the transportation sector is affected relatively more, leading to flows of capital and labor out of transportation and into other sectors of the economy. This labor and capital reallocation has a short-term negative 24 Q1 2007 Business Review effect on output as unemployed and underemployed resources seek new uses. Empirical studies have attempted to quantify the extent of reallocation in response to changes in the price of oil. Research by Steven Davis and John Haltiwanger found that oil-price increases account for about 20 to 25 percent of the variability of employment growth in the manufacturing sector. Furthermore, firms that had higher capital intensity and higher energy intensity made greater adjustments to their workforces in response to oil-price increases. Oil-price increases may also lead consumers and firms to delay their purchases of certain types of goods. For example, a household may decide that it wants to purchase an SUV. If oil prices jump up, the household might decide to hold off on the purchase until it becomes clearer how long-lasting the oil price increase is likely to be. Similarly, firms may delay investing in certain types of equipment until uncertainty about the future price of oil is somewhat resolved. Thus, whether an oil-price hike is perceived as temporary — lasting only a month or two — or long-lasting can potentially have a significant impact on spending decisions by consumers and businesses. The Asymmetric Effect of OilPrice Changes. How can we pin down the link between oil prices and the macroeconomy? As we saw in Figure 2, some oil-price increases could lead to recessions. What about oilprice decreases? Do they lead to faster output growth? Interestingly enough, the answer is no. A significant feature of the empirical relationship between oil prices and real output is that oil prices have an asymmetric effect on output: Oil-price increases slow output growth, but oil-price decreases do not boost output growth. A possible reason behind this asymmetric effect of oil prices on growth is the interaction of the supply, demand, and reallocation effects. Rising oil prices affect supply because firms now find it more expensive to www.philadelphiafed.org FIGURE 3 Share of Oil Consumption in GDP Percent 5.0 4.5 4.0 3.5 3.0 2.5 2.0 1.5 1.0 0.5 Source: 2004 1999 1994 1989 1984 1979 1974 1969 1964 1959 1954 1949 0 U.S. Energy Information Administration and Haver Analytics. produce goods because of higher energy costs. Demand may be affected as well, since consumers and firms are likely to be uncertain about how long oil prices will remain high and what the implications are for investment and purchases of durable goods. Both of these factors decrease real output. In addition, the reallocation of resources across sectors of the economy in response to higher oil prices slows economic growth. Now consider what happens when oil prices fall. Again, there is a supply effect: Firms now find it cheaper to produce goods, which encourages increased production. Lower oil prices are likely to increase demand as well. But the reallocation effect still slows growth as resources move across sectors in response to lower oil prices. On net, all of these factors may wash out, so that the effect of a decrease in the price of oil is just about nil. This might www.philadelphiafed.org explain the asymmetric effect of oilprice changes on the economy. The asymmetric effect of oil-price shocks on output growth is key to understanding a second prominent feature of the link between oil and the macroeconomy: The relationship between oil-price changes and real output growth is much stronger before 1985 compared with after 1985.7 Before 1985, there is strong statistical evidence that oil-price changes predicted real output growth. After 1985, this relationship breaks down. What has happened? Recall from Figure 2 that before 1980, large changes in the price of oil were upward. If increases in oil prices lead to slower economic growth (asym7 Economists use the term shock to refer to unanticipated changes in economic variables. Examples include unanticipated changes in monetary and fiscal policy, extreme environmental conditions, and events that alter the world price of energy. metry), the pre-1980 data contain many instances of oil-price increases to examine that hypothesis. Indeed, before 1980, there is a strong prediction that oil-price increases lead to slower growth. After 1980, oil-price changes are both positive and negative. If only positive oil-price changes affect economic growth, all of the negative price changes in the data make it more difficult to uncover the effect of oil-price changes on output, since the negative price changes would attenuate the measured effect of the positive price changes. The net result of the asymmetric effect of oil-price increases coupled with lots of oil-price decreases in the data after 1985 would lead to a much weaker measured relationship between oil prices and economic growth. Of course, just because oil-price increases appear to predict slower real output growth does not mean that oilprice increases cause slower real output growth. It could be the case that when the economy is strong and real output growth is high, resulting high demand for oil pushes up the price of oil. (When the economy weakens and demand for oil slows, there is downward pressure on the price of oil.) This type of feedback from the economy to oil prices could end up looking a lot like oil-price increases causing recessions, even though it is really the economy that is driving up oil prices, because oil prices would be high prior to a slowdown in growth. If we want to understand whether oil-price increases actually cause recessions, we have to control for the feedback effect of the economy and demand on oil prices. OIL-PRICE INCREASES CAUSED BY EXTERNAL FACTORS Research by James Hamilton discusses the problem of feedback from the economy to oil prices and poses a solution. If we want to control for the Business Review Q1 2007 25 feedback from real output growth to oil prices, we should identify jumps in the price of oil that are not caused by U.S. economic conditions.8 That is, we want to identify oil-price increases that are external to U.S. economic conditions and then investigate whether these oil-price increases caused by external factors predict slower output growth.9 We can then plausibly argue that since those oil-price increases are not a result of U.S. economic conditions, any relationship between these price increases and slower real output growth is in the direction from oilprice increases to real output growth. This would be a key piece of evidence in arguing that oil-price increases can lead to economic downturns. How can we find external oil-price increases in the data? Recall from the table that Middle East conflicts have historically been associated with oilprice increases. It can reasonably be argued that these conflicts were not an immediate result of U.S. economic conditions. That is, the overall state of the U.S. economy at the time did not influence the unfolding of the Middle East conflicts and the associated rise in oil prices listed in the table. Hamilton has argued that these conflicts can indeed be thought of as external to the U.S. economy, and so they can be used to measure a causal effect of oil-price shocks on output growth. Statistical analysis that examines the effect of these external episodes that led to oil-price increases finds that these episodes do precede economic slowdowns. This evidence argues that oil-price increases cause slower output growth. Of course, there are many more oil-price increases in the data than just the five or so associated with Middle East conflicts. Researchers have used a variety of methods to isolate the important oil-price changes for predicting real output growth. One early method was to use only oil-price increases in statistical analyses and ignore oil-price decreases. However, researchers have subsequently found that more sophisticated measures of oil-price increases have a more stable relationship with real output growth. In particular, a measure of the net increase in oil price is often used. This series is constructed as follows: Compare oil prices in the current period with the highest oil price over a previous period, say, the last 36 months. If the current price is higher than the preceding 36-month peak, calculate the percentage difference between the two. If the current price is lower than the preceding 36month peak, set the series to zero. This measure of net oil-price increases, in effect, says that if the current price of oil is increasing only because it is moving back up to a previous peak (over the last three years), we don’t expect it to have an effect on real output growth. However, if the current price is higher than it has been over the last three years, we can expect an effect on real output growth.10 Figure 4 shows that the net increase in oil prices, measured quarterly, tends to rise significantly before U.S. recessions, and this series does a good job of picking up the price movements associated with the Middle East conflicts reported in the table.11 Note that 8 See the two articles by James Hamilton. Hamilton’s 2003 article contains many references to the literature on quantifying the effect of oil shocks on the U.S. economy. 10 In his 2004 article, Hamilton argues that the net oil-price increase over a 36-month period has good statistical properties and summarizes well a complicated nonlinear link between oil prices and real output growth. 9 By external we mean events that are not caused by U.S. economic conditions. Economists use the term exogenous to describe such external events. 26 Q1 2007 Business Review 11 The quarterly measure of a net oil-price increase is constructed by averaging the monthly net oil-price increase series. this series is quite often zero. In fact, from the early 1950s until the end of 2004, there are about 700 months of data. But in only about 75 of those months is the net oil-price increase positive; the rest of the time it is zero. By this measure, oil shocks are fairly infrequent events. In his 2004 article, Hamilton demonstrates that net oil-price increases basically capture the historical tendency of the U.S. economy to do poorly after the five major Middle East conflicts listed in the table. So far, we have talked about the effect of oil prices on the economy. However, we could alternatively look directly at the quantity of oil produced each year and ask how changes in the world supply of oil affect the economy. Lutz Kilian, in a 2006 working paper, did just that. He examined data on the quantity of oil produced by the world’s suppliers, focusing principally on suppliers from the Middle East. To develop a series of data on the effects of external shocks on oil quantities, Kilian posed the question of what oil production would have been had a country not experienced a conflict such as a war. The difference between the amount of oil a country would have produced had there been no conflict and the quantity that was produced during the conflict gives a measure of supply shortfall that is external to developments in the U.S. economy (Figure 5). The Middle East supply disruptions listed in the table show up as large downward movements in the quantity of oil. As with the net oil-price series, we see that dramatic movements in the series preceded U.S. recessions. In this case, it is a dramatic falloff in the supply of oil. Interestingly, note that there is no sharp falloff in supply that helps explain the dramatic post-1999 increase in the price of oil. This suggests that the latest upward movement in the price of oil is a con- www.philadelphiafed.org sequence of growth in demand for oil outpacing growth in supply. FIGURE 4 Net Oil-Price Increase Percent 0.5 0.45 0.4 0.35 0.3 0.25 0.2 0.15 0.1 0.05 Source: Jan-05 Jan-03 Jan-01 Jan-99 Jan-97 Jan-95 Jan-93 Jan-91 Jan-89 Jan-85 Jan-87 Jan-83 Jan-81 Jan-79 Jan-77 Jan-75 Jan-73 Jan-71 Jan-69 Jan-67 Jan-65 Jan-63 Jan-61 Jan-59 Jan-57 Jan-55 Jan-53 Jan-51 Jan-49 0 Author’s calculations FIGURE 5 Exogenous Oil Supply Shocks Percent 4 2 EMPIRICAL EVIDENCE ON HOW MUCH OIL SHOCKS MATTER The evidence presented so far indicates that oil shocks, in the form of higher oil prices or reduced supplies of the quantity of oil, have a negative effect on U.S. real output growth. How strong is this negative relationship? We can use statistical analysis to estimate how much an increase in the price of oil caused by external factors reduces real output growth. The effect of oil-price shocks caused by external factors on real output growth can be measured by running a statistical analysis (called a regression) of real output growth on lagged real output growth and lags of the net oil-price increase. The estimated regression is described more fully in Quantifying the Effect of OilPrice Shocks. We can estimate this regression and get meaningful results because Hamilton's measure of net oilprice increases has been shown to be a good proxy for changes in oil price caused by external factors.12 Thus, we don’t have to worry so much about feedback from the U.S. economy to the increase in net oil prices when interpreting the results.13 0 12 See, for example, the 2003 paper by James Hamilton and the 2006 paper by Lutz Kilian. -2 -4 -6 Source: Lutz Kilian’s web page at http://www-personal.umich.edu/~lkilian/ www.philadelphiafed.org Jan-02 Jan-03 Jan-04 Jan-98 Jan-99 Jan-00 Jan-01 Jan-97 Jan-91 Jan-92 Jan-93 Jan-94 Jan-95 Jan-96 Jan-88 Jan-89 Jan-90 Jan-85 Jan-86 Jan-87 Jan-81 Jan-82 Jan-83 Jan-84 Jan-78 Jan-79 Jan-80 Jan-73 Jan-74 Jan-75 Jan-76 Jan-77 Jan-71 Jan-72 -8 13 While oil prices spike prior to U.S. recessions, interest rates also spike prior to recessions. Thus, in their study, Ben Bernanke, Mark Gertler, and Mark Watson conjecture that it is really monetary policy responding to oilprice shocks that causes recessions, since their model implies that an alternative policy could have greatly mitigated the effect of oil shocks. However, the article by James Hamilton and Ana Maria Herrera and my article with Sylvain Leduc argue that it is unlikely that alternative monetary policies would have completely avoided recessions in the face of the historical oil shocks that hit the U.S. economy. See also the Business Review article by Sylvain Leduc. Business Review Q1 2007 27 Quantifying the Effects of Oil-Price Shocks he dynamic effect of an exogenous oil shock on real output growth can be analyzed by running a regression of real output growth on its own lags and lags of the oil-shock measure. A key to interpreting the regression is that the oil-shock measure is exogenous, which means it is not itself influenced by real output growth. To measure exogenous oil shocks, we use the measure of net oil-price increases discussed in the text. This measure is calculated as the greater of zero and the percentage difference of the current oil price from its previous 36-month peak. To measure real output growth, we use real GDP. The regression uses quarterly data and is estimated over the period 1948:4 to 2005:4. To capture the dynamics of the relationship, we included four lags of output growth and the net oil-price increase. The regression takes the form: T yt yt-1 yt-2+ yt-3+ yt-4+ ot-1+ ot-2+ ot-3+ ot-4 where yt is quarterly real GDP growth at time t and ot is the net oil-price increase in quarter t. The equation can be estimated by ordinary least squares. The estimated coefficients, t-statistics, and probabilities that the coefficients are zero are: Estimate 0.01 0.25 0.10 -0.10 -0.12 -0.02 -0.04 -0.02 -0.04 t-stat 7.54 3.74 1.39 -1.48 -1.90 -0.99 -2.20 -1.18 -2.41 Prob 0.00 0.00 0.16 0.13 0.06 0.32 0.03 0.23 0.01 The regression results indicate that the coefficients on the net oil-price increase are negative and statistically significant at lags two and four and that the maximal impact of the oil shock occurs at lag four (when using three decimal places). We can test whether the oil shocks have joint significance in the regression, which is a test of whether, statistically, we get just as good a fit if the oil shocks are dropped. When we test that hypothesis, it is strongly rejected, which means that oil-price increases have predictive power for real GDP growth. A similar regression of headline CPI inflation on the net oil-price increase gives the following estimates: Estimate 0.66 0.65 -0.02 0.39 -0.19 2.61 5.09 2.88 -4.64 t-stat 3.22 9.45 -0.37 5.17 -2.90 0.71 1.37 0.77 -1.25 Prob 0.00 0.00 0.71 0.00 0.00 0.48 0.17 0.44 0.21 The coefficient estimates on the net oil-price increase variable are not significantly different from zero, which indicates that oil shocks are not helping to explain the path of inflation. Indeed, a formal statistical test of whether the net oilprice increase variable helps predict CPI inflation finds that it does not. 28 Q1 2007 Business Review www.philadelphiafed.org The analysis indicates that the largest effect of an oil-price shock occurs about four quarters after the shock, indicating that it takes some time for the maximal effect of an oil shock to hit the economy. The implied path of an oil shock on real output growth can be calculated using an impulse response function. This type of graph shows how real output growth responds over time to a one-time increase in the price of oil caused by external factors. More specifically, this type of graph can tell us how much real output growth rises or falls over time in response to a temporary 10 percent increase in the net price of oil that lasts only one period. In our case, we use quarterly data to estimate the regression and generate the impulse response. We can see how much real output growth is affected an arbitrary number of quarters in the future, given a one-quarter increase in the net price of oil today (Figure 6). Figure 6 shows that an increase in the net price of oil leads to a drop in real output growth that gradually increases over time until it reaches a maximum four quarters after the shock hits. After that, the growth rate of real output gradually recovers, so that after about three years, the effect of the oil shock has largely worn off and real output growth is back on its trend path (in the figure, the trend growth rate of real output has been removed). The impulse response function indicates that a 10 percent increase in the price of oil results in real output growth falling 0.55 percent at its maximum impact. This translates into about a 1.4 percent permanent reduction in the level of real output. Even a modest external increase in the net price of oil has a significant impact on real output, according to our analysis. A similar analysis can be performed to investigate the effect of oil-price shocks on CPI inflation. The www.philadelphiafed.org FIGURE 6 Real Output Response to Increase in Oil Prices Real output growth response to 10% increase in net oil price Percent 0.2 0 -0.2 -0.4 -0.6 0 5 15 10 20 25 Quarters After Shock Real output level response to 10% increase in net oil price Percent 0 -0.5 -1 -1.5 -2 0 5 10 15 25 20 30 35 40 Quarters After Shock results from that regression are also reported in the box on page 28. In this case, though, it turns out that oil-price shocks do not have a statistically significant effect on inflation. In the context of our analysis, this means that a jump in oil prices does not help predict the path of future CPI inflation. It appears then that net oil-price increases affect real output growth and not inflation. We can also examine how changes in the quantity of oil caused by external factors affect output growth and inflation using the data series put together by Lutz Kilian. Recall that Kilian developed a series of external shocks to oil quantity that measure supply disruptions in the Middle East. Kilian's analysis indicates that a 10 percent decrease in the oil supply caused by external factors (which is about the magnitude of the disruptions documented in the table) leads to about a 2 percent drop in real GDP growth about five quarters after the shock hits. Kilian also investigated the effect of external oil-supply disruptions on CPI inflation. His analysis indicates that the effect on CPI inflation is negligible. Inflation is up only about 0.75 percent three quarters after the shock hits (which is the maximal impact of oil shocks on inflation). The data on both oil-price shocks and oil-quantity shocks give a similar impression of what happens to Business Review Q1 2007 29 the economy after an oil shock that reduces supply and raises the price of oil. Real output declines steadily for several quarters, reaching a maximum decline about one year after the shock hits. Output then recovers gradually, and after a few years, the shock has largely worn off. Oil shocks appear to have little if any effect on CPI inflation. THE INTERNATIONAL PERSPECTIVE Is the U.S. unique in its response to oil shocks, or do other developed countries display similar behavior? A 2005 working paper by Lutz Kilian examines this question using data on real output growth, inflation, and his series on external oil production shocks. The sample of countries investigated includes the United States, the United Kingdom, Japan, Germany, Italy, Canada, and France.14 Using regressions similar to those reported in Quantifying the Effect of Oil-Price Shocks, Kilian finds a fair degree of similarity in the real output response of G7 countries to negative oil production shocks. A 10 percent external disruption in oil supply typically leads to about a 2 percent reduction in real output growth that occurs between one and two years after the shock hits. The weakest response among the G7 countries is in Japan, but when the data are analyzed on the basis of cumulative inflation and real growth responses, Italy and France also have fared well historically when confronted with oil supply shocks. For inflation, Kilian finds that oil supply disruptions do not lead to sustained inflation in the G7 countries. There is some evidence for stagflation (a simultaneous rise in inflation and 14 These countries are known as the Group of 7, or the G7. 30 Q1 2007 Business Review FIGURE 7 World Oil Demand Barrels / Day (000s) 25,000 Asia & Oceania 20,000 United States Western Europe 15,000 10,000 Eastern Europe & Former U.S.S.R. 5,000 0 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 Source: Energy Information Administration fall in real output growth) for the U.S., the U.K., and Italy. There is no such evidence for stagflation in response to oil shocks for Germany, Japan, and Canada. Thus, the evidence from other developed countries is broadly similar to what we have described for the U.S. Oil shocks caused by external factors that lower supply and raise price do appear to have a negative effect on real output growth. The evidence for oil shocks’ effects on inflation is more varied, but it is largely consistent with the view that oil shocks do not have strong inflation effects. Recent Developments. The principal reason for recent increases in the price of oil is strong world demand as developing countries increase their oil consumption (Figure 7). What is striking about the figure is the recent strength in demand for oil coming from Asian countries. Principally, this demand growth is from China, India, and Indonesia. As these countries become wealthier, their demand for goods such as automobiles is rising, which leads to increased oil consumption. Note as well that U.S. demand has been strong recently as the economy has experienced strong real output growth. Interestingly enough, strong demand for oil from regions of the world such as Asia can, from the perspective of the U.S., look very similar to an oilprice shock. To the extent that trade ties between the U.S. and Asia are weak, strong growth or weak growth in the U.S. may have little effect on economic growth in Asia. Consequently, Asian demand for oil that boosts the worldwide price of oil, and hence the price the U.S. pays for oil, is external to U.S. economic conditions and so may be no different from a conflict in the Middle East that results in a higher price for oil. In the case of the post-1999 oil-price increases, though, the rise in price has been fairly gradual compared with the external price increases we have talked about. Consumers and firms have had time to www.philadelphiafed.org adjust to the price increase and, given the strength of the U.S. economy, certainly part of the oil-price increase has been due to strong U.S. demand for oil. As a consequence, the effect on the economy of the most recent rise in the price of oil may not be as strong as predicted based on the periods of Middle East crises. Nonetheless, we can conduct a back-of-the-envelope simulation using the estimated relationship between real output and net oil-price increases that generated the impulse responses in Figure 6. If we simulate the model using the net oil-price increases that occurred between 2004Q1 through 2006Q2, the prediction is that the level of real GDP (holding all else constant) is currently about 3.2 percent lower than what it would have been had there been no oil shocks over that period. CONCLUSION Historically, the U.S. economy has tended to perform poorly following major disruptions in the supply of oil that coincide with large increases in the price of oil. Typically, these disruptions have been associated with conflicts in the Middle East that significantly affected the world supply of oil. The nature of these conflicts is that they are external to developments in the U.S. economy. Consequently, these episodes provide evidence that oil-price increases may directly cause slower real output growth, both for the U.S. and for the other major industrial countries. The empirical evidence suggests that a 10 percent increase in the price of oil is associated with about a 1.4 percent drop in the level of U.S. real GDP. Interestingly, increases in oil prices have no significant effect on U.S. in- flation, a finding that largely holds up when we look at the major industrial economies. Since 1999, there has been a dramatic increase in the world price of oil. The evidence suggests that this increase is driven not so much by supply disruptions as by strong demand from the U.S., Western Europe, and Asia, especially China, India, and Indonesia. From the perspective of the U.S., some of this price increase is tantamount to an external oil shock. However, because the rise in price has been gradual and has occurred in the face of strong growth, the U.S. economy has not experienced an oil-induced recession. Nevertheless, the evidence suggests that the recent rise in oil prices has worked to restrain domestic output growth. BR Hamilton, James D., and Ana Maria Herrera. “Oil Shocks and Aggregate Macroeconomic Behavior: The Role of Monetary Policy: Comment,” Journal of Money, Credit, and Banking, 36 (2004), pp. 265-86. Leduc, Sylvain. “Oil Prices Strike Back,” Federal Reserve Bank of Philadelphia Business Review (First Quarter, 2002), pp. 21-30. REFERENCES Bernanke, Ben, Mark Gertler, and Mark Watson. “Systematic Monetary Policy and the Effects of Oil Price Shocks,” Brookings Papers on Economic Activity, 1 (1997), pp. 91-142. Davis, Steven, and John Haltiwanger. “The Sectoral Job Creation and Destruction Responses to Oil Price Changes,” Journal of Monetary Economics, 48 (2001), pp. 465-512 Hamilton, James D. “Oil and the Macroeconomy Since World War II,” Journal of Political Economy, 91 (1983), pp. 228-48. Hamilton, James D. (2003), “What Is an Oil Shock?” Journal of Econometrics, 13 (2003), pp. 363-98. www.philadelphiafed.org Kilian, Lutz. “The Effects of Exogenous Oil Supply Shocks on Output and Inflation: Evidence from the G7 Countries,” Working Paper, University of Michigan (2005). Leduc, Sylvain, and Keith Sill. “A Quantitative Analysis of Oil Price Shocks, Systematic Monetary Policy, and Economic Downturns,” Journal of Monetary Economics, 51, (2004), pp. 781-808. Kilian, Lutz. “Exogenous Oil Supply Shocks: How Big Are They and How Much Do They Matter for the U.S. Economy?” Working Paper, University of Michigan (2006). Business Review Q1 2007 31 RESEARCH RAP Abstracts of research papers produced by the economists at the Philadelphia Fed You can find more Research Rap abstracts on our website at: www.philadelphiafed.org/econ/resrap/index. html. Or view our Working Papers at: www.philadelphiafed.org/econ/wps/index.html. AGGLOMERATION ECONOMIES AND THE SPATIAL CONCENTRATION OF EMPLOYMENT This paper seeks to quantify the contribution of agglomeration economies to the spatial concentration of U.S. employment. A spatial macroeconomic model with heterogeneous localities and agglomeration economies is developed and calibrated to U.S. data on the spatial distribution of employment. The model is used to answer the question: By how much would the spatial concentration of employment decline if agglomeration economies were counterfactually suppressed? For the most plausible calibration, the answer is about 48 percent. More generally, the general equilibrium contribution of agglomeration economies appears to be substantial, with empirically defensible calibrations yielding estimates between 40 and 60 percent. Working Paper 06-20, “A Quantitative Assessment of the Role of Agglomeration Economies in the Spatial Concentration of U.S. Employment,” Satyajit Chatterjee, Federal Reserve Bank of Philadelphia HOW DO ENFORCEMENT COSTS AFFECT THE OWN VS. LEASE DECISION? The authors develop a legal contract enforcement theory of the own versus lease decision. The allocation of ownership rights will minimize enforcement costs when the legal system is inefficient. In particular, when legal enforcement of contracts is costly, there will be a shift from arrangements that rely on such enforcement (such as a rental agreement) toward other forms that do not (such as direct ownership). The authors then test this prediction and show that costly enforcement of rental contracts hampers the development of the rental housing market in a cross-section of countries. They argue that this association is not the result of 32 Q1 2007 Business Review reverse causation from a developed rental market to more investor-protective enforcement and is not driven by alternative institutional channels. The results provide supportive evidence on the importance of legal contract enforcement for market development and the optimal allocation of property rights. Working Paper 06-21, “Owning Versus Leasing: Do Courts Matter?,” Pablo Casas-Arce, Universitat Pompeu Fabra, and Albert Saiz, University of Pennsylvania, and Visiting Scholar, Federal Reserve Bank of Philadelphia IMMIGRATION AND NEIGHBORHOOD DYNAMICS What impact does immigration have on neighborhood dynamics? Within metropolitan areas, the authors find that housing values have grown relatively more slowly in neighborhoods of immigrant settlement. They propose three nonexclusive explanations: changes in housing quality, reverse causality, or the hypothesis that natives find immigrant neighbors relatively less attractive (native flight). To instrument for the actual number of new immigrants, the authors deploy a geographic diffusion model that predicts the number of new immigrants in a neighborhood using lagged densities of the foreign-born in surrounding neighborhoods. Subject to the validity of their instruments, the evidence is consistent with a causal interpretation of an impact from growing immigration density to native flight and relatively slower housing price appreciation. Further evidence indicates that these results may be driven more by the demand for residential segregation based on race and education than by foreignness per se. Working Paper 06-22, “Immigration and the Neighborhood,” Albert Saiz, University of Pennsylvania, and Visiting Scholar, Federal Reserve Bank of Philadelphia, and Susan Wachter, University of Pennsylvania www.philadelphiafed.org