Changing Business Cycle Dynamics V2.1

Changing Business Cycle Dynamics John Maynard Keynes famously expressed in his 1923 work, A Tract on Monetary Reform, that "… [The] long run is a misleading guide to current affairs. In the long run, we are all dead. Economists set themselves too easy, too useless a task if in tempestuous seasons they can only tell us that when the storm is long past the ocean is flat again."1 Much of macroeconomics is dedicated to explaining business cycles. This is not surprising, as times of economic volatility affect nearly every segment of society. The motivation for much of Keynes’ later work was clearly to explain one of the most dramatic business cycle fluctuations in recent history, the Great Depression. Fortunately for the U.S. economy, both innovations in the private sector and in monetary policymaking have resulted in a dramatic reduction in business cycle volatility. Since the 1980s, the U.S. macroeconomy has undergone a great transformation. Periods of dramatic boom and bust, which have characterized economic behavior since the Civil War, have largely disappeared. Not only has output growth moderated to a more predictable and comfortable pace, but price behavior has also stabilized. Accompanying this moderation in key macroeconomic variables has been both a change in the structure of the macroeconomy and several innovations in policymaking. Over the last quarter century, the United States has largely shifted from a manufacturing-intensive economy to a predominantly service-based economy, in which new technologies play a large role. Policymakers are also often credited with a better understanding of the tools of monetary policy and their implementation. Additionally, recent advancements in information technology have allowed firms and workers to better adapt to economic conditions, resulting in less pronounced business cycles with more consistent growth and full employment. Even traditional economic relationships between output and inflation have 1

John Maynard Keynes. A Tract on Monetary Reform. London: Macmillan and Company, 1923.

been challenged as the economy has undergone significant changes. Macroeconomists have sparred over the theories explaining the reduction in business cycle volatility, but the impacts of the reduction in output variation are irrefutable and omnipresent. The reduction in U.S. business cycle volatility since 1980 has been a great boon to households, firms, and policymakers as the economic environment has become more certain. Reduced volatility seen in output and prices has several benefits, allowing people to be more consistently employed and reducing resourced devoted to economic planning and hedging against inflation. This paper will examine the roles of structural changes in the economy and monetary policymaking with regard to the recent reduction in business cycle volatility. For the most part of the 20th century, the U.S. macroeconomy performed extremely well, with an average growth rate of 3.5% per year.2 This average rate of annual growth, however, masks the fluctuations around the underlying trend, or natural rate. For much of the 1900s, steady output growth was punctuated by periods of dramatic contraction and expansion, known as the business cycle. In the last thirty years, however, these fluctuations in output have been less pronounced, as contractions and expansions beyond the underlying trend have become more short-lived and less extreme. As documented by macroeconomists Olivier Blanchard and John Simon, the standard deviation of output has declined markedly since 1950. Evidence of the decline in business cycle fluctuation is data on the frequency and length of economic expansions. Blanchard and Simon note that the average length of an expansion during the 1947 to 1981 period was nineteen quarters, compared to an expansion of thirty six quarters for the 1982 to 2000 period. Also accompanying the significant reduction in output volatility has been a moderation in both price variability and the unemployment rate. Essentially, the evidence has

2

See Figure 1: Real GDP Yr. chart xls from 1930 to 2006

shown that gyrations in the price level have become dampened and unemployment has been more consistently at its long-run natural rate. The persistent decline in macroeconomic volatility, known as the “Great Moderation,” is the result of both structural changes in the economy and better monetary policymaking. Structural changes contributing to the decline in volatility include the smoothing of the components of output, innovations in information technology, and the increased sophistication of financial markets. Related to macroeconomic policymaking, better monetary policy, specifically an anchoring of inflation expectations, enhanced central bank credibility, and an understanding of past mistakes, has contributed to the decline in volatility. Together both structural changes and innovations in policymaking have significantly changed the U.S. economic landscape. Economists Olivier Blanchard and John Simon note in their 2001 paper, The Long and Large Decline in U.S. Output Volatility, that the decline in output variation has largely been a result of the behavior of government spending, consumption, and investment activity. 3 They attribute volatility in the early period to erratic fiscal policies during both the Korean and Vietnam wars. Fiscal expansion also affected the volatility of output as spending increased during Johnson’s Great Society program. With respect to consumption and investment activity, Blanchard and Simon attribute the lower volatility to increased competition and liquidity in financial markets. Viewed thought the intertemporal lens, with better access to credit and savings vehicles, firms and consumers face a more linear budget constraint through their initial endowment point. For the representative household, with imperfect credit markets, the interest rate charged on borrowing generally exceeds the yield obtainable for savings, resulting in a kinked budget constraint. Under these conditions, as current income changes, current consumption will increase proportionally, making consumption more volatile than it would be 3

Blanchard 30

with otherwise with consumption smoothing. Increased sophistication and competition in financial markets have decreased the spread between borrowing and lending rates, enabling households and firms to smooth consumption and investment across periods as income and profits change.4 The result is a smoothing of the components of output. Blanchard and Simon explain that there has been a secular decline in output volatility because of the smoothing effect and that recessions will become less frequent in the future. They warn, however, that there is no “New Economy” as was touted in the late 1990s, but that the economy has indeed changed. Some critics dismiss the idea that there has been any secular trend in output volatility, citing the prevalence of adverse supply shocks of the 1970s, when the price of oil reached record highs. According to critics, simply because of good luck, no significant shocks have affected the economy since the 1980s, contributing to the decline in output and price volatility during the recent period.5 Blanchard and Simon respond to this claim by removing the recessions caused by supply shocks from their data. Their findings confirm that even with the severe recessions due to oil price shocks in the 1970s, the behavior of output has still exhibited a decline in volatility.6 Blanchard and Simon reject the random walk hypothesis that the decline in volatility could merely be the result of good luck. While Blanchard and Simon offer an analytically robust explanation of the decline in macroeconomic volatility, their analysis fails to explain it entirely. Specifically, the role of information technology and its productivity enhancing characteristics within the firm is omitted from their analysis.7 While the concept of a “New Economy,” as proclaimed during the exuberance of the 1990s is debatable, there is little doubt that information technology, including productivity applications, telecommunications, and the rise of the internet, has greatly impacted 4

Blanchard, 38. Blanchard, 13. 6 Blanchard, 14. 7 Blanchard piece does not contain either words information or technology 5

productivity. Nobel laureate Robert Solow was ahead of his time when he commented in 1987 that, “We see the computer age everywhere but in the productivity statistics.”8 Between 1973 and 1995 labor productivity grew only at 1.3% per year, compared to 2.5% per year from 1995 to 1999.9 This explosion in productivity growth, not seen since the 1960s, has both challenged the traditional relationship between economic variables and contributed to a distinct smoothing in output. Indeed, the effects of the computer age have been seen in the productivity statistics, as nearly one-third of the growth in labor productivity during the 1990s came from information technology capital.10 Advancements in information technology have been seen not only in productivity and growth statistics, but in the smoothing of macroeconomic variables as discussed earlier. Specifically, inventory and supply chain management, along with the increased dissemination of management technique through information technology have contributed to a rapid decline in macroeconomic volatility. One of the chief explanations for the decline in output volatility is better inventory and supply chain management techniques employed by firms. The widespread adoption of information technology by businesses has impacted both how firms plan for the future and the nature of the production process. Economists Kahn, McConnell, and Perez-Quiros argue in their 2002 paper that inventory behavior has changed significantly since the adoption of new information technologies. The new inventory and production management techniques allowed by these technologies has played a direct role in the moderation of output overall.11 They find that the most of the decrease in volatility that began in the early 1980s can be attributed to a reduction the volatility in the durable goods sector. The 50% decline in durable goods volatility, as measured by the standard deviation, occurred at nearly the same time as aggregate output 8

Robert Solow 1987 Stiroh, Kevin J. What Drives Productivity Growth? FRBNY March 2001 10 Oliner and Sichel: the resurgence of the growth in prod. IT? 11 Kahn, McConnell, Perez-Quiros, 183. 9

variation contracted. They find through econometric modeling techniques that the volatility reduction in the durables sector was large enough to explain 66% of the reduction in aggregate output volatility.12 Furthermore, this reduction in volatility does not arise from merely a reduction in final sales volatility, but from production volatility. Kahn estimates that only 15% of the decline in output volatility in the durables sector can be attributed to a change in sales volatility.13 Volatility seen in sales figures provides an obvious justification for a decline in output volatility, as production is performed to fulfill sales orders. With much of the volatility, however, coming from production, the role of inventory management becomes apparent. Over the last quarter century, both target inventory to sales ratios and deviations from target inventory levels have significantly declined. As calculated by the U.S. Department of Commerce, the inventory to sales ratio for the U.S. economy declined by nearly 16% during the 1992 to 2007 period.14 In his 2002 paper, Kahn decomposes the inventory to sales ratio into trend and transitory components. The trend components is the target inventory to sales ratio, which is driven by the advancement of information technologies. The transitory component, however, is the deviation in the inventory to sales ratio from the desired trend component. He shows that the target inventory to sales ratio has declined markedly the early 1980s. In addition to the decline in the target inventory to sales target ratio, deviations of the transitory component around the target have dampened significantly.15 As the desired amount of inventory held by firms relative to sales has fallen, businesses have also become more adept at managing the inventory they hold and targeting their desired levels. The reduction in the target inventory to sales ratio shows the impact of information technology, specifically powerful computers, productivity applications, and telecommunications. These innovation have revolutionized how 12

Kahn, McConnell, Perez-Quiros, 185. Kahn, McConnell, Perez-Quiros, 186. 14 U.S. Department of Commerce, Inventory to Sales Ratio 15 Kahn, McConnell, Perez-Quiros, 187. 13

firms relate to inventory and manage their supply chains. The result of these innovations has been a reduction in the level of production volatility. To better understand the exact role inventory and supply chain management play in affecting output volatility and the impact of information technologies, I have created a model based on Kahn’s presentation in On the Causes of the Increased Stability of the U.S. Economy. In my model of inventory and production decisions, there are three firms, the Low Information Firm, the High Information Firm, and the Very High Information Firm. Each firm faces the same amount of final sales for each period, which are determined by a random number generator and lie between 50 and 150 units of sales.16 Each firm also must make a production decision at the start of each period, and is unable to alter its decision after it has dedicated itself to a specific amount. This production decision is based partly on what each firm believes will be sales next period and also has an inventory accumulation or exhaustion component. Additionally, all three firms have a target inventory to sales ratio which is determined exogenously. With respect to production, the Low Information Firm bases its production decision solely off of last period’s sales, which is based upon the underlying assumption that next period’s sales are equal to this period’s sales plus a random error term. The fact that the low information firm bases production off last period’s sales is intuitive assuming the stochastic behavior of sales for each period. The other two firms, the High Information Firm and the Very High Information Firm are more advanced in their understanding of next period’s sales. They base their production decisions with perfect information of the next period’s final sales. That means they are able to produce exactly enough to meet sales, but must alter production if they deviate from their target inventory level given their inventory to sales ratios. Inventory to sales ratios for both the Low Information Firm and the High Information Firm are fixed at a level of two, while the Very High Information Firm 16

The first four final sales figues are were determined by myself for ease of calculation and grasping the model.

has a declining inventory to sales ratio to represent the adoption of inventory eliminating information technologies and better supply chain management. Throughout the model, production is used to both fill sales projections and to attain the desired inventory to sales ratio. Inventories are used fulfill sales in the event of a shortfall in production for the Low Information Firm and can be interpreted as a security buffer in the case of misjudgments in sales projections by both the High and Very High Information Firms. Data for each of the firms is can be seen in Figures 1, 2, and 3. The results of a 20 period simulation reveal exactly how new technologies can smooth production and offer striking implications for the role of inventories and new technologies. The results of the simulation show the standard deviations of production for the Low, High, and Very High firms at 176.13, 94.57, and 36.05 respectively. For the Very High Information, the volatility of production declined by a staggering 80% relative to the Low Information Firm. After the adoption of information technology, production volatility declined as theorized. The new technology impacted the firm in two distinct ways. Contrasting the Low Information Firm with the High Information Firm, the only change was that the firm equipped with high information was able to accurately predict future sales, mitigating the need to draw down on inventories in the event of a large spike in sales. Instead in volatility in production came from maintaining desired inventory to sales ratios. While the idea that firms can accurately predict future sales in the next period is a bit extreme, it is not far from reality given econometric modeling techniques coupled with integrated barcode systems and the emerging radio frequency identification (RFID) technology found at the retail and manufacturing level. With the wide availability of accurate data, manufacturers are able to see sales and production needs in real-

time, enabling more accurate forecasts of future sales.17 Besides the benefit of knowing exactly what future sales will be, the Very High Information firm had a declining inventory to sales ratio, reflecting innovation in supply chain management and the use of just-in-time production. Using information technologies, the representative firm is able to reduce the need to hold costly inventories, which not only take up physical space but occupy labor in completely unproductive activities. As telecommunications technologies have improved, coordinating inputs from various suppliers for production has become easier, eliminating the need for inventories and speeding up the production process. As inventories are eliminated, the inventory to sales ratio obviously falls, reducing the need to vary production widely in an effort to obtain target inventory levels. In addition to the inventory channel, advancements in information technology have affected the smoothing of output through the production function. The information technology boom has impacted volatility through the multifactor productivity variable. Consider a simple model of the production process, Y = ZLαK1-α, where Y is real output, Z is a measure of total factor productivity, and L and K are labor and capital respectively. Economists Stephen Oliner and Daniel Sichel note that during the 1974 to 1999 period, total factor productivity growth made up more than one-third of the growth in labor productivity.18 Besides the productivity enhancing characteristics of new technologies, they have likely influenced a moderation in output volatility. As new information technologies, specifically productivity applications, data management software, and telecommunications devices, have become more widespread, shocks to total factor productivity, the Z variable, have become smaller. With better data and communications devices, management techniques and the best practices within firms have become more consistent and 17

Erik Brynjolfsson and Lorin Hitt discuss the impacts of computers and productivity applications at great length and specificity in their 2000 paper Beyond Computation: Information Technology, Organizational Transformation, and Business Performance. Their granular, but highly enlightening, analysis, however, is beyond the scope of this paper. 18 Stephen Oliner and Daniel Sichel. The Resurgence of Growth in the Late 1990s: Is Information Technology the Story? 2000, JEP

widely shared throughout firms and industries. Interactions between capital and labor have not only become more efficient, but have become more reliably efficient as management technique and know how abounds in the modern information-laden firm. Total factor productivity has also been affected by the decline in inventories relative to sales. With no inventories to manage, labor and capital resources are not distracted from productive activities to manage inventory accumulation and exhaustion. With the inventory shock to total factor productivity lessened or eliminated, output is able to grow at a smoother rate without period interruption. Besides the impact of structural factors which have contributed to the decline in macroeconomic volatility, much research has focused on how monetary policymaking has impacted the recent stability of macroeconomic variables. In contrast to the Great Inflation period of the 1970s, the subsequent Volcker-Greenspan era has been characterized by a marked stabilization in both inflation and output. During much of the 1970s, the U.S. economy was characterized by high inflation and deep recessions. The blame for the increase in inflation during this period falls squarely on the shoulders of an accommodative Fed. The VolckerGreenspan era of policymaking, however, has brought about a distinct smoothing of macroeconomic variables through an enhanced understanding of inflation expectations, implementation of the Taylor principle, and flattening of the Phillips curve. The recent decline in price volatility relative to the pre-Volcker era can be characterized as a decrease in both inflation persistence and an anchoring of inflation expectations. Frederic Mishkin, member of Board of Governors, notes that since the early 1970s, the persistence of inflation has declined markedly. Given an inflationary shock, inflation reverts more quickly to its long run level than it has in the past.19 By regressing inflation on twelve lags of itself, Mishkin notes that the coefficients sum to approximately 0.6 and has declined significantly since 19

Mishkin inflation expectations 2

the 1970s.20 A decline in the sum of lagged coefficients signifies that the impact in inflation only produces a small amount of additional pressure in the future, a boon to policymakers. James Stock and Mark Watson explain a similar phenomenon in their 2007 paper with regard to inflation persistence. They decompose inflation into trend and transitory components, and note that trend inflation has fallen significantly while the transitory component has become less important in determining inflation.21 The anchoring of trend inflation signifies a reduction in inflation expectations. The decline in inflation expectations and actual inflation are mutually reinforcing. As expectations remain anchored actual inflation remains low, fulfilling the low expectations and offering additional credibility to the central bank. Credibility is important because even with a temporary rise in inflation, long-run expectations will be anchored based on the idea that the Fed is competent. While the decline in the variability of prices has been less pronounced than the moderation of output, it is a significant change to the macroeconomic environment that can be attributed to better monetary policymaking. In addition to the taming of inflation expectations, better monetary policy has affected the variability of inflation through an adherence to the Taylor principle. While the principles of sound money were certainly well known during much of the 20th century, data shows that only over the last 25 years have monetary policymakers demonstrated a competence in reducing inflation and its volatility. Specifically, under the chairmanship of Paul Volcker, the Federal Reserve began to tame inflation with an adherence to the Taylor principle. Upon arriving at the Federal Reserve, Chairman Volcker faced double-digit rates of inflation, but his hawkish stance and use of the Taylor principle quickly reduced both actual inflation and inflation expectations. The Taylor principle is that idea that given an increase in the rate of inflation, the nominal Fed

20 21

Mishkin inflation expectations 2 James Stock and Mark Watson. Why Has U.S. Inflation Become Harder to Forecast?

Funds rate must increase by more than the rate of inflation for the interest rate increase to have a real effect. This simple principle, while well know during the Great Inflation, was not adhered to by the Federal Reserve. Economist Richard Clarida explains that the Fed’s response to inflation had changed significantly since the 1970s. In his 2000 study of the behavior of the Federal Reserve, he finds that the Fed’s response to inflationary pressure changed dramatically from the pre-Volcker period to the Volcker-Greenspan period. Using John Taylor’s specification of the Taylor Curve, Clarida constructs a model of the nominal Fed Funds rate using the equilibrium real rate of interest, the deviation of inflation from the implicit target rate, and a measure of the output gap. Clarida finds that the coefficient on the inflation term increased noticeably across the pre-Volcker and Volcker-Greenspan periods. Clarida finds that the coefficient increased from 0.83 to 2.15, both statistically significant.22 Additionally, for the output gap term, Clarida shows that the coefficient changed from 0.27 to 0.93, with the Volcker-Greenspan coefficient marginally significant.23 The increase in the Fed’s responsiveness shows their adherence to the Taylor rule, as increases in inflation are met with increases in the real Fed Funds rate. Also contributing to the moderation in inflationary volatility has been a changing relationship between inflation and changes in output. This relationship is typically expressed through the Phillips curve, which relates inflation to unemployment. The Phillips curve can be expressed as, Π = Πe – β(U-Un) + ν, in which Π represents the rate of inflation, Πe is expected inflation, (U-Un) is the deviation of unemployment from its natural rate, β is the responsiveness of inflation to capacity utilization, and ν is a supply shock. Recent evidence has shown that the coefficient on the unemployment term has declined since the 1980s.24 The finding that inflation is less responsive to changes in capacity utilization has two striking macroeconomic 22

Richard Clarida. Monetary Policy Rules. Richard Clarida. Monetary Policy Rules. 24 Mishkin Inflation Dynamics Page 5 23

implications. First, if the coefficient on the unemployment gap has declined, that means the Phillips curve has flattened, suggesting that changes in resource utilization do not have such a great impact on inflation. The Fed’s hawkish stance on inflation has likely influenced a flattening of the Phillips curve, as price increases have become less frequent and the Fed’s ability to manage demand shocks has been more credible. The result is that people expect inflation to remain contained and deviations in resource utilization do not cause large changes in inflation. As discussed at great length before, output has moderated significantly since the 1980s, but there are still fluctuations in the business cycle. The flattening in the Phillips curve explains also why inflation has moderated over the same time period of output. Not only has the variability of output decreased which would suggest that inflation should have moderated, but inflations response to the business cycle has also dampened. Secondly, the decline in inflationary expectations has reduced the Πe term, shifting the Phillips curve closer to the origin. The fact that inflation expectations can shift a flatter Phillips curve also indicates a potential danger to policymakers and the economy. If inflation expectations were to increase suddenly, policymakers would be forced to sacrifice a large amount of output to bring price pressures back to an acceptable level. Bringing together the two chief macroeconomic variables, output and inflation, is the Taylor curve, which represents the volatility tradeoff between the two variables. The downward sloping Taylor curve shows the different combinations of output and inflation volatility available to policymakers. Given an exogenous supply shock, the Taylor curve is a convenient way of describing the optimal choices policymakers face, as they must choose between keeping prices stable or output at its natural rate. Recent volatility in both prices and output can be explained by an inward shift of the Taylor curve. In his 2004 speech, Ben Bernanke explain, “If monetary

policies during the late 1960s and he 170s were sufficiently far from optimal, the result could be a combination of output volatility and inflation volatility lying well above the efficient frontier define by the Taylor curve.25 For example, during the 1970s, policymakers had an overly optimistic view of the economy’s potential, and sought to exploit a long-run version of the Phillips curve to attain higher output. The result was elevated prices and more volatile output as money growth created Using better data, having a informed idea of where potential output is, and competent policymaking can achieve a lower inflation and output volatility. The recent decline in macroeconomic variability, known as the Great Moderation, has its roots in both structural changes and innovations in monetary policymaking. The components of output have moderated significantly through the development of deeper financial markets, technology, and inventory management. On the monetary side, the use the anchoring of inflation expectations coupled with better policymaking has reduced the volatility of output and kept prices consistently low. The result of the decrease in macroeconomic volatility has been a great boon for society. The moderation in macroeconomic variables has not only affected an understanding of macroeconomics, but has improved the quality of life and business for all people. More consistent incomes, sales, employment, and prices lessened the need for economic planning and the amount of resources required to hedging fluctuations in the business cycle and the impacts of inflation.

25

Bernanke great moderation speech 3.

Correlation Between Inflation and Unemployment

1960-1970 1970-1980 1980-1990 1990-2007

Correlation -0.7868 0.3401 0.0829 0.2115

1960 1970 1980 1990-Present

Std. of Prices Std. of Output 1.54% 2.13% 2.90% 2.73% 3.88% 2.67% 1.11% 1.41%

d

Pe ri o

Pe ri o d

d

Pe ri o

Pe ri o d

Pe ri o d

20

18

16

14

12

10

8

6

4

1995

1991

1987

1983

1979

1975

1971

1967

1963

1959

1955

1951

1947

1943

1939

1935

Sales

Production

1999

4%

Pe ri o d

Pe ri o d

Pe ri o d

Pe ri o d

2

0

Standard Deviation of Output Changes (5YR Period)

200 180 160 140 120 100 80 60 40 20 0

Pe ri o d

Pe rio d

48 19 52 19 56 19 60 19 64 19 68 19 72 19 76 19 80 19 84 19 88 19 92 19 96 20 00 20 04

44

19 19

36

32 40

19

19

19

Figure 1

Natural Log of RGDP

10

9.5

9

8.5

8

7.5

7

6.5

6

Natural Log of RGDP

Figure 2

dFIgue

14%

12%

10%

8%

6%

Very High Information Firm

2%

0%

rio d 0

Pe 1 rio d Pe 2 rio d Pe 3 rio d Pe 4 rio d Pe 5 rio d Pe 6 rio d Pe 7 rio d Pe 8 rio d Pe 9 rio d Pe 10 rio d P e 11 rio d Pe 12 rio d Pe 13 rio d Pe 14 rio d Pe 15 rio d Pe 16 rio d P e 17 rio d Pe 18 rio d Pe 19 rio d 20

-200 rio d

-100

Pe

Pe

Figure 3 Low Information Firm

500

400

300

200

100 Sales Production

0

-300

Figure 4