Chapter 11 Classical Business Cycle Analysis: Market-Clearing Macroeconomics

Transcription

1 Chapter 11 Classical Business Cycle Analysis: Market-Clearing Macroeconomics Economists generally agree about the basic business cycle facts outlined in Chapter 8. They know that economic growth is not necessarily smooth and that occasionally there are periods of recession in which output declines and unemployment rises. They know that recessions are typically followed by periods of recovery, in which the economy grows more strongly than normal. And they also know a great deal about how other macroeconomic variables such as productivity, interest rates, and inflation behave during recessions. Recall that recessions and booms in the economy raise two basic questions: (1) What are the underlying economic causes of these business cycles? (2) What, if anything, should government policymakers do about them? Unfortunately, economists agree less about the answers to these two questions than about the basic business cycle facts. The main disagreements about the causes and cures of recessions are between two broad groups of macroeconomists, the classicals and the Keynesians. As discussed first in Chapter 1 and again in Chapter 9, classicals and Keynesians although agreeing on many points differ primarily in their views on how rapidly prices and wages adjust to restore general equilibrium after an economic shock. Classical macroeconomists assume that prices and wages adjust quickly to equate quantities supplied and demanded in each market; as a result, they argue, a market economy is largely self-correcting, with a strong tendency to return to general equilibrium on its own when it is disturbed by an economic shock or a change in public policy. Keynesians usually agree that prices and wages eventually change as needed to clear markets; however, they believe that in the short run price and wage adjustment is likely to be incomplete. That is, in the short run, quantities supplied and demanded need not be equal and the economy may remain out of general equilibrium. Although this difference in views may seem purely theoretical, it has a practical implication: Because Keynesians are skeptical about the economy s ability to reach equilibrium rapidly on its own, they are more inclined than are classicals to recommend that the government act to raise output and employment during recessions and to moderate economic growth during booms. In Chapters 9 and 10, our discussion of the effects of shocks on the economy, and the possible policy responses to those shocks, was framed within an assumption

2 Chapter 11 Classical Business Cycle Analysis: Market-Clearing Macroeconomics 363 that what households and firms expect to see and experience in the future is not affected by what they see and experience today. This is a restrictive assumption. We might expect, for example, that international tensions threatening war in the future may influence trade decisions today. Similarly, news of growing levels of government debt may cause households to expect to pay higher taxes in the future and so increase their savings or their work effort today. In this chapter and the next, we relax that assumption. Taking account of expectations of future events will require a more sophisticated description of how households and firms decide to respond to price changes than we have used to this point. In Chapters 9 and 10 we used a simple description of this decision on the part of firms; for a period of time, firms will passively expand or contract output without changing prices but will eventually respond by increasing (decreasing) prices should output exceed (fall below) full-employment output Y. 1 In this chapter and the next we will introduce a more sophisticated description of macroeconomic price-setting behaviour, one that takes into account how households and firms expect future events to unfold. This description, what economists know as the theory of aggregate supply (AS), will be joined with the theory of aggregate demand (AD) developed in Chapter 9 to give us the model of AD AS, another of the key diagrams of macroeconomic analysis. In this chapter we focus on describing how adherents of the classical model describe business cycles. We begin with a description of real business cycle theory and how it can be used to explain the business cycle facts presented in Chapter 8. A high degree of price flexibility is a hallmark of that model. Confronting the real business cycle model with evidence suggesting that money is non-neutral leads us to a description of the aggregate supply relationship deemed appropriate by classical economists. That model of aggregate supply will then be joined with the model of aggregate demand to provide a summary of the classical approach to understanding the business cycle Business Cycles in the Classical Model We have identified two basic questions of business cycle analysis: What causes business cycles? What can (or should) be done about them? Let s examine the classical answers to these questions, beginning with what causes business cycles. The Real Business Cycle Theory In general, a complete theory of the business cycle must have two components. The first component is a description of the types of shocks or disturbances believed to affect the economy the most. Examples of economic disturbances emphasized by various theories of the business cycle include supply shocks, changes in monetary or fiscal policy, and changes in consumer spending. The second component is a model that describes how key macroeconomic variables, such as output, employment, and prices, respond to economic shocks. The model preferred by classical economists is the market-clearing version of the IS LM FE model or some similar 1 Classical economists judge this period of time to be very short, whereas Keynesians believe it may be considerably longer.

3 364 part III BUSINESS CYCLES AND MACROECONOMIC POLICY framework. However, the issue of which shocks are crucial in driving cyclical fluctuations remains. An influential group of classical macroeconomists developed a theory that takes a strong stand on the sources of shocks that cause cyclical fluctuations. This theory, the real business cycle theory (or RBC theory), argues that real shocks to the economy are the primary cause of business cycles. 2 Real shocks are disturbances to the real side of the economy, such as shocks that affect the production function, the size of the labour force, the real quantity of government purchases, and the spending and saving decisions of consumers. Economists contrast real shocks with nominal shocks, or shocks to money supply or money demand. In terms of the IS LM FE model, real shocks directly affect only the IS curve or the FE line, whereas nominal shocks directly affect only the LM curve. Although many types of real shocks could contribute to the business cycle, RBC theorists give the largest role to production function shocks what we have called supply shocks and what the RBC theorists usually refer to as productivity shocks. Productivity shocks include the development of new products or production methods, the introduction of new management techniques, changes in the quality of capital or labour, changes in the availability of raw materials or energy, unusually good or unusually bad weather, changes in government regulations affecting production, and any other factor affecting productivity. According to RBC theorists, most economic booms result from beneficial productivity shocks, and most recessions are caused by adverse productivity shocks. The Recessionary Impact of an Adverse Productivity Shock Does the RBC theorists idea that adverse productivity shocks lead to recessions (and, similarly, that beneficial productivity shocks lead to booms) make sense? We examined the theoretical effects on the economy of a temporary adverse productivity shock in Chapters 3 and 9. 3 In Chapter 3, we showed that an adverse productivity shock (or supply shock), such as an increase in the price of oil, reduces the marginal product of labour (MPN) and the demand for labour at any real wage. As a result, the equilibrium values of the real wage and employment both fall (see Figure 3.9, p. 72). The equilibrium level of output (the full-employment level of output Y) also falls, both because equilibrium employment declines and because the adverse productivity shock reduces the amount of output that can be produced by any amount of capital and labour. We later used the complete IS LM FE model (see Figure 9.8, p. 269) to explore the general equilibrium effects of a temporary adverse productivity shock. We confirmed our earlier conclusion that an adverse productivity shock lowers the general equilibrium levels of the real wage, employment, and output. In addition, we showed that an adverse productivity shock raises the real interest rate, depresses consumption and investment, and raises the price level. 2 For a more detailed introduction to real business cycles, see Charles Plosser, Understanding Real Business Cycles, Journal of Economic Perspectives, Summer 1989, pp , and Robert G. King and Sergio Rebelo, Resuscitating Real Business Cycles, in J. Taylor and M. Woodford, eds., Handbook of Macroeconomics, Elsevier, The original research is described by Finn E. Kydland and Edward C. Prescott, Time to Build and Aggregate Fluctuations, Econometrica, November 1982, pp , and John B. Long and Charles I. Plosser, Real Business Cycles, Journal of Political Economy, February 1983, pp Kydland and Prescott won the 2004 Nobel Prize in Economics in part for their contributions to the development of real business cycle theory. 3 RBC theorists analyze permanent as well as temporary productivity shocks; we focus on temporary shocks because it is the slightly easier case.

4 Chapter 11 Classical Business Cycle Analysis: Market-Clearing Macroeconomics 365 Broadly, then, our earlier analyses of the effects of an adverse productivity shock support the RBC theorists claim that such shocks are recessionary in that they lead to declines in output. Similar analyses show that a beneficial productivity shock leads to a rise in output (a boom). Note that in the RBC approach, output declines in recessions and rises in booms because the general equilibrium (or fullemployment) level of output has changed and because rapid price adjustment ensures that actual output always equals full-employment output. As classical economists, RBC theorists would reject the Keynesian view (discussed in Chapter 12) that recessions and booms are periods of disequilibrium, during which actual output is below or above its general equilibrium level for a protracted period of time. Real Business Cycle Theory and the Business Cycle Facts Although the RBC theory which combines the classical, or market-clearing, version of the IS LM FE model with the assumption that productivity shocks are the dominant form of economic disturbance is relatively simple, it is consistent with many of the basic business cycle facts. First, under the assumption that the economy is being continuously buffeted by productivity shocks, the RBC approach predicts recurrent fluctuations in aggregate output, which actually occur. Second, the RBC theory correctly predicts that employment will move procyclically that is, in the same direction as output. Third, supporters of RBC theory claim that the prediction of the model suggesting that the real wages will be higher during booms than during recessions (procyclical real wages) is consistent with business cycle facts. However, as we discussed in Chapter 8, definitive conclusions about the cyclicality of real wages remain elusive. A fourth business cycle fact explained by the RBC theory is that average labour productivity is procyclical; that is, output per worker is higher during booms than during recessions. This fact is consistent with the RBC theorists assumption that booms are periods of beneficial productivity shocks, which tend to raise labour productivity, whereas recessions are the results of adverse productivity shocks, which tend to reduce labour productivity. The RBC theorists point out that without productivity shocks allowing the production function to remain stable over time average labour productivity would not be procyclical. With no productivity shocks, the expansion of employment that occurs during booms would tend to reduce average labour productivity because of the principle of diminishing marginal productivity of labour. Similarly, without productivity shocks, recessions would be periods of relatively higher labour productivity instead of lower productivity, as observed. Thus, RBC theorists regard the procyclical nature of average labour productivity as strong evidence supporting their approach. A business cycle fact that does not seem to be consistent with the simple RBC theory is that inflation tends to slow during or immediately after a recession. The theory predicts that an adverse productivity shock will both cause a recession and increase the general price level. Thus, according to the RBC approach, periods of recession should also be periods of inflation, contrary to the business cycle fact. Some RBC theorists have responded by taking issue with the conventional view that inflation is procyclical. They recognize that while the standard view that the price level and inflation are procyclical (they increase when output increases) does seem to hold before World War II, and especially during the Great Depression, this was a period when the economy had a different structure and was subject to different types of shocks than the more recent economy. Since World War II, large adverse supply shocks have caused the price level to rise while output fell. Most notably,

5 366 part III BUSINESS CYCLES AND MACROECONOMIC POLICY inflation surged during the recessions that followed the adverse oil price shocks of and , an observation consistent with the RBC theory. The issue of the cyclical behaviour of prices remains controversial, however. 4 Application Calibrating the Business Cycle If we put aside the debate about price level behaviour, the RBC theory can account for some of the business cycle facts, including the procyclical behaviour of employment and productivity. However, real business cycle theorists argue that an adequate theory of the business cycle should be quantitative as well as qualitative. In other words, in addition to predicting generally how key macroeconomic variables move throughout the business cycle, the theory should predict numerically the size of economic fluctuations and the strength of relationships among the variables. To examine the quantitative implications of their theories, RBC theorists developed a method called calibration. The idea is to work out a detailed numerical example of a more general theory. The results are then compared with observations to see whether model and reality broadly agree. The first step in calibration is to write a simple classical model of the economy such as the classical version of the IS LM FE model except that specific functions replace general functions. For example, instead of representing the production function in general terms as Y = AF(K, N), the person doing the calibration uses a specific algebraic form for the production function, such as 5 Y = AK a N 1 - a, where a is a number between 0 and 1. Similarly, specific functions are used to describe the behaviour of consumers and workers. In the second step, the specific functions chosen are made even more specific by expressing them in numerical terms. For example, for a = 0.3, the production function becomes Y = AK 0.3 N 0.7. In the same way, specific numbers are assigned to the functions describing the behaviour of consumers and workers. Where do these numbers come from? Generally, they are not estimated from macroeconomic data but are based on other sources. For example, the numbers assigned to the functions in the model may come from previous studies of the production function or of the saving behaviour of individuals and families. The third step, which must be carried out on a computer, is to find out how the numerically specified model behaves when it is hit by random shocks, such as 4 Another RBC response to this criticism is to note that in reality the money supply is not literally fixed, as we assumed in our analysis of the effects of a productivity shock in Chapter 9. To the extent that the money supply declines in recessions, the tendency for prices to rise will be less. 5 This production function is the Cobb Douglas production function (see Chapter 3). As we noted, although it is relatively simple, it fits data quite well.

6 Chapter 11 Classical Business Cycle Analysis: Market-Clearing Macroeconomics 367 productivity shocks. The shocks are created on the computer with a random number generator, with the size and persistence of the shocks (unlike the numbers assigned to the specific functions) being chosen to fit the actual macroeconomic data. For these shocks, the computer tracks the behaviour of the model over many periods and reports the implied behaviour of key macroeconomic variables, such as output, employment, consumption, and investment. The results are then compared with the behaviour of the actual economy to determine how well the model fits reality. One of the developers of RBC theory, Edward Prescott 6 of the University of Minnesota, performed an early and influential calibration exercise. Prescott used a model similar to the RBC model we present here, the main difference being that our version of the RBC model is essentially a two-period model (the present and the future), and Prescott s model allowed for many periods. The results of Prescott s computer simulations are shown in Figures 11.1 and Figure 11.1 compares the actually observed volatilities of six macroeconomic variables, as calculated from post World War II U.S. data, with the volatilities predicted by Prescott s calibrated RBC model. 7 Prescott set the size of the random productivity shocks in his simulations so that the volatility of GNP in his model would match the actual volatility in U.S. GNP. 8 That choice explains why the actual FIGURE 11.1 Actual versus simulated volatilities of key macroeconomic variables The figure compares the actual volatilities of key macroeconomic variables observed in post World War II U.S. data with the volatilities of the same variables predicted by computer simulations of Edward Prescott s calibrated RBC model. Prescott set the size of the random productivity shocks in his simulations so that the simulated volatility of GNP would match the actually observed volatility of GNP exactly. For these random productivity shocks, the simulated volatilities of the other five macroeconomic variables (with the possible exception of consumption) match the observed volatilities fairly well. Standard deviation (percentage) 6 Actual Simulated using the RBC model Real GNP Consumption Investment Inventory stocks Total hours worked Productivity Variable 6 Theory Ahead of Business Cycle Measurement, Carnegie-Rochester Conference Series on Public Policy, Volume 25, Autumn 1986, pp Reprinted in Quarterly Review, Federal Reserve Bank of Minneapolis, Fall 1986, pp The measure of volatility used is called the standard deviation. The higher the standard deviation, the more volatile the variable being measured. 8 At the time of Prescott s study, the national income and product accounts of the United States focused on GNP rather than GDP, so Prescott also focused on GNP.

7 368 part III BUSINESS CYCLES AND MACROECONOMIC POLICY FIGURE 11.2 Actual versus simulated correlations of key macroeconomic variables with GNP How closely a variable moves with GNP over the business cycle is measured by its correlation with GNP, with higher correlations implying a closer relationship. The figure compares the correlations of key variables with GNP that were actually observed in the post World War II U.S. economy with the correlations predicted by computer simulations of Prescott s calibrated RBC model. Except for productivity, whose predicted correlation with GNP is too high, the simulations predicted correlations of macroeconomic variables with GNP that closely resemble the actual correlations of these variables with GNP. Correlation Real GNP Consumption Investment Inventory stocks Actual Simulated using the RBC model Total hours worked Productivity Variable and simulated volatilities of GNP are equal in Figure But he did nothing to guarantee that the simulation would match the actual volatilities of the other five variables. Note, however, that the simulated and actual volatilities for the other variables in most cases are quite close. Figure 11.2 compares the actual economy with Prescott s calibrated model in another respect: how closely important macroeconomic variables move with GNP over the business cycle. The statistical measure of how closely variables move together is called correlation. If a variable s correlation with GNP is positive, the variable tends to move in the same direction as GNP over the business cycle (that is, the variable is procyclical). A correlation with GNP of 1.0 indicates that the variable s movements track the movements of GNP perfectly (thus, the correlation of GNP with itself is 1.0), and a correlation with GNP of 0 indicates no relationship to GNP. Correlations with GNP between 0 and 1.0 reflect relationships with GNP of intermediate strength. Figure 11.2 shows that Prescott s model generally accounts well for the strength of the relationships between some of the variables and GNP, although the correlation of productivity and GNP predicted by Prescott s model is noticeably larger than the actual correlation. The degree to which relatively simple calibrated RBC models can match the actual data is impressive. In addition, the results of calibration exercises help guide further development of the model. For example, the version of the RBC model discussed here has been modified to improve the match between the actual and predicted correlations of productivity with GNP.

8 Chapter 11 Classical Business Cycle Analysis: Market-Clearing Macroeconomics 369 Are Productivity Shocks the Only Source of Recessions? Although RBC theorists agree, in principle, that many types of real shocks buffet the economy, in practice much of their work rests on the assumption that productivity shocks are the dominant, or even the only, source of recessions. Many economists, including both classicals and Keynesians, have criticized this assumption as being unrealistic. For example, some economists challenged the RBC theorists to identify the specific productivity shocks that they believe caused each of the recessions since World War II. The critics argue that except for the oil price shocks of 1973, 1979, and 1990, and the tech revolution of the late 1990s, historical examples of economywide productivity shocks are virtually nonexistent. An interesting RBC response to this argument is that, in principle, economywide fluctuations could also be caused by the cumulative effects of a series of small productivity shocks. Moreover, the effects of shocks can persist as the economy adjusts to them. To illustrate the point that small shocks can cause large fluctuations, Figure 11.3 shows the results of a computer simulation of productivity shocks and the associated behaviour of output for a simplified RBC model. In this simple RBC model, the change in output from one month to the next has two parts: (1) a fixed part that arises from normal technical progress or from a normal increase in population and employment; and (2) an unpredictable part that reflects a random shock to productivity during the current month. 9 The random, computergenerated productivity shocks are shown at the bottom of Figure 11.3, and the FIGURE 11.3 Small shocks and large cycles A computer simulation of a simple RBC model is used to find the relationship between computergenerated random productivity shocks (shown at the bottom of the figure) and aggregate output (shown in the middle of the figure). Even though all the productivity shocks are small, the simulation produces large cyclical fluctuations in aggregate output. Thus, large productivity shocks are not necessary to generate large cyclical fluctuations. Output productivity shocks Simulated Level of Aggregate Output 4 2 Simulated Productivity Shocks Time (months) 9 Specifically, the model is Y t = Y t e t, where Y t is output in month t, Y t-1 is output in the previous month, and e t is the random productivity shock in month t. The productivity shocks are randomly chosen numbers between 1.0 and 1.0. A similar example is given in Numerical Problem 6 at the end of this chapter.

9 370 part III BUSINESS CYCLES AND MACROECONOMIC POLICY implied behaviour of output is displayed above them. Although none of the individual shocks is large, the cumulative effect of the shocks causes large fluctuations in output that look something like business cycles. Hence business cycles may be the result of productivity shocks, even though identifying specific, large shocks is difficult. Does the Solow Residual Measure Technology Shocks? Because productivity shocks are the primary source of business cycle fluctuations in RBC models, RBC theorists have attempted to measure the size of these shocks. The most common measure of productivity shocks is known as the Solow residual, which is an empirical measure of total factor productivity, A. The Solow residual is named after the originator of modern growth theory, Robert Solow, who used this measure in the 1950s. 10 Recall from Chapter 3 that to measure total factor productivity A, we need data on output, Y, and the inputs of capital, K, and labour, N. In addition, we need to use a specific algebraic form for the production function, such as Y = AK a N 1 a. We then compute the value of the productivity parameter, A, also known as the Solow residual, as Y Solow residual = K a = A. (11.1) 1 - a N The Solow residual is called a residual because it is the part of output that cannot be directly explained by measured capital and labour inputs. When the Solow residual is computed from actual data, using Eq. (11.1), it turns out to be strongly procyclical, rising in economic expansions and falling in recessions. Figure 11.4 shows Canadian values of Solow residuals from 1961 to The shaded areas identify years of recession. Note that the residuals fell during the recessions of , , , and , and increased during economic expansions. This procyclical behaviour is consistent with the premise of RBC theory that cyclical fluctuations in aggregate output are driven largely by productivity shocks. Recently, however, some economists have questioned whether the Solow residual should be interpreted solely as a measure of technology, as RBC proponents tend to do. If changes in the Solow residual reflect only changes in the technologies available to an economy, it should be unrelated to such factors as government purchases or monetary policy that do not directly affect scientific and technological progress (at least in the short run). However, statistical studies reveal that the Solow residual is, in fact, correlated with such factors as government expenditures, suggesting that movements in the Solow residual may also reflect the impacts of other factors. 11 To understand why measured productivity can vary, even if the actual technology used in production does not change, we need to recognize that capital and labour are sometimes used more intensively than at other times and that more intensive use of inputs leads to higher output. For instance, a printing 10 Technical Change and the Aggregate Production Function, Review of Economics and Statistics, 1957, pp In Chapter 6, we described Solow s contributions to growth theory. 11 Changes in technology might well be correlated with changes in government spending on research and development (R&D). However, the Solow residual is also highly correlated with non R&D government spending and with lags too short to be accounted for by the effects of spending on the rate of invention.

10 Chapter 11 Classical Business Cycle Analysis: Market-Clearing Macroeconomics 371 FIGURE 11.4 Canadian Solow residuals, Solow residuals are constructed from annual output, employment, and capital series, with a = 0.3. Calling it total factor productivity, we previously calculated the Solow residual in Table 3.1 (p. 50). The shaded areas identify recession years. Note that the Solow residual is procyclical and leads the cycle. Sources: See Table 3.1. Solow residual Year press used full time contributes more to production than an otherwise identical printing press used half the time. Similarly, workers working fast (for example, restaurant workers during a busy lunch hour) will produce more output and revenue than the same number of workers working more slowly (the same restaurant workers during the afternoon lull). To capture the idea that capital and labour resources can be used more or less intensively at different times, we define the utilization rate of capital, u K, and the utilization rate of labour, u N. The utilization rate of a factor measures the intensity at which it is being used. For example, the utilization rate of capital for the printing press run full time would be twice as high as for the printing press used half the time; similarly, the utilization rate of labour is higher in the restaurant during lunch hour. The actual usage of the capital stock in production, which we call capital services, equals the utilization rate of capital times the stock of capital, or u K K. Capital services are a more accurate measure of the contribution of the capital stock to output than is the level of capital itself because the definition of capital services adjusts for the intensity at which capital is used. Similarly, we define labour services to be the utilization rate of labour times the number of workers (or hours) employed by the firm, or u N N. Thus, the labour services received by an employer are higher when the same number of workers are working hard and fast than when they are working slowly (that is, the utilization of labour is higher). Recognizing that capital services and labour services go into the production of output, we rewrite the production function as Y = AF(u K K, u N N) = A(u K K) a (u N N) 1 - a, (11.2) where we have replaced the capital stock, K, with capital services, u K K, and labour, N, with labour services, u N N. Now we can use the production function in Eq. (11.2)

11 372 part III BUSINESS CYCLES AND MACROECONOMIC POLICY to substitute for Y in Eq. (11.1) to obtain an expression for the Solow residual that incorporates utilization rates for capital and labour: Solow residual = A(u kk) a (u N N) 1 - a K a N 1 - a = Au a 1 - K u a N. (11.3) Equation (11.3) shows that the Solow residual, as conventionally measured, includes not only parameter A (which reflects technology and perhaps other factors affecting productivity) but also utilization rates of capital and labour u K and u N. Thus, even if technology were unchanging, the calculated Solow residual would be procyclical if the utilization rates of capital and labour were procyclical. There is evidence that utilization is procyclical (so that capital and labour are worked harder in boom periods than in economic slumps). For example, Craig Burnside, of the World Bank, and Martin Eichenbaum and Sergio Rebelo, of Northwestern University, studied the cyclical behaviour of capital utilization by using data on the amount of electricity used by producers. 12 Their rationale for using data on electricity is that additional electricity is needed to increase capital utilization, whether the increased utilization is achieved by operating capital for an increased number of hours per day or by increasing the speed at which the capital is operated. This study revealed that electricity used per unit of capital rises in economic upturns, leading the authors to conclude that capital utilization is strongly procyclical. In addition, this study showed that a measure of technology, analogous to the term A in Eq. (11.2), is much less procyclical than is the Solow residual. Measuring the cyclical behaviour of labour utilization is more difficult, but various studies have found evidence that the utilization rate of labour is also procyclical. For example, Jon Fay and James Medoff, 13 of Harvard University, sent questionnaires to large manufacturing enterprises, asking about employment and production during the most recent downturn experienced at each plant. Fay and Medoff found that during a downturn, the average plant surveyed cut production by 31% and cut its total use of blue-collar hours to 23% below the normal level. Plant managers estimated that total hours could have been reduced by an additional 6% of the normal level without further reducing output. Of the 6% of normal hours, about half (3% of normal hours) were typically assigned to various types of useful work, including equipment maintenance and overhaul, painting, cleaning, reworking output, and training. The remaining 3% of normal hours were assigned to make-work and other unproductive activities. These numbers suggest that firms utilize labour less intensively during recessions. Another interesting study by R. Anton Braun and Charles L. Evans, 14 of the Federal Reserve Banks of Minnesota and Chicago, respectively, examined the behaviour of the U.S. Solow residual over the seasons of the year. Using data that had not been adjusted for normal seasonal variation, Braun and Evans found that the measured Solow residual was especially high during the period around Christmas, growing by 16% between the third and fourth quarters of the year and then dropping by 24% between the fourth and first quarters. The most reasonable explanation for this finding, as Braun and Evans pointed out, is that U.S. manufacturers and retailers work particularly hard during the Christmas 12 Capital Utilization and Returns to Scale, in B. Bernanke and J. Rotemberg, eds., NBER Macroeconomics Annual, Labor and Output Over the Business Cycle, American Economic Review, September 1985, pp Seasonal Solow Residuals and Christmas: A Case for Labor Hoarding and Increasing Returns, Journal of Money, Credit, and Banking, August 1998, part 1, pp

12 Chapter 11 Classical Business Cycle Analysis: Market-Clearing Macroeconomics 373 shopping season to meet the surge in demand. In other words, the temporary rise in the Solow residual around Christmas is due, in large part, to higher labour utilization rates rather than to changes in productive technology. The tendency to use workers less intensively in recessions than in expansions has been referred to as labour hoarding. Labour hoarding occurs when, due to the costs of firing and hiring workers, firms retain some workers in a recession that they would otherwise lay off. Firms keep these workers on the payroll to avoid costs of laying off workers and then rehiring them or hiring and training new workers when the economy revives. Hoarded labour either works less hard during the recession (there is less to do) or is put to work doing tasks, such as maintaining equipment, that are not measured as part of the firm s output. When the economy revives, the hoarded labour goes back to working in the normal way. In much the same way, it may not pay the restaurant owner to send her workers home between the lunch and dinner rush hours, with the result that restaurant workers are less productive during the slow afternoon period. This lower rate of productivity during recessions (or during the afternoon slow period in the restaurant) does not reflect changes in the available technology, but only changes in the rate at which firms utilize capital and labour. Hence we should be cautious about interpreting cyclical changes in the Solow residual (equivalently, total factor productivity, A) as solely reflecting changes in technology. In fact, after looking more closely at the data on productivity, some economists have called into question whether technology shocks really lead to procyclical productivity as the RBC theory suggests. Susanto Basu, of the University of Michigan, and John Fernald, of the Federal Reserve Bank of Chicago, 15 find that technology shocks are not procyclical, but rather are not closely related to the cyclical movements of output. Their view of the data is that although a technological improvement helps to produce additional output, when the shock occurs there is normally a period of transition in which the demand for capital and labour changes. This reallocation of resources, along with changes in the utilization rates of capital and labour, means that technological improvements are initially associated with declines in the use of capital and labour because less capital and labour are now needed to produce the same amount of output. Only after time passes, and resources are adjusted, do technological improvements lead to increased output. Although changes in technology or the utilization rates of capital and labour might cause aggregate cyclical fluctuation, history suggests that shocks other than productivity shocks also affect the economy; wars are but one obvious example. Thus, many classical economists favour a broader definition of classical business cycle theory that allows for both productivity and other types of shocks to have an impact on the economy. Because the models they use allow for shocks other than real productivity shocks, the models are not called RBC models but rather dynamic, stochastic, general equilibrium (DSGE) models, as they model behaviour over time (dynamic), allow for shocks to the economy (stochastic), and are based on general equilibrium concepts. The macroeconomic effects of shocks other than productivity shocks can be analyzed with the classical IS LM FE model. Let s use it to examine the effects of a fiscal policy shock. 15 Why Is Productivity Procyclical? Why Do We Care? Working paper , Federal Reserve Bank of Chicago, 2000.

13 374 part III BUSINESS CYCLES AND MACROECONOMIC POLICY Fiscal Policy Shocks in the Classical Model Another type of shock that can be a source of business cycle fluctuations in the classical model is a change in fiscal policy, such as an increase or decrease in real government purchases of goods and services. 16 This claim is seemingly at odds with the classical assumption that prices adjust quickly to ensure output Y is equal to fullemployment output Y. Recalling our discussion of the IS LM FE model in Chapter 9, an increase in government spending shifts the IS curve up and to the right, causing output to expand. Classical economists believe this elicits a speedy response in the form of a higher price level P, which in turn shifts the LM curve up and to the left. In this way general equilibrium is re-established at the original level of full-employment output. Classical economists argue, however, that an increase in government purchases will not only shift the IS curve but may also shift the FE line to the right. If so, an increase in government purchases causes full-employment output Y to increase. The mechanism by which fiscal policy might shift the FE line and so affect real output in the classical model is through the possible effect that a change in government purchases has on wealth. If government increases the share of the nation s output that it takes for public spending, then people may feel their wealth has been reduced. If so, then, as discussed in Chapter 3, the decrease in wealth leads workers to increase their labour supply because someone who is poorer can afford less leisure. 17 As a sketch of the simple labour demand and supply diagram shows, an increase in the supply of labour increases the equilibrium level of employment N and, consequently, the full-employment level of output Y. This result is useful for classical economists because it helps their model fit the data. In particular, it suggests that government purchases are procyclical with real output and government purchases moving in the same direction, something we noted previously is observed in the data. The result also means that, because of diminishing marginal productivity of labour, the increase in employment that accompanies an increase in government purchases results in a decline in average labour productivity. This prediction is particularly important because, as we reported in Figure 11.2 (p. 368), a weakness of the RBC model that considers only productivity shocks is that it predicts that average labour productivity and GNP should be highly correlated while the data suggest they are not. If shocks to government purchases are positively correlated with GNP and negatively correlated with average labour productivity, then a classical model containing both productivity shocks and shocks to government purchases can better match the empirical evidence of GNP and average labour productivity being only weakly correlated. For this to be useful as an explanation for the weak correlation between GNP and the average labour productivity in the data, classical economists must be 16 Another example of a change in fiscal policy is a change in the structure of the system of taxation without any change in government purchases. All economists recognize that different types of taxes affect people s incentives to work, save, and invest differently. A change in the structure of the tax system a switch from income to consumption taxes, for example may therefore affect the position of the FE line and so full-employment output. We have not emphasized this connection in our examination of the business cycle in part because these effects are complicated and depend on the nature of the tax, the type of income or revenue that is taxed, and so on. In addition, fiscal policies intended to dampen the business cycle focus on temporary changes to the total amount of taxes collected (as opposed to changes to the structure of the tax system), and so we have emphasized the impact of temporary tax changes on the aggregate demand (AD) curve. Classical economists accept the Ricardian equivalence proposition, and so would not expect changes in the amount of taxes collected without an accompanying change in government purchases to have much effect on the economy. 17 In Chapter 3 we noted that a change in wealth produces a pure income effect. Just as winning a lottery is expected to increase the consumption of leisure and so reduce the supply of labour, a fall in wealth will result in less leisure being chosen and so a greater supply of labour.

14 Chapter 11 Classical Business Cycle Analysis: Market-Clearing Macroeconomics 375 comfortable with the claim that an increase in government purchases lowers the wealth of individuals and so causes them to increase their supply of labour. This can be a difficult connection to make. For example, if new government spending is on exactly the same goods and services households would have bought had they not funded the new government purchases with new taxes, then clearly there is no impact on household wealth; the exact same goods are purchased by government that would have otherwise been purchased by households. Of course, this is unlikely; we usually ask governments to add our tax payments with those of others to maintain things like a police force, a system of courts, and a road system, things we as individuals desire but may find difficult to build and maintain without collective effort. Examples like these suggest that an increase in government purchases may provide direct benefits to households and firms that increase, rather than decrease, private wealth. Other examples of government spending, however, may not provide the same level of comfort to those believing government purchases provide direct benefits to households and firms. Those who believe significant amounts of government spending are wasteful would argue that increases in government purchases clearly decrease private wealth and so encourage an increase in labour supply. Should Fiscal Policy Be Used to Dampen the Cycle? Our analysis shows that changes in government purchases may have real effects on the economy. Potentially, then, changes in fiscal policy could be used to offset cyclical fluctuations and stabilize output and employment; for example, the government could increase its purchases during recessions. This observation leads to the second of the two questions posed in the introduction to the chapter: Should policymakers use fiscal policy to smooth business cycle fluctuations? Classical economists generally oppose attempts to dampen cyclical fluctuations in the economy. This skepticism about the value of active anti-recessionary policies is rooted in the belief of classical economists that since rapid price responses ensure that markets always clear so that the economy is always in general equilibrium there are no inefficiencies for policymakers to fix. In the classical model, a fall in full-employment output is an optimal response to an adverse productivity shock. While an increase in government purchases may cause fullemployment output to increase, this possibility is associated with a fall in household wealth. In the classical model, then, government s effort to reverse the effects of an adverse productivity shock simply makes people worse off. Classical economists conclude that government purchases should be increased only if the benefits of the expanded government program exceed the cost to taxpayers. They apply this criterion for useful government spending that the benefits should exceed the costs whether the economy is in recession or not. Unemployment in the Classical Model A major weakness of the classical model is that it does not explain why unemployment rises during business downturns. Indeed, in the simple classical, or supplyand-demand, model of the labour market, unemployment is literally zero: Anyone who wants to work can find a job at the market-clearing wage. Of course, in reality, unemployment is never zero. Furthermore, the sharp increases in unemployment that occur during recessions are a principal reason that policymakers and the public are so concerned about economic downturns.

15 376 part III BUSINESS CYCLES AND MACROECONOMIC POLICY Classical economists are perfectly aware of this issue, and they have developed more sophisticated versions of the classical business cycle model to account for unemployment. The main modification they make to the simple supply-and-demand model of the labour market is to drop the model s implicit assumption that all workers and jobs are the same. Rather than all being the same, workers in the real world have different abilities, skills, and interests, among other things; jobs entail different skill requirements, work environments, locations, and other characteristics. Because workers and jobs both vary in so many ways, matching workers to jobs is not instantaneous and free but rather time consuming and costly. The fact that someone who has lost a job or has just entered the labour force must spend time and effort to find a new job helps explain why there are always some unemployed people. Some classical economists suggest that differences among workers and among jobs explain not only why the unemployment rate is always greater than zero but also why unemployment rises so sharply in recessions. They argue that productivity shocks and other macroeconomic disturbances that cause recessions also often increase the degree of mismatch between workers and firms. 18 Thus, a major adverse productivity shock might affect the various industries and regions within a country differently, with jobs being destroyed in some sectors but new opportunities emerging in others. An oil price shock, for example, would eliminate jobs in energy-intensive industries but create new opportunities in industries that supply energy or are light energy users. Following such a shock, workers in industries and regions where labour demand has fallen will be induced to search elsewhere for jobs, which raises the frictional component of unemployment. Some of these workers will find that their skills do not match the requirements of industries with growing labour demand; these workers may become chronically unemployed, raising structural or long-term unemployment. 19 With many unemployed workers looking for jobs, and because creating new jobs takes a while, the time necessary to find a new job is likely to increase. For all these reasons, an adverse productivity shock may raise unemployment as well as reduce output and employment. Note that this predicted rise in frictional and structural unemployment during recessions is the same as an increase in the natural rate of unemployment (the sum of frictional and structural unemployment rates). What is the evidence of worker job mismatch and unemployment? To examine this issue, we start with evidence on just how much potential exists for worker job mismatch. A recent study by the IMF examined job creation and destruction in six sectors of the Canadian economy over the period , 20 concluding that in those six sectors job destruction averaged about 9% of all existing jobs each year, while the annual rate of job creation was equal to about 11% of existing jobs. Thus, on average over this period, the net addition of new jobs was 2% per year. The study also showed that the rates of job creation and destruction vary across provinces. For example, over the period the rate of job destruction in Alberta was about 11% per year while the rate of job creation was almost 14% per year. The comparable figures for Ontario were 9% (job destruction) and 10% (job creation). This evidence suggests that a great deal of churning of jobs and workers occurs in the economy. Much of this churning reflects the closing of old plants and the opening of new ones within the same industry. But it also reflects the growth of some 18 This idea was proposed in David Lilien, Sectoral Shifts and Cyclical Unemployment, Journal of Political Economy, August 1982, pp See Chapter 3 for definitions and discussion of frictional and structural unemployment. 20 See Ravi Balakrishnan, Canadian Firm and Job Dynamics, IMF Working Paper, February 2008.