A History of the Federal Reserve, Volume 2 (99 page)

The new Federal Reserve chairman, Arthur Burns, opposed administration efforts to influence Federal Reserve policy, but he did not hesitate to instruct the administration. He continued to give greater weight to reducing the unemployment rate than to reducing inflation. And he proposed repeatedly that the government adopt wage and price guidelines to keep prices from rising. After repeatedly opposing wage-price policy, President Nixon changed course, imposing price and wage controls in 1971. Later, he eased controls. The efforts, both more and less restrictive, failed. Reported inflation rose especially after controls ended.

Burns and other advocates never made a convincing argument showing why controls of prices would work to control rates of price change. One of their mistakes was to mix control of one or more relative prices with control of the sustained rate of price change. Suppose controls prevented a rise in steel prices. Unless the change somehow reduced total spending and increased saving, buyers would have more money to spend elsewhere. The lower relative price of steel might change a price index, but it would not change the (correctly measured) rate of price change.

By suppressing the rise in the price level, controls encouraged the government to choose expansive policies to reduce unemployment. Unemployment was President Nixon’s main political concern, lest it deny him reelection. Controls were a political success; he was reelected. But inflation rose.

The Federal Reserve sacrificed much of its remaining independence to lower unemployment in time for the 1972 election. Burns met frequently with President Nixon, who urged him repeatedly to increase money growth. Burns agreed, but he could not do it alone. The other members of the Board had been appointed by Presidents Kennedy and Johnson. They had no reason to support President Nixon’s reelection. They acted on the belief that it was more important to reduce the unemployment rate than to control inflation. Like other members of the FOMC and principal members of Congress, they supported inflationary actions based on their beliefs about relative social costs. In some cases, FOMC members believed either that they would reduce inflation later or that inflation would not occur as long as price controls or slack remained. They were misled by their economic analysis and beliefs.

Failure to distinguish one-time price level changes from sustained rates of change became a more serious problem following oil price increases in 1973 and after. The Federal Reserve and many others did not separate the two reasons for higher prices. Both were taken as evidence of inflation. There was a difference, however. The one-time change in oil prices was a change in a relative price. Once it passed through the economy, the re
ported rate of price change would fall. If policy, especially monetary policy, remained unchanged, the reported rate of inflation would return to the rate preceding the oil price increase. The Federal Reserve’s actions and public statements would work to prevent the public from anticipating a persistent increase in the rate of price change.

Slowing money growth in response to the oil price increase reduced aggregate demand in the mistaken belief that lower aggregate demand could offset the effect of reduced aggregate supply. Since the first effect of reduced money growth is on output, the unemployment rate rose to the highest level since the 1937–38 depression. Predictably, the Federal Reserve responded to increased unemployment by increasing money growth, thereby converting some of the one-time price level change into persistent inflation.

This was an analytic error. The oil price increase reduced wealth. The wealth loss had to be borne by reductions in real wages and profits. By mistaking the wealth loss for a recession, the Federal Reserve postponed the loss and increased inflation

Economic theory does not give a firm answer to the question, Should the central bank stabilize the price level or the rate of price change? The former gives the public an opportunity to plan on stable prices, a valuable contribution to those planning investments and allocating wealth over time. This benefit carries a cost. It requires the central bank to reverse any increase or decline in the price level. The classical gold standard produced such a result. Over long periods, the price level remained approximately constant. The costs of achieving stable prices under the gold standard proved higher than the modern public is willing to bear. In democratic countries especially, the public expected the state to respond to unemployment. We do not have the gold standard not because we do not know about the gold standard; it is because we know about its costs.

Inflation control, unlike price level control, does not promise to keep the price level stable. The price level becomes a random walk around the constant (perhaps zero) average inflation rate. Positive oil shocks cause the reported price level to rise and negative oil shocks cause it to fall. Once the shocks pass through, the reported rate of inflation returns to its previous zero or perhaps 2 percent per annum. Currency devaluations or revaluations, excise tax changes, and changes in productivity similarly affect the price level but not the maintained rate of inflation. Sustained changes in the productivity growth rate would change the sustained rate of inflation unless money growth adjusted.

The Federal Reserve is in the money business. It is responsible for growth of the money stock and its consequences. By controlling money
growth it influences growth of aggregate demand. It has no direct influence on aggregate supply, so it can respond to positive aggregate supply shocks only by introducing an unanticipated reduction in aggregate demand, reducing spending to lower prices. Such action complicates unnecessarily the consumer’s problem of determining expected future inflation and adds a monetary disturbance to the one-time reduction in measured output growth from the supply disturbance. Better to let the disturbance pass through. If government response is unavoidable, the supply shock should be seen as an increased tax paid to oil producers; offset it by lowering domestic tax rates.

The Federal Reserve treated the 1973 and 1979 supply shocks as inflationary. In 2006, they allowed the oil price increase to pass through and did not try to force other prices to offset it. They concentrated attention on preventing other prices from rising. This was a step forward. The Federal Reserve should complete the policy change by informing the public that they seek to control the maintained rate of price change, not the price level.

After 1979, “practical monetarism” replaced what remained of Keynesian analysis. In practice, practical monetarism suffered from three principal problems. First, monetarism is a medium- to long-term theory of inflation. Second, short-term money growth rates are relatively variable and difficult to forecast. Third, timing was poor. The Federal Reserve tried practical monetarism just at the time that Congress reduced financial regulation. This made it difficult to estimate how much money growth to permit. In 1982 the Federal Reserve gave up whatever remained of the policy of monetary control.

Paul Volcker recognized that monetary policy could not succeed based on quarterly or semiannual forecasts. At the policy discussion in December 1985 and elsewhere, he explicitly rejected short-term forecasts and fine tuning.
⁷
In contrast, some FOMC members describe policy as “data driven,” suggesting that they change decisions based on noisy current data (Yellen, 2006, 3). That problem continues. Chairman Bernanke claimed that policy action should change only if new data change the outlook, and he has responded to noisy data in practice.

7. See Chapter 9 for the discussion and quotation. Stern and Feldman (2004, 2) reject the short-term focus that dominates policy decisions. “We conclude that current procedures put too much emphasis on short-term countercyclical policy and too little emphasis on longterm inflation control. We argue that these shortcomings could lead to significant monetary policy mistakes.” At the time this is written, May 2009, the FOMC has responded again to market and congressional pressures by giving main weight to recession and unemployment and little weight to longer-term inflation.

The staff continued to use a Phillips curve to forecast inflation. Research has shown that a key input to the forecast, the full employment level or natural rate of unemployment, has not been estimated accurately (Stock and Watson 1999). Orphanides and van Orden (2004) showed that main errors in forecasting inflation resulted from the use of contemporary data. These data differed substantially from the later revised data. Atkeson and Ohanian (2001, 10) concluded that “for the last 15 years, economists have not produced a version of the Phillips curve that makes more accurate inflation forecasts than those from a naïve model.” A principal reason is variability of the natural rate of unemployment or NAIRU.
⁸

The text of Chapters 8 and 9 has many comments by Paul Volcker that praise the staff but dismiss their forecasts as inaccurate and unreliable. Alan Greenspan (2007, 170) also did not find staff inflation forecasts useful. “The ‘natural rate,’ while unambiguous in a model, and useful for historical analyses, has always proved elusive when estimated in real time. The number was continually revised and did not offer a stable platform for inflation forecasting or monetary policy.” I believe economists and central bankers should take seriously this rejection of Phillips curve forecasts by the two most successful chairmen of the Board of Governors.

Orphanides and van Orden (2004) compared alternative inflation forecasts using information available to policymakers at the time of the forecast. They found that using information about unemployment, as in the standard Phillips curve, was less accurate and less reliable than other models they studied.

Current academic researchers and many central bank staffs work within the analytic framework developed in Woodford (2003). Michael Woodford’s work is carefully and elegantly presented.
⁹
It uses the rational expectations paradigm. Two equations determine output and inflation. The central bank sets the only interest rate in the model. All other rates and relative prices depend on the model and rational expectations.

The money stock, bank credit, and other financial variables have no independent role. Setting the interest rate determines the money stock from the demand function. Long-term interest rates depend only on current and expected future short rates. The model cannot analyze the financial failures that were a main concern in 2007.

8. Dennis (2007, 3) has surveyed recent Phillips curves and concluded that they were useful pedagogically but were not useful for practical policymaking. Arthur Okun (1980, 818), who relied on Phillips curve forecasts as a presidential adviser, described the work as “seriously undermined when that empirical generalization collapsed.”

9. The model builds on earlier work by Goodfriend and King (1997) andClarida, Gertler, and Gali (1999).

A central bank that used the Woodford model to set the interest rate would have to judge whether the policy was correct by observing output and inflation. These respond with a lag to interest rate changes so in practice central bankers rely on other measures, including money and credit growth, housing starts, industrial production, employment and unemployment, and many other variables, including anecdotal reports from their districts. They believe that these variables provide relevant, independent information.

Alan Greenspan’s interpretation of wage and profit data to infer a change in the productivity growth rate is a now famous example of data-based judgment improving on and replacing model-based forecasts. Model forecasts at the time called for higher interest rates. Greenspan resisted. Several years passed before the models confirmed the productivity increase.

Further, the Woodford model is not the first to dismiss the role of money. In the history of economics, money has been dismissed many times, only to return. The early Keynesian era and Chairman Martin’s tenure are examples; the nineteenth-century Banking School is another. The high variability of short-period money growth makes money an unreliable indicator at this frequency. Long ago, the Federal Reserve learned that, using the St. Louis equation, quarterly changes in money did not promptly affect the rate of inflation, but persistent changes did. The European Central Bank ignores short-period changes but pays attention to maintained growth rates. The many charts in the text showing the relation of real base growth to real GDP growth before each recession is evidence of a reliable influence but not a constant lead. Researchers at the Bank of England wrote, “Understanding the role of money in the economy has always been an important issue for policymakers. . . . Monetary data can potentially provide important corroborative or incremental information about the outlook for inflation” (Berry et al., 2007, abstract). Earlier, Nelson (2002) showed that the monetary base growth had a significant effect on inflation in the United Kingdom.

A common result of research on the term structure of interest rates rejects the theory that explains long-term rates as dependent only on current short rates and expectations.
¹⁰
Rudebusch, Sack, and
Swanson (2007) studied changes in the term premium. They found that the premium changes by amounts too large to ignore (Federal Reserve Bank of San Francisco, 2004).
¹¹
Litterman and Scheinkman (1991) concluded that changes
in longer-term bond yields reflect three factors: changes in the level of all interest rates, changes in the slope of the yield curve that raise or lower longer-term rates relative to short-term rates, and changes in the curvature of the yield curve. For example, an increase in energy prices that is expected to persist but not increase further raises short-term rates relative to long rates. An increase in anticipated inflation that is expected to persist raises all rates in a parallel shift. To interpret changes in level slope and curvature require more than a single short-term rate.

10. A simple regression of current and lagged changes in the short-term rate on the change in the long-term rate using monthly data explains about 25 percent of the long-rate change in the period 1951–66 and 45 percent for 1967–81.

11. Many studies show a significant relation between recessions and the term structure of interest rates. These studies have two deficiencies. They relate two endogenous variables,
usually an unreliable procedure. And they seek to explain real output using differences in nominal interest rates.