A new approach to understanding how people behave during an economic crisis and how they prepare themselves in anticipation of discrete policy changes
Changes in economic conditions and policies create challenges for policymakers implementing the changes, for ordinary citizens and businesses who have to cope with the changes, and for researchers attempting to model and explain behavior after-the-fact. The policymakers and ordinary citizens, of course, somehow adjust to periods of change, but researchers face a different situation. Too often, historical data around a period of reform or policy change seem to look “peculiar” and are pushed aside or treated superficially, with the analysis focusing on activity during “normal times.” This is unfortunate, because people behave in surprising and revealing ways during periods of crisis and policy change.
Recently, a wave of literature termed policy switching has attempted to understand that behavior by studying the effects of discrete changes in economic policy and the preparations people make in anticipation of these changes. Combining the entire experience—economic behavior in times of rapid change with behavior during more normal times—gives rise to powerful predictions and insights about patterns of economic organization. One branch of the policy-switching literature concentrates on private adjustments to discrete changes in government policy; another branch concentrates on discrete shifts in private behavior, such as opening a new plant or abandoning a production process. This article will examine private responses to policy switching by the government, with a brief look at switching in the private sector.
Policy switching by government
To introduce the idea of policy switching, let us take a hypothetical example. Suppose that during an election year two political parties, the Liberals and the Conservatives, differ on the need to impose a one-time capital levy—a tax on capital already in place; the Liberals favor the tax and the Conservatives oppose it. Public opinion polls give investors information about the parties’ prospects and thus about the probability that a capital levy will be imposed in the future. When the polls show the Conservatives in the lead, investors bid up the price of existing capital; conversely, when the Liberals move ahead, capital prices drop. These price movements, which may be quite substantial, would not reflect changes in the current income stream to capital but would reflect possible changes in future after-tax returns.
To a researcher attempting to explain asset prices using actual income streams, such election-year price movements present a problem—typically, asset prices would be too volatile, based on the variability of current income from the asset. In the past, researchers dealt with similar problems either by dropping the atypical data or by creating a variable designed to account for this behavior. The policy switching way of handling this is to delve into the impact of the prospective policy switch on individuals who are pricing the assets and incorporate explicitly the changing beliefs about future income prospects into the evaluation of asset prices.
In this example, the size of the proposed capital levy and the opinion poll results would be treated as variables fundamental to asset pricing. Movements in these variables, which result in movements in asset prices, would help to reveal structural aspects of the asset-pricing process. This policy-switching approach to understanding the asset market allows the researcher to use periods of possible extraordinary volatility to refine our understanding of the normal working of markets.
We examine below three government policies studied in the recent policy-switching literature that involve discrete shifts in some aspect of policy: (1) monetary policy reform at the end of a hyperinflation, (2) an exchange-rate crisis culminating in a devaluation, and (3) establishment of an exchange rate target zone. These cases emphasize the key role of the private sector in pursuing profit opportunities in the face of uncertain government policy actions. The pursuit of profit—the basic motivation in most of economics—is the organizing principle of the policy-switching approach. Individuals who most accurately anticipate government actions will profit and market prices will reflect those anticipations.
Monetary policy reform. Let us consider how policy switching methods can be applied to understand monetary behavior in the final months of the German hyperinflation in the early 1920s. During this period, inflation was out of control, reaching over 500 percent a week in October 1923. In his classic hyperinflation study, “The Monetary Dynamics of Hyperinflation,” Philip Cagan deleted the last four months of the German experience. This omission apparently reflected the fact that those final months had high inflation and relatively high real money holdings, contradicting his theory that real money balances should fall in response to an acceleration of inflation. Discarding such data was unfortunate, as it was the final months that contained the largest variations in data in terms of the most extreme inflation rates. It is, therefore, these final months that may also have revealed economic behavior most clearly.
Cagan’s topic was revisited in 1980 by Peter Garber and this author when we attempted to include the final months of the German hyperinflation to study systematically the entire experience (see Robert Flood and Peter Garber, “An Economic Theory of Monetary Reform,” Journal of Political Economy, Vol. 88, No. 1, February 1980). In the policy change that we modeled, German authorities switched from their policy stance of high-inflation runaway money supply growth to a stable-price money supply policy. In the months just before the reform, the German people knew the inflation had to end soon because it was wrecking their country, but they could not be sure exactly when. Nonetheless, during the final months of the inflation, people thought there was a good chance things would soon be stabilized, and their inflation anticipations incorporated the possibility of a stabilizing reform.
It is reasonable to compare the situation of the German people to the passengers in a speeding car. The passengers may expect the driver to begin braking at any moment, and if they had to forecast the car’s speed over an interval, their forecast would incorporate the effects of the expected braking. The German people expected the inflation brakes to be put on. They realized that the German economy simply could not continue to support the inflation and they began bracing themselves for a panic stop during the last few months of 1923. Naturally, the bracing took place a bit before the panic stop itself, and it was that bracing that gave Cagan so much trouble in his study.
The accuracy of inflation predictions during such periods are crucially important for profit estimates on contracts denominated in nominal terms. Someone who forecasts inflation more accurately than others will profit. Therefore, during a period when monetary reform is likely, the best inflation forecasts will incorporate the possibility of monetary reform.
We proposed and implemented a procedure for how people living through the German hyperinflation might have estimated the chances of monetary reform and how they might have revised their inflation forecasts in light of these estimates. The results of our effort are portrayed in Chart 1. Cagan’s model of hyperinflation is based on a negatively sloping linear relationship between expected inflation and the logarithm of real money balances—money divided by the price level. The dots in the figure represent the data Cagan actually used in his study and the solid straight line is Cagan’s model of money demand. Cagan dropped the hyperinflation’s final four months (August, September, October, and November of 1923), which are shown as asterisks labeled with the month number. The points in the figure marked with an “x” and labeled with a month number represent Cagan’s data points revised by policy-switching considerations. To make these revisions, we developed a theory of when a hyperinflating economy would reform its currency. Our revisions to Cagan’s data adjusted his expected inflation numbers to allow for the possibility that a monetary reform would take place during the next month. Almost all of the adjusted points lie close to the extended portion of Cagan’s fitted line (dashed line in the figure). Most dramatic is the revision of the expected rate of inflation for November 1923. Before accounting for policy switching, this point was far off Cagan’s model of money demand, while after revision it fits well with the other data.
Chart 1Effect of policy switching
Source: Robert Flood and Peter M. Garber, “An Economic Theory of Monetary Reform,” Journal of Political Economy, Vol. 88, No.1, February 1980.
An exchange rate crisis. When a country tries to control its exchange rate, while at the same time following other policies or facing other conditions that are inconsistent with the announced exchange-rate policy, it can precipitate an exchange rate crisis. This situation sets up potential, almost risk-free profits for speculators at the expense of the exchange authority and requires the private sector to rethink its nominal contracting to account for the possibility of an exchange rate crisis.
To illustrate, let us examine French efforts in the early 1980s to peg the French franc to the German deutsche mark (DM). In the days prior to the devaluation of the French franc, which took place on Monday, March 21, 1983, franc interest rates leaped to seldom-seen heights. The two-day annualized franc rate reached over 350 percent a year during the week before the devaluation (see Chart 2)!
Chart 2Accounting for an exchange-rate crisis Devaluation of the French franc, 1983
Source: International Financial Statistics (IFS).
This leap in French interest rates caused a huge interest differential that is consistent with a policy-switching model. The 1983 devaluation of 3.75 percent against the DM, which took place over a two-day weekend, resulted in a yield of 675 percent at an annual rate to someone borrowing francs on the Friday before the devaluation, converting them to DM over the weekend, and repaying the franc loan with depreciated currency.
In 1983, foreign-exchange market signals strongly suggested an impending devaluation as expectation of an exchange rate change intensified, even though no one in the market knew precisely when, or by how much, the franc would be devalued. If foreign exchange traders attached, for example, a 50 percent probability to devaluation over a given weekend, then those traders would insist on an interest rate over the weekend that would at least compensate them for holding a franc position rather than a DM position. In the example, a 50 percent probability of devaluation and the expectation of a 3.75 percent devaluation would require a two-day interest rate in the neighborhood of 325 percent at an annual rate to compensate lenders of francs over the weekend, which is of the same order of magnitude as the interest rates observed.
An exchange rate target zone. In the previous examples, we approached policy switching as a onetime event—once the policy is actually switched, the example ends. Most studies of policy switching during the early 1980s were modeled in this fashion because analysis of repeated switches proved cumbersome. With repeated switches, the modeler has to cope with not just the probabilities of the first policy switch but of all future policy switches as well. This quickly becomes an analytical nightmare.
While studying exchange rate target zones, however, Paul Krugman noticed an elegant way to model simply and explicitly a situation of repeated policy switches. His technical innovation has led to over 100 research papers written in the last few years on the subject of exchange rate target zones. It is perhaps unfortunate that Krugman’s innovation was displayed only in the area of exchange rate economics, for it has wide applicability. Academic exchange rate economics has been a disappointment because of its inability to develop a robust explanation of exchange rates in the modern flexible rate period. Therefore, as we study the Krugman example, our emphasis will be on the lessons learned from the approach to the policy switching problem rather than the specifics of the exchange rate application.
A government seeking to establish a central parity for the value of its currency relative to some other currency or currencies adopts an exchange rate target zone whereby it pledges to maintain the exchange rate within explicit margins. In the Exchange Rate Mechanism of the European Monetary System, for example, most member countries have announced the intention of maintaining the value of their currency within certain upper and lower limits—2.25 percent in either direction—around a central parity.
In a stylized version of a target zone, suppose that the foreign exchange authority intervenes in favor of the currency at the weak edge of the band and intervenes against the currency at the strong edge of the band. According to Krugman, speculators would recognize the foreign exchange authority’s commitment to defend the zone at the zone edges. Knowing the trigger point and direction of official intervention in the foreign exchange market in some circumstances would be a profit opportunity for speculators possessing the information. But with the intervention information generally available, profits are competed away by speculators “lining up” and bidding against each other to accommodate the anticipated intervention at the most advantageous price to the foreign exchange authority. In Krugman’s example, where the size of the intervention at the zone edges is anticipated correctly, the profits to speculators will be bid away entirely. This means that at the zone edges, intervention will have no effect on the exchange rate because the action of the authorities will be undone by the speculators.
We have just developed the now-famous “smooth pasting” result applied to the foreign exchange market. The lesson from the example is that speculators will bid away foreseen risk-free profit opportunities. In more realistic applications, government intervention is foreseen less perfectly and speculators will sometimes profit and sometimes take a loss.
Some microeconomic directions
A parallel literature to that described above, which uses similar ideas, is tailored to describe private decisions that involve discrete changes in variables controlled by the private sector. But such decisions, like opening a new plant or abandoning a production activity are, in some ways, poorly modeled by traditional methods. For reasons not at all apparent in these models, unprofitable lines of production are kept operating far longer than predicted and the opening of new plants delayed longer than expected.
Models that incorporate a fixed cost of starting a new plant or production process have fared much better. These models allow researchers to describe the firm’s investment policy using ideas similar to those developed for options-pricing models. For instance, consider a firm that has paid the fixed cost of entering a production line. Even if this product line is currently unprofitable, it is always possible that it will become profitable again in the future. By continuing to run the unprofitable process, perhaps on a diminished scale, the firm retains the option of restarting the production process later if conditions should improve. Thus, a firm’s decision to shut down a process or a plant involves looking at the current price-cost margin and realizing that shutting down the activity means writing off the value of the restarting-option.
The policy switching approach began as an academic undertaking designed to help investigators studying both microeconomic and macroeconomic data come to grips with the “peculiar” behavior that surrounds crises and other discrete events. The approach has turned out to be particularly worthwhile, allowing economists to combine economic behavior in times of crisis with behavior during more normal times. This permits the development of more reliable descriptions of behavior that may be useful to policymakers considering radical reforms. Studies of policy switching also remind us of the value of careful analysis of the historical record surrounding data. In each of the examples we looked at, correct analysis required a deeper understanding of the situation than would be available to a researcher looking only superficially at the data.
Suggestions for further reading
Thomas Sargent and Neil Wallace introduce many of these ideas in their paper, “The Stability of Monetary Models with Perfect Foresight,” Econometrica, 41 (November 1973). Paul Krugman and Marcus Miller collect some of the early papers on exchange rate target zones in Exchange Rate Targets and Currency Bands (Cambridge University Press, 1992). In Policy Switching, forthcoming (The MIT Press), Peter Garber and Robert Flood cover many of the topics mentioned here and on related areas. Avinash Dixit gives a very readable version of his work on the private-sector parallel to policy switching in his paper “Investment and Hysteresis,” The Journal of Economics Perspectives, Vol. 6 (Winter 1992).