Printer Friendly

Economic forecasting in the private and public sectors.

The similarities and differences between public and private forecasting are highlighted in this article. The advantages and limitations of economic models and judgmental forecasts are reviewed, and a process that incorporates features of both is recommended. Forecasting is also complicated by difficulties in determining where we are at the present time as well as an increasingly elusive and complex economic structure.

OVER THE YEARS, I have been involved with economic forecasting from a variety of perspectives-as producer and consumer, both inside nd outside government. I would like to take this opportunity to reflect on some of the similarities and differences between economic forecasting in the private and public sectors. The broad approaches taken and the conceptual difficulties faced by forecasters are quite similar in both sectors. The principal differences surround the context and focus of the forecasts and the ends that they serve.

Let me begin by highlighting what I see as the chief similarities between private and public sector forecasting. This group is, of course, more aware than most that I address of the opportunities, challenges, and limitations presented by economic forecasting. The same cannot often be said of the constituency served by the forecaster. In both the private and public sectors, a large gap commonly exists between the expectations of consumers of forecasts and the abilities of the forecaster. in some cases the forecaster must overcome considerable skepticism that economic projections are of any value. In other cases, expectations reach far beyond the abilities of the practitioner. In either situation, the clarity with which the forecaster can communicate the key conditioning assumptions and the uncertainties surrounding a forecast can be as important as the predictions themselves.

Whether employed by the government or by private firms, it is vital that forecasters have a clear understanding of what economic events they are attempting to anticipate and over what time periods. Success in this effort requires a thorough knowledge of how the focus of the forecast relates to the objectives of the decisionmaker and reflects the critical features of the economic environment in which he or she must operate. Too often one observes forecasts that seem to focus on a set of economic statistics because they are readily available or the traditional object of analysis, rather than because of their immediate relevance to the decisionmaker The adept forecaster is capable of drawing the distinction.

MODELS VS. JUDGMENTAL FORECASTING

Over the years, one recurring theme in discussions of forecasting - both within and outside government - has been the debate over the relative merits of economic models and so-called "intuitive" or judgmental approaches. This is a distinction with little meaning or practical relevance. With a few exceptions, it is rare to find pure practitioners from either camp. Most of us are involved in some combination of these efforts. To be sure, there is considerable variation in how economists achieve this melding of models and judgment, and in the weights implicitly assigned to each approach. But the mix is almost always present, and this is appropriate.

I would even take the argument a step further and suggest that, in some respects, it is difficult to distinguish models from judgments. At their core, the two approaches can be quite similar, frequently being based on the same economic theories and similar bodies of empirical evidence. Of course, the intuitive forecaster generally does not have a thousand equations ready to execute at a moment's notice. More often, he or she relies on a handful of key economic relationships, with the relative importance of these key relationships shifting as the economic landscape changes. Much the same is true, in a sense, of model-based forecasters. For a given economic episode, usually only a few key equations in an econometric model drive the forecast produced by the model. A skillful forecaster of either persuasion recognizes and exploits the critical economic relationships in play at any point in time.

I do not mean to imply that there are no meaningful differences between the use of economic models and methods that rely primarily on judgment. Both approaches have their particular strengths and weaknesses. Forecasters, regardless of their preferred modus operandi, should be aware of these differences and should be looking for ways to take advantage of the complementarities offered by the approaches.

ADVANTAGES AND LIMITATIONS OF MODELS

Perhaps the greatest advantage of a fully articulated model is that it helps the forecaster keep track of the interrelationships among the primary variables of interest. I have in mind two kinds of relationships. The first type is the simple accounting identity, such as the one that links government budget deficits, the current account balance, and the excess of domestic saving over investment. These identities play a much larger role that is generally recognized. They enforce a common discipline on forecasters that is unrelated to their theoretical predispositions. Regardless of how formal or informal the model, these identities serve as a powerful check on the internal logic of any forecast.

The second type of relationship reflects behavioral interdependencies. These relationships usually are subject to substantial uncertainty and, as a result, tend to be the focus of greater controversy. Taken together, identities and behavioral equations can aid the forecaster in tracing out a sequence of complicated interactions. For example, it would be difficult without a model to quantify the net impact on domestic interest rates of a change in the fiscal deficit, because it may involve simultaneous links among domestic demand, international capital flows, domestic and foreign monetary policy responses, exchange rates, and so on.

Another advantage of the econometric approach, if it is based on appropriate statistical methods, is that it permits the forecaster to assess systematically the historical accuracy of economic relationships, providing information over time on which have been most and least reliable. these historical measures can be used, in turn, to quantify the uncertainty surrounding the forecaster's assessment of the future.

There are limits, however, to the apparent power of the econometric model as a forecasting tool. In spite of significant progress toward accommodating more sophisticated - and we hope more realistic formal models, it is still fair to say that, on the whole, our econometric models are at best very crude approximations of the true economy. the economy we are attempting to model is exceedingly complex, best characterized by continually evolving institutions and economic relationships. The widespread use of addfactors in most model-based projections is the clearest manifestation of the difficulty that our large-scale models have in representing a complicated reality. At this stage in their development, statistical models still require large doses of judgment if they are to be useful to decisionmakers.

Another set of limitations of econometric models might fall under the general label of "model uncertainty." By this I mean simply that we cannot be sure that our characterizations of the fundamental relationships incorporated in our models are accurate representations of the underlying economic processes. For example, econometric models in the 1950s and 1960s did not devote much attention to the determinants of inflation and its role in the course of macroeconomic adjustment. The failure to recognize fully the role of inflation expectations led initially, at least, to the generally poor record of the profession on forecasting inflation in the 1970s.

Another facet of model uncertainty surrounds the standard econometric practice of estimating fixed economic relationships under the assumption that the structure of the economy is unchanging. If the structure of the economy is more like a moving target than a sitting duck, we will rarely accumulate enough observations from any given structure to estimate accurately the parameters for our models. Tests for structural change have been developed, but these tests work best when a reasonable number of observations from both structures have been collected so that a change may be reliably detected. If the change is occurring now, standard statistical tests may not discover it until one, two, or five years from now.

Developments in financial markets provide a prime example of these difficulties. Twenty years ago we did not anticipate the degree to which financial innovation and deregulation would make the prediction of money demand difficult, with its corresponding consequences for defining a monetary aggregate that could be monitored usefully by policymakers. Looking ahead, it seems reasonable to assume that similar events will occur that will alter our understanding of some of the fundamental relationships in the economy.

A final source of uncertainty may be attributed to the functional form of our models. Most models are essentially linear, in part because historical data are not rich enough to distinguish among the myriad nonlinear forms that might be entertained. The linear approximation is convenient, and no doubt reasonably accurate, for many historical periods. However, it seems possible that the linear approximation may break down during critical economic episodes. For example, the gradual expansion and steep contraction of the business cycle may not be represented well by a linear model.

Moreover, the precision of the estimated parameters in our models if often overstated. The large t statistics that are supposed to represent our confidence in parameter estimates can be quite misleading, because they are frequently the product of an extensive "data-mining" process during which hundreds of alternative equations are estimated and discarded. As a consequence of this biased procedure, our confidence that such relationships represent true economic structure, rather than random chance, must be considerably less than that implied by the reported statistics.

ADVANTAGES AND LIMITATIONS OF JUDGMENTAL FORECASTING

Some of the weak points of the intuitive forecasting approach are simply mirror images of the strengths of the model-based approach. For example, in the intuitive approach, it may be difficult, if not impossible, to keep track of the numerous interactions and simultaneities that exist among the variables of interest. Moreover, it can be exceptionally difficult for consumers of these forecasts to identify the critical underlying assumptions and gauge the sensitivity of the forecast to changes in these assumptions.

For the most part, the strengths of intuitive forecasting complement the weaknesses of model-based prediction. The flexibility of the intuitive approach may allow its practitioner to adjust more quickly to shifts in key parameters or to perceived changes in the economic structure. At times of rapid change, such as at business cycle turning points, intuitive forecasters may be able to pick up on and react to the nonlinear response of the economy better than those who are relying solely on conventional econometric models. Moreover, intuitive forecasters may catch important developments early on by recognizing the signals or anomalies in weekly or monthly data as they are received. While some work has been done to formalize this process in statistical models, at present the judgmental forecaster seems to have the edge on this front.

Given the strengths and weaknesses of these approaches, it seems obvious that the best forecasting strategy will incorporate features of both model-based and intuitive forecasting.

Indeed, a healthy mix of the two techniques is used in economic forecasting at the Federal Reserve. Model-based results often provide a useful starting point for framing the overall outlook. They also help us to gauge quickly the likely influence of incoming information on the outlook and to estimate the sensitivity of forecasts to key conditioning assumptions. However, in spite of the usefulness of models, the role of judgment remains substantial. For example, a significant degree of judgment must be used when reconciling results from a variety of formal, econometric equations, all of which have some degree of plausibility as representations of economic behavior. Moreover, incorporating anecdotal evidence, which may reveal important economic changes before they are reflected in any data, can only be accomplished judgmentally. In that regard, the Federal Reserve benefits substantially from the timely information reported by the District Banks from their extensive contacts with businesses within their regions. Given the tremendous quantity of data with which we are faced - much of it of an idiosyncratic nature - and given the changing economic environment and institutions, the Federal Reserve relies heavily on judgment in evaluating economic prospects.

PRIVATE VS. PUBLIC POLICY FORECASTING

As I have suggested, private and public forecasters share many of the same basic concerns and face similar analytical issues regarding forecast methodology. Nonetheless, some important distinctions can be made between the activities of private and public forecasters.

Because a firm's or industry's ultimate measure of success or failure is its profitability, the most valued private forecasters will be those who accurately anticipate factors that influence the bottom line. These include factors that characterize the demand for the firm's product, such as market share, relative prices, and developments in competing markets. They also include components of the firm's cost structure, such as its cost of raising capital, its energy mix and intensity, and conditions in the specific labor markets from which it hires. For the most part, forecasts of the aggregate economy are required as a backdrop for critical industry-specific developments. To be sure, for some industries, such as durable goods, the macro backdrop looms relatively large. However, for many other industries, macroeconomic considerations are dominated by the influences of changing technologies, tastes, and other developments in closely related markets. It will almost always be the case that the private forecaster must perform well on the firm- or industry-specific variables. Thus, it is reasonable that private forecasters devote more resources to forecasting in much greater detail a more narrow set of microeconomic variables than does the economist in the public sector.

The policy forecaster, on the other hand, necessarily focuses on those aspects of the economy that policy, most directly influences. For example, it is generally agreed that monetary policy affects the general price level in the long run, and aggregate output and employment in the short run. These are the variables by which the success of monetary policy most often is judged. Consequently, they are the variables of primary interest to the policy forecaster. Changes in monetary and fiscal policies may alter the relative price of cold rolled sheet steel and the cost of capital for farm machinery producers as well. And, because firm-specific data often provide important clues to the macroeconomic puzzle, the policy forecaster must retain some grasp of industry-specific details in forming his or her macro projections. But understanding all of the microeconomic ramifications of macroeconomic policy is beyond the scope of public-sector forecasters, who must concentrate their resources on the effort to predict aggregate outcomes and the consequences of policy actions.

Let me conclude with a final observation that I believe holds in both the public and private sectors, and whether one emphasizes formal models or more intuitive approaches. Economic forecasting is really the art of identifying tensions or imbalances in the economic process and understanding in what manner they will be resolved over the short to intermediate term. For example, at the microeconomic level, consider the dynamic relationships among production, prices, inventories, and consumption. An unexpected change in consumption creates a tension or imbalance at the firm or in the industry. It may lead to a change in prices, production schedules, or inventories, with corresponding implications for subsequent output. There may be substantial uncertainty about how important each channel will be in resolving the tension, and the exact sequence in which each channel will come to the forefront of the resolution process. But we can be sure that the initial tension in the system will be resolved over time.

A macroeconomic example might be the tensions created when growth in nominal income exceeds the real growth potential of the economy. In the long run, such a discrepancy is reflected in the pace of inflation. But in the short run, the tensions created by outsized nominal income growth can result in changes in real output, changes in inflation, or both. The timing and composition of the responses of production and inflation to this tension are the focus of much macroeconomic attention.

Clearly, detecting key imbalances is a crucial element in the forecast process and is one reason why determining where the economy is at any particular moment is so important in assessing where it may be headed. Much of a forecaster's success in predicting the future clearly depends on how well he or she can determine existing conditions. Given the difficulties we face in determining where we are at present, we should have only modest expectations for our ability to predict the future. While our forecasting tools have improved considerably over the postwar period, our forecast accuracy has not. This observation suggests that we are engaged in a continual struggle in which the benefits of improved techniques are eroded by an increasingly elusive and complex economic structure. Since inevitably the structure will become increasingly more complex in the years ahead, forecasters in both the private and public sectors face a constant challenge to develop more reliable forecast procedures that combine the flexibility of the intuitive approach with the systematic discipline of the model-based approach.

* Alan Greenspan is Chairman of the Board of Governors of the Federal Reserve System, Washington DC. He is a Fellow and former President of NABE. This article was presented at the 32nd Annual Meeting of the National Association of Business Economists, September 23-27, 1990, Washington DC.
COPYRIGHT 1991 The National Association for Business Economists
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1991 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Greenspan, Alan
Publication:Business Economics
Date:Jan 1, 1991
Words:2855
Previous Article:The debt explosion of the 1980s: problem and opportunity.
Next Article:The statistics corner.
Topics:

Terms of use | Copyright © 2016 Farlex, Inc. | Feedback | For webmasters